🎧 Listen Ads-Free: Tired of interruptions? Subscribe to AI Unraveled directly on Apple Podcasts to enjoy all our daily episodes completely ads-FREE at https://djamgamind.com/daily

🚀 Welcome to AI Unraveled. Today, we cut through the PR and look at the forensics. Anthropic leaks a potential zero-day weapon, MIT proves AI isn’t replacing engineers, and Meta open-sources a model that outperforms real human brain scans.

This episode is made possible by our sponsor:

🎙 DjamgaMind: High-Fidelity Intelligence for the C-Suite. If you are a modern decision-maker, DjamgaMind delivers strategic audio forensics in Healthcare, Energy, and Finance. Stop reading headlines and start understanding the systemic impact with our human-verified, technical-grade analysis. 👉 Explore the Forensics: https://DjamgaMind.com/regulations

In Today’s Briefing:

• Claude Mythos Leak: Anthropic’s next-gen model exposed as a potential tool for automated cyber espionage and zero-day discovery.

• The MIT Layoff Study: The forensic proof that 95% of tech layoffs were not caused by AI, and why “Vibe Coding” is failing in production.

• Meta TRIBE v2: A brain encoding model that simulates neural responses across video, audio, and text, outperforming real fMRI recordings.

• Nvidia Nemotron 3 Super: The fastest open-weights model in its class, clocking 442 tokens per second via hardware-software co-design.

• Apple’s Siri Extensions: Opening the iPhone moat to Gemini, Claude, and ChatGPT in iOS 27.

• Quantum Warning 2029: Google’s roadmap to survive “Harvest Now, Decrypt Later” attacks.

• Wikipedia’s AI Ban: Why the world’s knowledge base is holding the human line against “enshittification.”

Strategic Signal: The Shift from Generative Hype to Technical Utility. Credits: Created and produced by Etienne Noumen.

Keywords: Claude Mythos Leak, Anthropic Zero-Day, MIT AI Layoff Study, Meta TRIBE v2, Nvidia Nemotron 3 Super, Google Quantum 2029, Apple iOS 27 Siri, ChatGPT Ad Revenue, Wikipedia AI Ban, DjamgaMind, AI Unraveled.

🔗 RESOURCES & CAREERS

Find AI Jobs (Mercor): Apply Here – https://work.mercor.com/?referralCode=82d5f4e3-e1a3-4064-963f-c197bb2c8db1

⚗️ PRODUCTION NOTE: We Practice What We Preach.

AI Unraveled is produced using a hybrid “Human-in-the-Loop” workflow.

Anthropic just leaked details of its next‑gen AI model Mythos – and it’s raising alarms about cybersecurity

A configuration error exposed ~3,000 internal documents from Anthropic, including draft blog posts about a new model codenamed Claude Mythos. According to the leaked drafts, the model is described as a “step change” in capability, but internal assessments flag it for serious cybersecurity risks:

• Automated discovery of zero‑day vulnerabilities

• Orchestrating multi‑stage cyberattacks

• Operating with greater autonomy than any previous AI

The leak confirms what many have suspected: as AI models get more powerful, they also become more dangerous weapons. Anthropic has previously published reports on AI‑orchestrated cyber espionage, but this time the risk is baked into their own pre‑release model.

ChatGPT hits $100M in ad revenue:

It has only been six weeks since OpenAI began experimenting with showing ads in ChatGPT results. Nonetheless, the company already reports that they’ve hit $100 million in annual ad revenue from the campaign. Most of the users signed up for the Free or Go tiers are eligible for the ad program, but so far, OpenAI reports that just 20% of ChatGPT users have actually enjoyed an ad-supported experience. So there is way more room for this program to grow. Up next: “self-serve” access for advertisers in April, while Canada, Australia, and New Zealand are likely to start seeing ads soon.

Google makes it easier to switch to Gemini:

In a blog post, the company announced a line-up of new Gemini tools allowing users to upload their chat histories and context from other AI apps, like ChatGPT or Claude. The “import” option is available to free and paid users, and Google even provides a recommended prompt to use, so your former chatbot will compile and contribute the most crucial information and context. So for those keeping score, we’re now we’re having the AIs train other AIs to take over their jobs. The betrayal!

Meta’s brain model beats real fMRI scans

Image source: Meta

Meta just open-sourced TRIBE v2, an AI model trained on brain scans from 700+ people that simulates neural activity across vision, hearing, and language — with its synthetic predictions actually outperforming real fMRI recordings.

The details:

• Trained on 1,000+ hours of brain data, v2 leaps from 1,000 brain regions to 70,000, with 700+ subjects up from just 4 volunteers in the original.

• TRIBE v2’s predictions matched population-level brain activity better than most real scans, which often get clouded by heartbeats, movement, and noise.

• The team replicated decades of neuroscience findings in software, correctly pinpointing brain regions for faces, speech, and text with zero scans.

• Meta open-sourced the code, weights, and a live demo, letting any researcher start running virtual brain experiments without building from scratch.

Why it matters: Neuroscience has long required putting people inside expensive scanners for every new experiment, a bottleneck that’s kept entire fields moving one study at a time. TRIBE v2 could do for brain research what AlphaFold did for protein structure: compress months of scanning into seconds of compute.

Apple to unlock Siri for rival AI assistants

Image source: Apple

The Rundown: Apple plans to open up the upcoming Siri revamp for other models starting with iOS 27, according to Bloomberg — ending ChatGPT’s exclusive integration and letting users choose which AI handles their queries directly from the assistant.

The details:

• Users will be able to pick their preferred AI in ‘extensions’ settings and route questions to models of their choice via Siri with the incoming iOS 27.

• ChatGPT is currently the only model compatible with Siri commands via its 2024 deal, but use of that integration has reportedly been ‘minimal’.

• Bloomberg said chatbots in the App Store could also be a revenue stream, with Apple taking a cut of AI subscriptions purchased across its devices.

• Apple is expected to introduce the new Siri AI overhaul powered by Gemini at its WWDC developer event in early June.

Why it matters: Google is already rebuilding Siri’s underlying tech with Gemini, and ChatGPT has had a spot since 2024. Now, Apple is letting the rest of the field in to provide more user choice. It’s a smart move — skip the model war entirely, layer the best AI on top of a billion iPhones, and let its hardware moat do the rest.

OpenAI pauses erotic chatbot plans indefinitely

• OpenAI has paused its plans to launch an erotic “adult mode” for ChatGPT indefinitely, confirming to the Financial Times that it is shifting focus toward its core products instead.

• The company wants more time to research the potentially harmful effects of sexually explicit chats and the emotional attachments they may create, while investors also weren’t excited about the project.

• This is the second major product OpenAI pulled back this week, after discontinuing its Sora AI video-generation app to redirect compute power to other higher-priority projects going forward.

Apple may build smaller AI models from Gemini

• Apple has gained full access to Google’s Gemini model and plans to distill it into smaller models that can run directly on Apple devices without an internet connection.

• The distillation process works by feeding Gemini’s high-quality answers and reasoning information into smaller, cheaper models that learn its internal computations while requiring less computing power.

• Apple is building a smarter, chatbot version of Siri for iOS 27 using Gemini, but has hit issues because Gemini was tuned for chatbot and coding tasks that don’t always match Apple’s needs.

Quantum computers could break encryption by 2029, warns Google

• Google published a formal plan to move all of its infrastructure to post-quantum cryptography by 2029, warning that quantum computers capable of breaking current encryption may arrive sooner than expected.

• The company highlighted “harvest now, decrypt later” attacks as an already active threat, where bad actors steal encrypted data today planning to unlock it once quantum machines become powerful enough.

• Over 6.8 million Bitcoin sitting in vulnerable addresses could eventually be at risk, but Bitcoin developers have started work on quantum-resistant upgrades through BIP 360, a new address format proposal.

Google TurboQuant cuts AI memory use by 6x

• Google Research announced TurboQuant, a new compression algorithm that can reduce AI working memory — known as the KV cache — by at least 6x without losing performance or accuracy.

• The method combines two techniques called PolarQuant and QJL, using vector quantization to clear cache bottlenecks, and the team plans to present their findings at ICLR 2026 next month.

• TurboQuant is still a lab breakthrough and only targets inference memory, not training, so it wouldn’t solve the wider RAM shortages even if successfully deployed in real-world systems.

Reddit to require human verification for suspicious accounts

• Reddit will now force accounts flagged for suspicious behavior to verify they are human, using passkeys, biometrics, and bot labeling as the platform removes around 100,000 automated accounts every day.

• CEO Steve Huffman said passkeys serve as a baseline check but cannot prove individuality, while biometric options like World ID’s iris-scanning and Face ID offer stronger proof but raise privacy concerns.

• Co-founder Alexis Ohanian expressed skepticism about selling face-scanning to Redditors, highlighting tension between the platform’s pseudonymous culture and the technical demands of proving personhood at scale.

Wikipedia bans AI from writing its articles

Image source: Wikipedia

Wikipedia’s volunteer editors banned the use of AI to write articles on the foundation’s English-language site, a move the policy’s author called a “pushback against enshittification and forceful push of AI by so many companies”.

The details:

• Prior attempts at broad AI rules failed to reach consensus, but mounting AI-generated errors pushed editors to a near-unanimous 40-2 vote.

• The ban covers writing or rewriting articles with LLMs, with editors still allowed to use AI for grammar fixes and translations with human review.

• The policy’s author said the change could “spark a broader change” and “empower communities on other platforms” to set AI rules on their own terms.

• StackOverflow and German Wikipedia have enacted similar bans, with Spanish Wikipedia going further to fully ban the use of AI, even for editing purposes.

Why it matters: AI text reportedly surpassed human output for the first time in 2025, and Wikipedia is trying to hold the human line, all while Elon pushes Grokipedia (an AI-created version of Wikipedia) in the exact opposite direction. The internet’s most-used knowledge base bet against the current, but how long that holds is anyone’s guess.

Open-Source Speed Demon

Elevate Your Career with AI & Machine Learning For Dummies PRO
Ready to accelerate your career in the fast-growing fields of AI and machine learning? Our app offers user-friendly tutorials and interactive exercises designed to boost your skills and make you stand out to employers. Whether you're aiming for a promotion or searching for a better job, AI & Machine Learning For Dummies PRO is your gateway to success. Start mastering the technologies shaping the future—download now and take the next step in your professional journey!



Download the AI & Machine Learning For Dummies PRO App:
iOS - Android
Our AI and Machine Learning For Dummies PRO App can help you Ace the following AI and Machine Learning certifications:

• AWS Certified AI Practitioner (AIF-C01): Conquer the AWS Certified AI Practitioner exam with our AI and Machine Learning For Dummies test prep. Master fundamental AI concepts, AWS AI services, and ethical considerations.


• Azure AI Fundamentals: Ace the Azure AI Fundamentals exam with our comprehensive test prep. Learn the basics of AI, Azure AI services, and their applications.


• Google Cloud Professional Machine Learning Engineer: Nail the Google Professional Machine Learning Engineer exam with our expert-designed test prep. Deepen your understanding of ML algorithms, models, and deployment strategies.


• AWS Certified Machine Learning Specialty: Dominate the AWS Certified Machine Learning Specialty exam with our targeted test prep. Master advanced ML techniques, AWS ML services, and practical applications.


• AWS Certified Data Engineer Associate (DEA-C01): Set yourself up for promotion, get a better job or Increase your salary by Acing the AWS DEA-C01 Certification.

Nvidia, the dominant supplier of AI chips, released a competitive open-source large language model whose speed tops its size class — the first open-weights leader to come from the United States since last year, when Meta delivered Llama 4.

What’s new: Nvidia released Nemotron 3 Super 120B-A12B, a large language model designed for agentic applications, including not only weights but also training datasets and recipes. It is the second in a planned family of three: Nvidia released Nemotron 3 Nano-39B-A3B in December 2025, and Nemotron 3 Ultra-500B-A50B is forthcoming.

• Input/output: Text in (up to 1 million tokens), text out (up to 1 million tokens)

• Knowledge cutoff: June 2025 (pretraining data), February 2026 (fine-tuning data)

• Architecture: Hybrid mamba-2/transformer/mixture-of-experts with multi-token prediction layers (120 billion parameters, 12 billion active per token)

• Training data: 25 trillion tokens of curated data scraped from the web and synthesized in 20 natural languages and 43 programming languages

• Features: Tool calling, structured outputs, seven languages (Chinese, English, French, German, Italian, Japanese, Spanish), reasoning modes (off, low, regular)

• Performance: Fastest open-weights model of its size (442 output tokens per second), leads open-weights models on PinchBench test of agentic tasks

• Availability/price: Weights and datasets free to download under a license that permits noncommercial and commercial uses (rights terminate if safety guardrails are removed without replacement or if the user files patent or copyright litigation against Nvidia), free chat via Nvidia and OpenRouter, API around $0.30/$0.80 per 1 million tokens of input/output via third-party providers

How it works: Nemotron 3 Super’s hybrid architecture interleaves mamba-2, attention, and modified MoE layers with multi-token prediction heads that generate a number of tokens per forward pass.

Behind the news: Nvidia plans to invest $26 billion over five years to develop open-weights models — a substantial commitment. The announcement coincides with shifts in the open-weights landscape that could affect Nvidia’s business. Chinese companies, including Alibaba, Moonshot AI, and Z.ai, lately have built the most capable open-weights models, and they are building alternatives to Nvidia GPUs and Cuda software. For instance, DeepSeek has reportedly trained an upcoming model entirely on Huawei’s Ascend chips and Cann software.

Why it matters: Nemotron 3 Super gives developers a fast, fully open model for agentic applications, with training data, recipes, and tools alongside the weights. This openness also serves Nvidia’s business goals. Chinese open-weights models are growing more capable and increasingly streamlined to run on non-Nvidia chips, creating a risk that developers who previously relied on Nvidia will look elsewhere. Nemotron gives them a reason not to.

We’re thinking: Who better to optimize a model for GPUs than the company that designs the GPUs? From custom numerical formats to inference software, Nvidia can co-design hardware and software in ways that few model developers can match. Nvidia is betting that building models will help sell chips and vice versa.

The “AI is replacing software engineers” narrative was a lie. MIT just published the math proving why. And the companies who believed it are now begging their old engineers to come back.

Since 2022, the tech industry has been running a coordinated narrative.

AI will replace 80 to 90% of software engineers. Learning to code is pointless. Developers are obsolete. but what if i tell you that It wasn’t a prediction. It was a headline designed to create fear. And it worked on millions of students and engineers who genuinely believed their careers were over before they started.

It’s 2026 now. Let’s look at what actually happened.

In 2025, 1.17 million tech workers were laid off. Everyone said it was AI. Companies said it was AI. The news said it was AI.

You want to know what percentage of those people actually lost their jobs because AI automated their work?…5%, I’m not lying atp, its literally around 5%, 55k people out of 1.17 million. That’s it.

And according to an MIT study, nearly 95% of companies that adopted AI haven’t seen meaningful productivity gains despite investing millions. The revolution that was supposed to make engineers obsolete couldn’t even pay for itself.

now coming to the main point, So if AI didn’t cause the layoffs, what did?

Source : https://arxiv.org/pdf/2505.10465

What Else Happened in AI on March 27th 2026?

Google rolled out Gemini 3.1 Flash Live, a new voice AI with upgrades in speed, task completion, and realism, to power convos across Search, Gemini Live, and its API.

Mistral released Voxtral TTS, a lightweight voice AI that clones any speaker from a 3-second clip and generates natural-sounding speech across 9 languages.

OpenAI has reportedly shelved its planned erotic chatbot mode indefinitely after pushback from staff and investors.

Novo Nordisk is deploying AI agents across clinical trial ops, with the pharma giant saying the tech is trimming approval timelines and reducing the need for contractors.

Suno launched v5.5 of its AI music generator, adding voice cloning, custom model tuning, and personalized style learning for Pro subscribers.

Cohere released Transcribe, a free open-source speech recognition model that tops HuggingFace’s accuracy leaderboard across 14 languages — taking the No. 1 spot.

Claude AI Maker Anthropic Considers IPO as Soon as October.

Meta Releases TRIBE v2: A Brain Encoding Model That Predicts fMRI Responses Across Video, Audio, and Text Stimuli.

Tencent AI Open Sources Covo-Audio: A 7B Speech Language Model and Inference Pipeline for Real-Time Audio Conversations and Reasoning.

LISTEN DAILY ADS-FREE at:

🎧 Listen Ads-Free: Tired of interruptions? Subscribe to AI Unraveled directly on Apple Podcasts to enjoy all our daily episodes completely ads-FREE at https://djamgamind.com/daily

🚀 Welcome to AI Unraveled. Today, the AI industry gets a massive reality check. A new benchmark proves the top LLMs still cannot reason, forcing the enterprise to pivot toward Small Language Models (SLMs) and inference compression. Meanwhile, Meta quietly buys up every autonomous agent startup on the market, and the Pentagon turns Palantir into permanent military infrastructure.

This episode is made possible by our sponsor:

🎙️ DjamgaMind: High-Fidelity Intelligence for the C-Suite. If you are a modern decision-maker, DjamgaMind delivers strategic audio forensics in Healthcare, Energy, and Finance. Stop reading headlines and start understanding the systemic impact with our human-verified, technical-grade analysis.. 👉 Explore the Forensics: https://DjamgaMind.com/regulations

🛑 AIRIA: As Meta and Sierra deploy agents that can build other agents, enterprise security is failing to keep up. AIRIA provides the centralized control plane and micro-VM sandboxing required to govern autonomous workflows, ensuring your internal agents never become a rogue liability.. 👉 Govern your agents: https://airia.com/request-demo/?utm_source=AI+Unraveled+&utm_medium=Podcast&utm_campaign=Q1+2026

In Today’s Briefing:

• The ARC-AGI-3 Humiliation: Why frontier models from OpenAI and Google are scoring below 1% on a reasoning test that humans ace.

• Google TurboQuant: A new algorithm that compresses AI model memory by over 6x with zero accuracy loss, threatening the AI hardware market.

• The Rise of SLMs: Neurometric launches an SLM Marketplace, proving that 75% of enterprise tasks are better suited for cheap, sub-20B parameter models.

• Meta’s Talent Monopoly: Mark Zuckerberg acqui-hires Dreamer, adding to a $2B shopping spree for agentic AI talent alongside a $135B capex budget.

• The Pentagon’s Palantir Lock-in: The U.S. military is spending $13.4 billion on AI in 2026, elevating Palantir’s Maven to core infrastructure.

• Reddit’s Bot Crackdown: CEO Steve Huffman deploys World ID and passkeys to separate humans from AI agents.

• The EU AI Act Omnibus: Europe delays enterprise AI compliance to 2028, but implements an immediate, strict ban on AI “nudifier” deepfake apps.

• Mistral Voxtral TTS: A new open-weight text-to-speech model that runs entirely on-device and supports 9 languages.

Strategic Signal: The Cap on LLM Reasoning vs. The Optimization of Inference.

Keywords: ARC-AGI-3 Benchmark, François Chollet, Google TurboQuant, Neurometric SLM, Meta Dreamer Acquisition, Palantir Maven Pentagon, Reddit Bot Crackdown World ID, EU AI Act Nudifier Ban, Mistral Voxtral TTS, Small Language Models, DjamgaMind, AIRIA, AI Unraveled.

🔗 RESOURCES & CAREERS

Find AI Jobs (Mercor): Apply Here – https://work.mercor.com/?referralCode=82d5f4e3-e1a3-4064-963f-c197bb2c8db1

⚗️ PRODUCTION NOTE: We Practice What We Preach.

AI Unraveled is produced using a hybrid “Human-in-the-Loop” workflow.

ARC’s new AGI test stumps every frontier AI

Image source: ARC Prize Foundation

François Chollet’s ARC Prize Foundation just released ARC-AGI-3, the newest version of its interactive reasoning benchmark, where humans can solve 100% of tasks on the first try but AI models struggle, with top systems not even scoring 1%.

The details:

• Labs spent millions training models on earlier versions of the test, pushing ARC-AGI-2 scores from 3% to around 50% in under a year.

• Agents face game-like scenarios with zero instructions, and must discover rules, form goals, and plan strategies entirely from scratch.

• Google’s Gemini Pro scored the highest among frontier models at just 0.37%, followed by GPT 5.4 High (0.26%), Opus 4.6 (0.25%), and Grok-4.20 (0%).

• A $1M prize backs the challenge, and cofounder Mike Knoop says frontier labs are paying far more attention to V3 than they did to earlier versions.

Why it matters: It’s always jarring to see the top models get reset below 1% on a new ARC-AGI release, but if the older tests are any indicator, even more surprising will be how quickly frontier labs climb the ladder. Whether that reflects genuine reasoning or just more expensive brute-forcing is exactly what Chollet built V3 to find out.

Reddit’s AI bot crackdown skips the ID check

Image source: Reddit

Reddit CEO Steve Huffman outlined a plan to separate humans from bots across the site, including labeling automated accounts, flagging suspicious users for verification, and letting sub-communities self-police without mass ID checks.

The details:

• Accounts running automation in approved ways on the social platform will carry an [App] label, with suspicious behavior leading to human verification.

• To confirm proof of humanity, Reddit will offer passkeys or Sam Altman’s World ID scanner, with government IDs as a last resort, only where laws require it.

• AI-written content isn’t being banned, with Huffman calling it ‘annoying’ but saying communities can set their own rules on AI-generated posts.

• Rival platform Digg recently folded after being overrun with bots, and Cloudflare data shows automated traffic on pace to surpass humans by 2027.

Why it matters: The Dead Internet Theory was already here before the AI agent acceleration we’ve seen over the past six months. Now, it’s a reality every social media site is dealing with. While this feels a bit like a band-aid, it is a small step towards every platform needing a serious human-first solution if it wants to remain usable to them.

OpenAI pauses erotic chatbot plans indefinitely LINK

• OpenAI has paused its plans to launch an erotic “adult mode” for ChatGPT indefinitely, confirming to the Financial Times that it is shifting focus toward its core products instead.

• The company wants more time to research the potentially harmful effects of sexually explicit chats and the emotional attachments they may create, while investors also weren’t excited about the project.

• This is the second major product OpenAI pulled back this week, after discontinuing its Sora AI video-generation app to redirect compute power to other higher-priority projects going forward.

Apple may build smaller AI models from Gemini LINK

• Apple has gained full access to Google’s Gemini model and plans to distill it into smaller models that can run directly on Apple devices without an internet connection.

• The distillation process works by feeding Gemini’s high-quality answers and reasoning information into smaller, cheaper models that learn its internal computations while requiring less computing power.

• Apple is building a smarter, chatbot version of Siri for iOS 27 using Gemini, but has hit issues because Gemini was tuned for chatbot and coding tasks that don’t always match Apple’s needs.

Quantum computers could break encryption by 2029, warns Google LINK

• Google published a formal plan to move all of its infrastructure to post-quantum cryptography by 2029, warning that quantum computers capable of breaking current encryption may arrive sooner than expected.

• The company highlighted “harvest now, decrypt later” attacks as an already active threat, where bad actors steal encrypted data today planning to unlock it once quantum machines become powerful enough.

• Over 6.8 million Bitcoin sitting in vulnerable addresses could eventually be at risk, but Bitcoin developers have started work on quantum-resistant upgrades through BIP 360, a new address format proposal.

Google shrinks AI memory with zero accuracy loss

Elevate Your Career with AI & Machine Learning For Dummies PRO
Ready to accelerate your career in the fast-growing fields of AI and machine learning? Our app offers user-friendly tutorials and interactive exercises designed to boost your skills and make you stand out to employers. Whether you're aiming for a promotion or searching for a better job, AI & Machine Learning For Dummies PRO is your gateway to success. Start mastering the technologies shaping the future—download now and take the next step in your professional journey!



Download the AI & Machine Learning For Dummies PRO App:
iOS - Android
Our AI and Machine Learning For Dummies PRO App can help you Ace the following AI and Machine Learning certifications:

• AWS Certified AI Practitioner (AIF-C01): Conquer the AWS Certified AI Practitioner exam with our AI and Machine Learning For Dummies test prep. Master fundamental AI concepts, AWS AI services, and ethical considerations.


• Azure AI Fundamentals: Ace the Azure AI Fundamentals exam with our comprehensive test prep. Learn the basics of AI, Azure AI services, and their applications.


• Google Cloud Professional Machine Learning Engineer: Nail the Google Professional Machine Learning Engineer exam with our expert-designed test prep. Deepen your understanding of ML algorithms, models, and deployment strategies.


• AWS Certified Machine Learning Specialty: Dominate the AWS Certified Machine Learning Specialty exam with our targeted test prep. Master advanced ML techniques, AWS ML services, and practical applications.


• AWS Certified Data Engineer Associate (DEA-C01): Set yourself up for promotion, get a better job or Increase your salary by Acing the AWS DEA-C01 Certification.

Image source: Google

Google Research introduced TurboQuant, an algorithm that compresses AI model memory over 6x without any retraining — while delivering up to 8x speed gains on Nvidia H100 chips and losing almost zero accuracy.

The details:

• AI models keep a running log of each conversation, and as chats get longer, that storage balloons, which slows responses and drives up costs.

• TurboQuant shrinks that storage by over 6x with zero accuracy loss, scoring perfectly on tests that bury a key detail in a large amount of text.

• On Nvidia’s top server chips, it also sped up response processing up to 8x compared to standard methods, without adding any extra cost to run.

• The paper, set to be presented at ICLR 2026 in April, also topped rival methods in vector search — the tech search engines use to match similar results quickly.

Why it matters: Despite being first published in April 2025, top AI memory companies felt the heat of the official release, with stocks dropping 3-5%. One compression paper won’t crater memory demand overnight, but the selloff shows Wall Street is pricing in a world where smarter software cuts into the premium AI memory commands.

TOGETHER WITH AIRIA

One-Stop Enterprise AI At Your Fingertips

These days, it seems like there’s a different AI tool for every single business need – sales, marketing, creative, team lunch orders (okay, maybe not that one, but you get the idea). And while that’s all well and good, it can get a bit complicated… not to mention expensive. Which is all the more reason to opt for a platform like Airia that does everything, and does it well.

Airia’s enterprise AI platform gives you all the tools you need to up your productivity and innovation, maintain security and compliance, and manage it all in one simple, seamless hub. That means confusion, less crossed wires, and more “getting the job done” – and isn’t that what AI is supposed to be about in the first place?

Try a demo of Airia’s enterprise AI for yourself right here.

Meta acqui-hires its 4th AI startup in 4 months

Mark Zuckerberg isn’t just building models; he’s assembling an entire talent army for agents. Meta just acqui-hired the team behind Dreamer (an agentic AI platform), adding them to Meta Superintelligence Labs.

• The move follows Meta’s $2B acquisition of autonomous web agent Manus in December, the acqui-hire of AI social network Moltbook in early 2026, and bringing on Scale AI’s Alexandr Wang as Chief AI Officer.

• Meta is concurrently developing an internal AI model (reported codename: “Avocado”) and has projected ~$135 billion in 2026 capital expenditures.

• The pattern shows Meta aggressively cornering the market on agentic talent to build a decentralized “decision-support layer” across its platforms.

Pentagon formalizes Palantir’s Maven AI with $13B investment

The Pentagon is formalizing Palantir’s Maven AI as a core military system with multi-year funding, pushing the platform’s investment umbrella to $13 billion (up from $480 million in 2024). The US Defense Department is spending a staggering $13.4 billion on AI this year alone.

• The jump from “defense contractor” to “core military infrastructure” concerns some civilian tech watchdogs, who note that infrastructure doesn’t get shut down when it hallucinates; it just gets patched.

• Military AI at this funding level is moving faster than civilian accountability frameworks, prioritizing verifiable targeting over explainable decisions.

European Parliament bans AI “nudifier” apps LINK The European Parliament adopted its position on a simplification (”omnibus”) proposal amending the Artificial Intelligence Act. While the EU is delaying the application of certain high-risk AI system rules to 2027/2028 to give companies breathing room, it took immediate action on deepfakes.

• MEPs introduced a strict new ban on “nudifier” systems that use AI to create or manipulate sexually explicit images resembling an identifiable real person without consent.

• The legislation also gives AI providers until November 2026 to comply with rules on mandatory watermarking for AI-created audio, image, video, or text content.

Google upgrades Lyria 3 Pro for full song generation

Google has massively upgraded its music AI model, rolling out Lyria 3 Pro across Gemini, Vertex AI, and Google Vids.

• The model can now generate full 3-minute songs with structured intros, verses, choruses, and bridges from a single prompt.

• The audio quality rivals studio-produced tracks, further threatening the economics of the commercial stock music industry.

Clarivate reveals the “AI50” Global Leaders LINK A new data-driven benchmark by Clarivate reveals that the vast majority of high-impact AI inventions are concentrated in just four regions. Of the top 50 global organizations leading AI invention, 15 are in China, 14 in the US, 6 in South Korea, and 6 in Japan.

Mistral gives hands-free productivity a boost

A month ago, Mistral conquered speech-to-text with its Voxtral Transcribe 2 model family. Now, the AI lab is flipping the script with its first text-to-speech model.

On Thursday, Mistral launched its Voxtral TTS, an open-weight text-to-speech model that is three times smaller than the industry standard, more natural-sounding, and multilingual. The company claims that it’s the best open-source text-to-speech model released to-date.

The model was built to be as natural-sounding as possible by understanding the context of what people say and modeling how people naturally speak, including pauses, rhythm, intonation, emotion and more. Some other highlights include:

• Support for nine languages: English, French, German, Spanish, Dutch, Portuguese, Italian, Hindi, and Arabic. It can accept input in one language and output in another.

• Speed: Adapts to voices with references of just five seconds and offers low-latency, taking a voice sample of 10 seconds and 500 characters, and producing audio within 90 milliseconds

• On-device: 3 billion parameter model that fits in a smartwatch, a phone, or a laptop

• Human evaluations: In comparative human evaluation of Voxtral TTS and ElevenLabs v2.5 Flash, Mistral’s model achieved superior naturalness

While there are many other text-to-speech models on the market, Pierre Stock, VP of science and the first employee hired at Mistral AI, said the model’s overall efficiency sets it apart.

“The model is optimal for both efficiency — including the cost, the speed, the ease of deployment – and state-of-the-art, conversational capability in which the model feels both super natural and conversational,” Stock told The Deep View.

Some use cases, Mistral notes, include customer support, sales and marketing, real-time translation and personal voice agents. The model is available to test out in Mistral Studio playground, and a model with several reference voices is available as open weights on Hugging Face.

Can SLMs solve AI’s biggest roadblock in 2026?

When I talk to people putting AI to work in organizations right now, I typically hear two things: either inference costs are out of control or they are paying too much for tokens. Spoiler: those two things mean the same thing. The cost of running AI is too high, and AI agents are making it a lot worse.

Here’s the good news (or bad news, depending on your POV): If you’re building AI agents, most of the tokens your agent consumes are going to a frontier model that’s wildly overqualified for the job, and you’re overpaying for it.

We’re talking about tasks like:

• Classifying a support ticket

• Extracting a date from an email

• Rewriting a subject line

These are not tasks that require models with hundreds of billions of parameters. A small, fine-tuned model can handle them in milliseconds at a fraction of the cost.

In an exclusive interview with The Deep View, Rob May, CEO of Neurometric, shared how the company is betting big on this mismatch.

On Wednesday, Neurometric launched its SLM Marketplace, a catalog of 115 task-specific small language models (SLMs). Each of these models is under 20 billion parameters and is fine-tuned to do one thing well. You can download any model for free, or let Neurometric host it with up to 100 million tokens per month at no charge. After that, it costs $2 per month per model.

“100 million tokens on one of these models costs us 40 cents,” May told me. “So why would we charge for it?”

May says they are seeing a 25/75 split: roughly 25% of tasks in a typical enterprise AI workflow genuinely need frontier-level reasoning. The other 75% are structured, repetitive, and narrowly focused, such as classification, extraction, formatting, routing, scoring, and summarization. SLMs can handle those tasks.

The models in the SLM Marketplace span 14 categories, from accounting and finance to developer tools. Neurometric expects to scale to 5,000 models by year’s end. For tasks not yet covered, an Auto-SLM Creator can generate custom models from a plain-language description.

One insight that stood out from my conversation with May was that the biggest challenge with small models is keeping them on track. Larger models handle this naturally, but SLMs need what May described as “micro harnesses,” or checkpoint prompts that remind the model what it’s doing and what comes next. It’s part of an emerging discipline called harness engineering.

The next generation of electricity generation

Elevate Your Career with AI & Machine Learning For Dummies PRO
Ready to accelerate your career in the fast-growing fields of AI and machine learning? Our app offers user-friendly tutorials and interactive exercises designed to boost your skills and make you stand out to employers. Whether you're aiming for a promotion or searching for a better job, AI & Machine Learning For Dummies PRO is your gateway to success. Start mastering the technologies shaping the future—download now and take the next step in your professional journey!



Download the AI & Machine Learning For Dummies PRO App:
iOS - Android
Our AI and Machine Learning For Dummies PRO App can help you Ace the following AI and Machine Learning certifications:

• AWS Certified AI Practitioner (AIF-C01): Conquer the AWS Certified AI Practitioner exam with our AI and Machine Learning For Dummies test prep. Master fundamental AI concepts, AWS AI services, and ethical considerations.


• Azure AI Fundamentals: Ace the Azure AI Fundamentals exam with our comprehensive test prep. Learn the basics of AI, Azure AI services, and their applications.


• Google Cloud Professional Machine Learning Engineer: Nail the Google Professional Machine Learning Engineer exam with our expert-designed test prep. Deepen your understanding of ML algorithms, models, and deployment strategies.


• AWS Certified Machine Learning Specialty: Dominate the AWS Certified Machine Learning Specialty exam with our targeted test prep. Master advanced ML techniques, AWS ML services, and practical applications.


• AWS Certified Data Engineer Associate (DEA-C01): Set yourself up for promotion, get a better job or Increase your salary by Acing the AWS DEA-C01 Certification.

By Bill Gates | March 26, 2026

I’m in Texas this week for the Breakthrough Energy Ventures Investors Summit, and the energy here (no pun intended) is incredible. It also matches what’s happening in the industry writ large, especially when it comes to electricity.

By 2050, the world will need to generate nearly three times as much electricity as we use today. And we need to generate it cleanly. That’s a tremendous challenge—but it’s also an exciting moment to be paying attention. Earlier this month, TerraPower, the nuclear company I started in 2008, received federal approval to begin construction at its Wyoming plant. Wind and solar are generating more electricity than fossil fuels in the EU for the first time ever. The way we power the world is changing fast.

What’s driving that change? New approaches to geothermal, fusion energy, and geologic hydrogen—some that are closer to commercialization than most people realize, and some that might still be decades away.

On my website, I’ve written about three companies in this space that I think are worth watching. You can read about them, and why I’m so optimistic about the future of clean electricity, on Gates Notes now.

What Else Happened in AI on March 26th 2026?

Oracle Data Deep Dive NYC, April 10th: Hands-on AI labs and direct access to Oracle experts. Learn more and register for free.*

OpenAI is raising another $10B to push its record funding round past $120B, with Microsoft, a16z, and T. Rowe Price joining the round.

Google upgraded its music AI model to generate full 3-minute songs with intros, verses, and choruses, with Lyria 3 Pro rolling out in Gemini, Vertex AI, and Google Vids.

Bret Taylor’s Sierra introduced Ghostwriter, an AI agent that builds other AI agents — letting companies create customer service bots across voice, chat, and 30+ languages.

The U.S. Department of Labor launched “Make America AI-Ready,” a free 7-day AI literacy course delivered entirely over text message to promote AI upskilling.

Subscribe to our AI Unraveled Feed at Apple to listen ADS-FREE:

🚀 Welcome to AI Unraveled. Today, the tech giants execute a massive strategic pivot. OpenAI axes its flagship video generator to focus on robotic world simulation, Apple prepares a total overhaul of Siri for iOS 27, and the US government locks down the hardware stack.

This episode is made possible by our sponsor:

🎙️ DjamgaMind: High-Fidelity Intelligence for the C-Suite. If you are a modern decision-maker, DjamgaMind delivers strategic audio forensics in Healthcare, Energy, and Finance. Stop reading headlines and start understanding the systemic impact with our human-verified, technical-grade analysis. 👉 Explore the Forensics: https://DjamgaMind.com/regulations

🎧 Listen Ads-Free: Tired of interruptions? Subscribe to AI Unraveled directly on Apple Podcasts to enjoy all our daily episodes completely ads-FREE!

🛑 AIRIA: With Anthropic’s new “Dispatch” feature taking remote control of your macOS desktop, security is no longer optional. AIRIA provides the enterprise-grade sandboxing required to run these autonomous remote agents safely, ensuring your corporate environment is protected from multi-turn adversarial attacks. 👉 Govern your agents: https://airia.com/request-demo/?utm_source=AI+Unraveled+&utm_medium=Podcast&utm_campaign=Q1+2026

In Today’s Briefing:

• The Death of Sora: Why OpenAI is winding down its video generator to free up compute for ‘Spud’ and the physical robotics economy.

• Apple’s ‘Ask Siri’: The standalone chatbot app coming to iOS 27 that reads emails and executes actions inside third-party apps.

• Cloudflare Dynamic Workers: New infrastructure allowing AI agents to safely write and execute code on the fly at 100x speeds.

• Enterprise Agent Panic: Insights from The Deep View on why 587 C-Suite leaders are terrified of autonomous agent security.

• Roche’s AI Factory: The pharmaceutical giant deploys 3,500 Nvidia Blackwell GPUs for in-house drug discovery.

• Microsoft Superintelligence: Mustafa Suleyman poaches former Allen Institute CEO Ali Farhadi.

• FCC Foreign Router Ban: The United States bans all new foreign-made routers to secure edge infrastructure.

Strategic Signal: The Pivot to Utility and Execution. Credits: Created and produced by Etienne Noumen.

Keywords: OpenAI Sora Shutdown, OpenAI Spud Model, Apple iOS 27 Ask Siri, Cloudflare Dynamic Workers, Roche AI Factory Blackwell, Microsoft Superintelligence Ali Farhadi, FCC Router Ban, Agentic Enterprise Security, World Simulation Robotics, DjamgaMind, AI Unraveled.

🔗 RESOURCES & CAREERS

Find AI Jobs (Mercor): Apply Here – https://work.mercor.com/?referralCode=82d5f4e3-e1a3-4064-963f-c197bb2c8db1

⚗️ PRODUCTION NOTE: We Practice What We Preach.

AI Unraveled is produced using a hybrid “Human-in-the-Loop” workflow.

OpenAI’s Sora gets the axe to prioritize ‘Spud’

Good morning, AI enthusiasts. Weeks ago, OpenAI’s Applications CEO Fidji Simo told staff to stop chasing “side quests.” Today, one of the AI leader’s most high-profile ones got the axe. Sora, the video generator that previously hit No. 1 on the App Store, is officially winding down.

• Altman reportedly told staff OAI would wind down all video products, including its mobile app and API, with employees calling Sora a “drag” on resources.

• The freed-up compute will go toward ‘Spud’, OAI’s next major model expected in the coming weeks, which Altman says “can really accelerate the economy”.

• Sora head Bill Peebles said the team will now pursue “world simulation” for robotics, calling the prize “automating the physical economy”.

• The move puts OAI’s December partnership and $1B investment by Disney on ice after the media giant initially agreed for its IP to be used in Sora.

Apple’s standalone Siri app, chatbot for iOS 27

Apple is reportedly testing a standalone Siri app alongside a new “Ask Siri” chatbot experience, according to Bloomberg insider Mark Gurman — with both slated to debut with its iOS 27 system at WWDC in June.

• Siri will get its own dedicated app for the first time, along with a redesigned interface that lets users type or speak requests.

• The assistant will reportedly read across iMessages, emails, and notes to build context, also capable of executing actions inside third-party apps directly.

• Apple is branding the experience “Ask Siri,” positioning it as a chatbot-style interaction rather than the rigid voice commands of the previous iterations.

Arm releases first in-house chip, with Meta as debut customer LINK

• Arm Holdings has released its first in-house chip, the Arm AGI CPU, after nearly 36 years of only licensing its designs to companies like Nvidia and Apple, with Meta as its debut customer.

• The Arm AGI CPU is a production-ready processor built for running inference in AI data centers, developed using the Arm Neoverse family of CPU IP cores through a partnership with Meta.

• Arm started developing the chips back in 2023, and they are already ready to order, marking a historic shift from exclusively licensing designs to now competing alongside many of its partners.

Cloudflare launches Dynamic Workers for AI Agents

Cloudflare launched Dynamic Workers, a new tool that lets AI agents quickly and safely run code they write on the fly at 100x the speed of other options. As agents move from generating text to executing tasks, secure serverless environments are becoming the ultimate bottleneck. Cloudflare’s new architecture allows autonomous systems to spin up, execute, and destroy code environments in milliseconds without exposing the host network.

The Deep View: Enterprise AI Adoption Blockers

In a new episode of The Deep View: Conversations, Salesforce SVP of Enterprise IT Strategy Shibani Ahuja revealed findings from meetings with 587 C-suite leaders over the past year. The consensus? The enterprise is terrified of agentic AI. Leaders relying on simple chatbots are falling behind, but organizations deploying autonomous agents are hitting massive security blockers. Gartner also named Blackbird the company to beat in disinformation narrative intelligence, as AI supercharges the speed and scale of malicious content.

Pharma giant Roche launches “AI factory”

Roche launched a new “AI factory” to speed up drug discovery and manufacturing, with the company now amassing over 3,500 Nvidia Blackwell GPUs. This represents a massive capital expenditure shift for the pharmaceutical industry, moving compute from cloud providers to on-premise, vertically integrated AI factories.

Microsoft Superintelligence poaches Allen Institute leaders

Microsoft AI poached three top researchers from the Allen Institute for AI, including former CEO Ali Farhadi, with the trio joining Mustafa Suleyman’s Superintelligence team. The move further solidifies Microsoft’s commitment to building its own frontier models to reduce its long-term reliance on OpenAI.

Techpresso: FCC bans all foreign-made routers

In a massive geopolitical tech shift, the FCC has officially banned the import and sale of all new foreign-made network routers in the United States. The mandate is designed to secure the lowest level of the hardware stack against foreign espionage as AI inference moves to the edge.

Meta ordered to pay $375 million in child safety case LINK

• A New Mexico jury found Meta liable on every count in a child safety case, ordering the company to pay $375 million for endangering children and concealing sexual exploitation on Instagram and Facebook.

• Attorney General Raul Torrez framed the case as a products-liability claim, arguing the platforms were defective — a legal strategy that sidesteps Section 230 and gives 40-plus state attorneys general a tested courtroom playbook.

• A bench trial expected in May will not ask how much Meta should pay but what it must change, including age verification, predator removal, and modifications to encrypted messaging across multiple states.

Amazon acquires ‘approachable’ humanoid maker Fauna Robotics LINK

• Amazon has acquired Fauna Robotics, a startup that makes “approachable” humanoid robots designed for consumers and businesses, though the companies did not share the financial terms of the deal.

• Fauna Robotics was founded in 2024 by former Meta and Google engineers and earlier this year launched Sprout, a $50,000 bipedal robot standing 3.5 feet tall and weighing 50 lbs.

• Fauna’s roughly 50 employees will join Amazon in New York City, and the company will continue to operate as Fauna Robotics under Amazon, according to CEO Rob Cochran.

Trump appoints tech CEOs to White House council LINK

• President Trump has appointed CEOs from Meta, NVIDIA, Dell, Oracle, and AMD, along with Google co-founder Sergey Brin and venture capitalist Marc Andreessen, to a White House science and technology advisory council.

• The President’s Council of Advisors on Science and Technology currently has 13 members, co-chaired by White House AI and cryptocurrency czar David Sacks and Trump’s science advisor Michael Kratsios, with room to grow to 24.

• Several of these tech leaders have direct financial ties to Trump, including donations to his inauguration, funding construction of his White House ballroom, and business deals like Oracle’s backing of the TikTok takeover.

What Else Happened in AI on March 25th 2026?

Microsoft AI poached three top researchers from the Allen Institute for AI, including former CEO Ali Farhadi, with the trio joining Mustafa Suleyman’s Superintelligence team.

Figma opened its design canvas to coding agents, letting tools like Claude Code create/ edit designs directly using a team’s existing components and brand standards.

Pharma giant Roche launched a new “AI factory” to speed up drug discovery and manufacturing, with the company now amassing over 3,500 Nvidia Blackwell GPUs.

Cloudflare launched Dynamic Workers, a new tool that lets AI agents quickly and safely run code they write on the fly at 100x the speed of other options.

The OAI Foundation committed $1B this year across disease, job displacement, and AI safety research, with co-founder Wojciech Zaremba joining as Head of AI Resilience.

LISTEN TO ADS-FREE Audio of this episode at

https://djamgamind.com

https://podcasts.apple.com/us/channel/djamgamind/id6760446113

🚀 Welcome to AI Unraveled. Today, the AI bubble meets geopolitical reality. The Iran-U.S. war is threatening global semiconductor cooling supplies, forcing hyperscalers to rethink their Middle East expansion. Meanwhile, Anthropic takes over the desktop, and OpenAI secures another $10 billion while shutting down its video generation platform.

This episode is made possible by our sponsors:

🛑 AIRIA: With Anthropic’s new “Dispatch” feature taking remote control of your macOS desktop, security is no longer optional. AIRIA provides the enterprise-grade sandboxing required to run these autonomous remote agents safely, ensuring your corporate environment is protected from multi-turn adversarial attacks. 👉 Govern your agents: https://airia.com/request-demo/?utm_source=AI+Unraveled+&utm_medium=Podcast&utm_campaign=Q1+2026

🎙️ DjamgaMind: Skip the ads and get the macroeconomic breakdown. Join our Ads-FREE Premium Feed at DjamgaMind for the technical deep-dive into the AI industry’s shift to physical hardware. 👉 Switch to Ads-Free: [DjamgaMind on Apple Podcasts / Spotify] at https://djamgamind.com

In Today’s Briefing:

• Geopolitical Tech Crisis: How the Iran-U.S. war, the Strait of Hormuz blockade, and strikes on Qatar’s helium plants are threatening the global semiconductor supply chain.

• Anthropic Dispatch: Claude gets direct remote control of your computer, completing tasks while you step away.

• Luma AI Uni-1: A new foundational image model that processes text and visuals through a single “thinking” pipeline.

• Jensen Huang on AGI: Nvidia’s CEO claims Artificial General Intelligence has already been achieved via agentic software.

• OpenAI’s Reality Check: A $10B funding round at a $730B valuation, the official shutdown of Sora, and IPO risk disclosures detailing a heavy reliance on Microsoft and TSMC.

• Zuck’s Internal Agents: Meta mandates AI usage in performance reviews as Zuckerberg builds a personal “CEO agent” to bypass middle management.

• Cisco’s LLM Security Leaderboard: Anthropic dominates the top 10 for multi-turn attack resistance, while open-weights models struggle.

• Apple Business: A new all-in-one device management and productivity platform launching in April.

Strategic Signal: Software AGI vs. Physical Supply Chain Fragility.

Keywords: Iran US War Tech Impact, Qatar Helium Shortage, Strait of Hormuz Semiconductors, Anthropic Dispatch Remote Computer Use, Luma AI Uni-1, Jensen Huang AGI Claim, OpenAI $10B Funding, OpenAI Sora Shutdown, Meta CEO Agent My Claw, Cisco LLM Security Leaderboard, Apple Business Platform, Fauna Robotics Sprout, DjamgaMind, AI Unraveled.

🚀 FOR LEADERS: DjamgaMind Audio Intelligence

Don’t Read the Regulation. Listen to the Risk. Drowning in dense legal text? DjamgaMind turns 500-page healthcare/energy/finance mandates into 15-minute executive audio briefings.

👉 Start your briefing: https://DjamgaMind.com/regulations

🔗 RESOURCES & CAREERS

Find AI Jobs (Mercor): Apply Here –

https://work.mercor.com/?referralCode=82d5f4e3-e1a3-4064-963f-c197bb2c8db1

⚗️ PRODUCTION NOTE: We Practice What We Preach.

AI Unraveled is produced using a hybrid “Human-in-the-Loop” workflow.

Anthropic ships remote computer use

Anthropic just released a research preview that hands Claude direct control of your desktop — letting it click, type, and navigate across any app on your Mac while you step away, with phone-based task assignment through Dispatch.

The details:

• The newly released Dispatch turns the combo into a remote setup, allowing users to fire off a task from mobile and letting Claude handle it on the computer.

• The system is built to avoid screen control when possible, checking for direct app integrations and browser access before resorting to clicking.

• The feature is only available to macOS users on Pro or Max plans currently via Cowork and Claude Code, with a Windows version also in the pipeline.

• Anthropic acquired computer use startup Vercept in February, with the new release marking the team’s first product launch after just four weeks.

Why it matters: Anthropic’s Alex Albert puts it well, saying, “the future where I never have to open my laptop to get work done is becoming real very fast”. While losing OpenClaw to OAI was considered by many to be a miss, the recent flurry of features has shown the building blocks forming to turn Claude into its own remote agent.

Luma AI’s new image model thinks as it generates

Image source: Luma AI

Luma AI rolled out Uni-1, an image model that processes text and visuals through the same pipeline — thinking through what it’s asked to do before and while it creates, with the company calling this approach “path to general intelligence.”

The details:

• Uni-1 runs on the same type of architecture as GPT Image 1.5 and Nano Banana Pro, processing text and images in a single pipeline instead of diffusion.

• The model also features real-world understanding, enabling creative decisions and use cases such as infographics, manga, and specific aesthetics.

• In testing, Uni-1 topped human preference rankings for style, editing, and reference-based work, trailing only Nano Banana Pro in text-to-image ELO.

• Uni-1’s API price of ~$0.09 / image at 2K resolution undercuts Nano Banana Pro’s $0.134 rate by roughly a third, though the API is waitlist-only for now.

Why it matters: Luma made its name in video, so an image model is a new direction. If the same system can extend into video, voice, and interactive worlds as Luma is teasing, Uni-1 could set the foundation for one model that can do it all creatively — moving into the creative agent territory that users are starting to expect.

War in Iran puts tech industry on fragile footing

The tech industry is notorious for operating within its own bubble — sometimes even its own reality distortion field — but the impacts of the Iran-U.S. war are threatening to bear down on it.

Multiple factors are now in play in the conflict that could disrupt tech companies and impact the pace of AI growth:

• Iran names U.S. tech firms as targets: The official news agency of the Iranian military listed Amazon, Microsoft, Palantir, and Oracle as the “enemy’s technological infrastructure” and made clear that it considers them military targets. This was connected to the U.S. threat to obliterate Iran’s power plants, a stance that has since been softened.

• Critical mineral shortage disrupts chip makers: Semiconductors run the world, especially AI, and the industry is facing a critical shortage of minerals because of the conflict. A third of the world’s helium comes from Qatar, and it’s essential for cooling systems and circuits in producing semiconductors. The closure of the Strait of Hormuz puts the semiconductor supply chain at risk, and Iran has already struck the Qatar helium plant at Ras Laffan and taken it offline.

• Hyperscalers rethink Middle East expansion: Tech companies had been preparing to invest billions of dollars in data centers and AI factories, but the instability and uncertainty of the conflict between the U.S./Israel and Iran has put those plans in jeopardy. Iran has already attacked AWS buildings in the UAE. OpenAI, Nvidia, Oracle, and Cisco have been collaborating on a potential 5-gigawatt facility in the UAE. But a prolonged conflict could redirect this and other projects to safer havens like India, Southeast Asia, or Northern Europe.

Apple announces Apple Business LINK

• Apple announced Apple Business, a free all-in-one platform that combines device management, productivity tools, and customer outreach into a single service replacing Apple Business Essentials, Apple Business Manager, and Apple Business Connect.

• The platform includes built-in MDM, new “Blueprints” for zero-touch deployment, Managed Apple Accounts with cryptographic separation between personal and work data, and integrated email, calendar, and directory services.

• Apple Business launches April 14 in over 200 countries, and existing data from the three discontinued services will automatically migrate, while Business Essentials customers will stop being charged monthly device management fees.

Jensen Huang claims AGI has already been achieved LINK

• NVIDIA CEO Jensen Huang told Lex Fridman on his podcast that he believes AGI has already been achieved, pointing to agentic tools that could theoretically build and run a viral app.

• The claim matters because OpenAI’s partnership with Microsoft includes escape clauses tied to AGI, though their contract defines it as an AI model generating $100 billion in profit.

• Microsoft has been preparing for a possible split by restructuring its AI division to focus on its own models, while tensions grow over OpenAI’s latest funding round and competing partnerships.

Zuck ramps up Meta’s internal AI agent use

Mark Zuckerberg is creating a personal “CEO agent” to shortcut the chain of command when he needs quick answers, according to the WSJ, coming as part of a company-wide mandate that now factors AI usage into performance reviews.

The details:

• Zuck’s agent is still in development, but already handles tasks like pulling answers that typically require going through multiple layers of Meta’s org chart.

• Staffers have spun up custom agent tools, including one called “My Claw” that reads their work files and negotiates with coworkers’ bots directly.

• Another Claude-powered internal tool called “Second Brain” acts as an AI chief of staff, pulling answers from any internal document on demand.

• Zuckerberg had previously courted OpenClaw creator Peter Steinberger, and also acquired Chinese agentic platform Manus in December.

Why it matters: Meta may have tens of thousands of employees, but that isn’t stopping the newer parts of the org from trying to move as fast and lean as some of its more AI-native rivals. With Zuck seemingly very invested in the AI agent boom, Meta’s integration of Manus will be one of the more interesting implementations to watch for.

OpenAI flags Microsoft dependence as IPO risk LINK

• OpenAI identified its heavy reliance on Microsoft as a business risk in a financial document shared with investors, noting that Microsoft provides “a substantial portion” of its financing and compute.

• The document also flagged risks including a global chip shortage, potential disruption to Taiwan Semiconductor Manufacturing Company from regional conflict, and roughly $665 billion in compute spend commitments through 2030.

• OpenAI disclosed at least 14 lawsuits from ChatGPT users or families blaming its products for mental illness leading to suicide or injury, plus three separate lawsuits from Elon Musk or xAI.

OpenAI’s latest raise:

In major OpenAI news, Bloomberg reports that the company is nearing a deal for $10 billion in fresh funding from a string of venture firms and funds, including Abu Dhabi’s MGX, Coatue Management, and Thrive Capital. This will value the company at a staggering $730 billion, according to the report, which suggests the deal will close by the end of the month. That’s on top of the $110 billion in funds announced last month, coming into the House of Altman from Amazon, Nvidia, and SoftBank. (For comparison’s sake, OpenAI’s fiercest rival Anthropic recently completed a $30 billion round — which also included MGX — valuing the Claude maker at $380 billion.

Not you, Sora: OpenAI Will Shut Down Sora Video Platform

To what will OpenAI dedicate all of this incoming capital? Unclear, but definitely not the Sora “slop feed” app, which the company announced plans to discontinue. In a post to the official Sora account on X, OpenAI confirms “we’re saying goodbye to Sora,” adding “what you made with Sora mattered, and we know this news is disappointing.” Disappointing, perhaps, but it’s not a COMPLETE surprise, though. Just one week ago, WSJ reported that OpenAI’s CEO of Applications Fidji Simo had told staffers the company was shifting focus to productivity applications for enterprises, and away from “side quests.” Sora clearly fell in the latter category.

Amazon picks up Fauna Robotics:

The New York-based robotics startup is developing a humanoid 3.5-foot domestic helper bot, named Sprout, designed for handling basic household chores like fetching small items and doing a little cleaning up. (Fauna’s also focused on “fun robots,” so naturally, Sprout is capable of human interaction and has some dance moves.) No announced plans yet for a Sprout consumer release, but the company started sending prototypes to “research and development partners” earlier this year.

Anthropic takes 8 spots in top 10 most secure LLMs

The promise of AI-driven productivity comes with a catch: every implementation hands over the keys to your company’s data and operations to new technology, unlocking a host of security risks.

The leaderboard results were calculated based on rigorous testing that measured single- and multi-turn attacks aimed at eliciting a harmful or malicious response from the model. Anyone can access the results for free, but here is a quick breakdown:

• Anthropic: The company dominated the leaderboard, holding 8 out of the top 10 spots, with Claude Opus 4.5, taking first place, followed by Sonnet 4.5 and Haiku 4.5.

• OpenAI: GPT-5.2 and GPT 5 Nano managed to make it into the top 10, too, coming in 7th and 9th place, respectively.

• Bottom of the leaderboard: Mistral took the last two places with its Magistral Small 2509 and Ministral 3 14b Instruct models. The list of the bottom 10 (least secure models) also includes models from DeepSeek, Cohere, Qwen and xAI.

What Else Happened in AI on March 24th 2026?

Nvidia CEO Jensen Huang appeared on the Lex Fridman Podcast, saying, “I think it’s now. I think we’ve achieved AGI” when asked about his intelligence timelines.

Apple announced its WWDC 2026 event will run June 8-12, teasing ‘AI advancements’ that are speculated to include its Siri overhaul powered by Google Gemini.

OpenAI is reportedly guaranteeing a 17.5% minimum return to lure private equity firms into its enterprise joint venture — outbidding Anthropic as both prep for IPOs.

Agentic personal software builder Dreamer announced it is licensing its tech to Meta, with its full team joining Meta Superintelligence Labs in an undisclosed deal.

OpenAI hired former Meta VP of global clients Dave Dugan to run its ad sales, coming as the company continues its initial advertising push into ChatGPT.

OpenAI Foundation pledges $1B in grants to ensure AI ‘benefits all of humanity’ [Link]

Steve Wozniak says he’s “disappointed a lot” by AI and rarely uses it [Link]

LISTEN TO ADS-FREE Audio of this episode at https://djamgamind.com

🚀 Welcome to DjamgaMind. Today, the digital models break into the physical world. Elon Musk launches a $25 billion chip factory, Peter Thiel backs AI cow collars at a $2B valuation, and DoorDash turns its gig workers into walking spatial-data cameras for robotics training.

This episode is made possible by our sponsors:

🛑 AIRIA: As executives follow Mark Zuckerberg’s lead and deploy “AI Chiefs of Staff” to access internal company data, the risk of unauthorized data exposure skyrockets. AIRIA provides the enterprise guardrails to ensure your internal agents execute tasks securely without leaking your intellectual property. 👉 Govern your agents: https://airia.com/request-demo/?utm_source=AI+Unraveled+&utm_medium=Podcast&utm_campaign=Q1+2026

🎙️ DjamgaMind: Skip the ads and get the macroeconomic breakdown. Join our Ads-FREE Premium Feed at DjamgaMind for the technical deep-dive into the AI industry’s shift to physical hardware. 👉 Switch to Ads-Free: [DjamgaMind on Apple Podcasts / Spotify] at https://djamgamind.com

In Today’s Briefing:

• Terafab: Elon Musk’s $25 billion mega-factory designed to bring chip fabrication in-house and deploy a terawatt of AI compute into space.

• Zuck’s AI Agent: Meta’s CEO is building a personal AI Chief of Staff to bypass corporate bureaucracy, amidst rumors of 20% workforce layoffs.

• The DoorDash Data Harvest: Why the new “Tasks” app is paying gig workers to capture spatial and biometric data to train delivery robots.

• The $2B ‘Cowgorithm’: Halter nears a massive valuation to replace agricultural fencing with AI-driven smart collars.

• The Wrapper Backlash: Cursor faces heat for building its Composer 2 on top of Kimi K2.5, while ChatGPT’s first advertisers see dismal 0.91% click-through rates.

• Claude Cowork Projects: Anthropic’s move to store project contexts locally on desktop, boosting interoperability and privacy.

• Congressional Crackdown: A bipartisan bill aims to ban sports and casino contracts on prediction markets like Polymarket and Kalshi.

Strategic Signal: Vertical Hardware Sovereignty & Physical Data Extraction.

Keywords: Elon Musk Terafab, $25B Chip Factory, Orbital Data Centers, DoorDash Tasks App, Spatial Data Harvesting, Halter AI Cowgorithm, Peter Thiel AgTech, Mark Zuckerberg AI Agent, Claude Cowork Projects, Cursor Composer 2 Kimi K2.5, Polymarket Prediction Market Ban, ChatGPT Ad Performance, DjamgaMind, AI Unraveled.

🚀 FOR LEADERS: DjamgaMind Audio Intelligence

Don’t Read the Regulation. Listen to the Risk. Drowning in dense legal text? DjamgaMind turns 100-page healthcare/energy/finance mandates into 5-minute executive audio briefings. Whether navigating Bill C-59 or HIPAA compliance, our AI agents decode the liability so you don’t have to.

👉 Start your briefing: https://DjamgaMind.com/regulations

🔗 RESOURCES & CAREERS

Find AI Jobs (Mercor): Apply Here – https://work.mercor.com/?referralCode=82d5f4e3-e1a3-4064-963f-c197bb2c8db1

⚗️ PRODUCTION NOTE: We Practice What We Preach.

AI Unraveled is produced using a hybrid “Human-in-the-Loop” workflow.

Musk unveils $25B Terafab chip factory LINK

• Elon Musk announced plans to build a chip factory called Terafab, a joint project between Tesla and SpaceX, with an estimated cost of $25 billion near Tesla’s Austin headquarters.

• Musk said semiconductor manufacturers aren’t making chips fast enough for his companies’ artificial intelligence and robotics needs, so he decided to build the facility himself.

• The factory aims to produce chips supporting 100 to 200 gigawatts of computing power per year on Earth and a terawatt in space, though Musk gave no timeline.

Zuckerberg builds an AI agent to help him run Meta LINK

• Mark Zuckerberg is reportedly building a personal AI agent to help him run Meta, mainly by speeding up information retrieval without going through multiple layers of people or teams.

• Meta employees are already using agentic tools like MyClaw for accessing work files and chat logs, and Second Brain, built on Anthropic’s Claude, which is described internally as an “AI chief of staff.”

• The push comes as Meta tries to compete with AI-native startups that have smaller teams, and a separate Reuters report claims the company may be planning layoffs affecting up to 20% of its workforce.

Congress moves to ban sports prediction markets LINK

• A bipartisan pair of U.S. senators, Democrat Adam Schiff and Republican John Curtis, plan to introduce a bill that would ban federally regulated prediction market platforms from offering contracts tied to sports and casino-style games.

• The bill targets platforms like Kalshi and Polymarket, where 90 percent of trading volume comes from sports contracts, meaning a federal ban would destroy the dominant revenue stream for companies seeking $20 billion valuations.

• States are already fighting prediction markets through lawsuits and criminal charges in multiple jurisdictions, while attorneys general from 39 states have urged federal courts to uphold state authority over sports gambling regulation.

ChatGPT’s first advertisers can’t prove ads work LINK

• OpenAI’s first advertising partners — WPP, Omnicom, and Dentsu — are struggling to prove that ChatGPT ads actually work, with click-through rates running nearly 7x below Google search benchmarks.

• One brand’s click-through rate on ChatGPT ads hit just 0.91% compared to Google’s 6.4%, and a separate advertiser spent only 3% of a $250,000 budget after several weeks.

• Measurement tools are also broken — a reporting glitch in OpenAI’s Ad Manager blocks advertisers from seeing their own data, making it impossible to optimize campaigns or justify continued spend.

DoorDash offers cash for AI data. Read the terms

A company is willing to pay people to capture real-world content to train robots. No, that’s not a Black Mirror episode, it’s DoorDash’s new app.

On Thursday, DoorDash unveiled Tasks, a feature that pays Dashers to provide businesses with ground-level insights, from photographing restaurant food to capturing hotel entrances for easier deliveries. Win-win, right? Well, tucked below that is a paragraph about a new standalone app DoorDash is piloting.

“We’re also piloting a new standalone app where Dashers can complete activities like filming everyday tasks or recording themselves speaking in another language,” said DoorDash. “This data helps AI and robotic systems understand the physical world.”

Real-world data is one of AI’s most valuable and limited resources. Without it, manufacturers are stuck training models on simulations that only approximate reality. That’s likely why DoorDash is willing to pay upfront: it’s a premium commodity they can upsell. DoorDash has not responded to a request for comment.

While getting paid to perform and record routine tasks may sound like a good way to make extra cash, users should soberly consider the privacy and security implications, which, according to Andrew Shimshock, CTO and Co-Founder of Mill Pond Research, a company focused on secure AI solutions, are numerous.

“Once an AI or robotic system is trained on biometric or spatial data and released into production, the underlying training data must be considered compromised,” said Shimshock. “Bad actors do not need to breach a server; they simply need access to query the AI.”

Shimshock also highlighted that once a person’s biological markers are baked into a multimodal AI, that data cannot be unlearned or deleted. Beyond the consenting individual’s data, there is also the question of what that means for individuals incidentally captured in the background of a task, without their knowledge or consent.

While dystopian, this approach isn’t new. Data from the popular Pokémon Go game, which had hundreds of millions of players wandering the streets to capture the game characters, has now been used to create a Visual Positioning System (VPS) to train delivery robots.

The lines of user consent are getting increasingly blurred in the era of AI. The question is, when is too much? It is an accepted truth that the entirety of the internet now lives in the datasets behind these powerful models, but what if our physical existence gets baked into them too?

Claude Cowork’s new Projects power up workers

Anthropic’s Claude Cowork has maintained steady momentum since its January launch, with no signs of slowing.

On Friday, Anthropic launched a Project feature for the Claude Cowork desktop app and it allows users to create different workspaces that contain the tasks, instructions, and files relevant to that one area of focused work.

The content in the Project is then stored locally in a folder on the user’s computer. This has several advantages. Users never need to reupload the same information between sessions, can access it without constant context-switching, and retain full control over their files.

It also helps with vendor interoperability. Often, users don’t want to switch between AI tools simply because of the time it took to tailor them to their preferences and needs. With the instructions and files being stored locally, users can then easily reupload the personalization information elsewhere.

To create a new project, users can either start from scratch by setting up a new folder with instructions and files, import from chat by bringing over instructions and files from an existing chat project, or use an existing folder they already work from. Anthropic said users can update or download the Claude desktop app to try it.

While this is in itself a major update, it doesn’t seem like Anthropic is stopping there. The company is reportedly developing a Sketch tool for Claude, according to an X post. This would allow users to input a quick sketch to explain what they’d like done, rather than having to describe it in words.

Halter’s AI ‘Cowgorithm’ nears $2B valuation

Image source: Halter

New Zealand’s Halter is reportedly nearing a new round led by Peter Thiel’s Founder’s Fund that would value the startup at $2B — with its ‘Cowgorithm’ collars using AI in agriculture to track, herd, and modernize pasture management.

The details:

• Halter’s solar-powered collars create virtual fences and let ranchers herd cattle remotely via app using vibration and audio cues.

• Collars send 6,000+ data points per minute to Halter’s proprietary AI, which it calls the ‘Cowgorithm’, to track health and optimize grazing.

• The round would double Halter’s valuation from roughly $1B after a $100M raise last June, with investor demand leaving the deal oversubscribed.

• Founded on a 300-cow NZ dairy farm, Halter has collected 7B+ hours of animal behavior data and is now expanding into the US market.

Why it matters: Some of the biggest AI unlocks might be coming in industries that haven’t materially changed in centuries. Halter replaces physical fences, manual herd checks, and more with real-time data, a collar, and an app — bringing practical modernization to an old-school process. Plus, ‘Cowgorithm’ is an excellent name.

What Else Happened in AI on March 23rd 2026?

Cursor revealed that its Composer 2 model was built on top of Kimi K2.5, coming after the company faced backlash over failing to include the detail in its release.

Anthropic rolled out Projects in its Claude Cowork, allowing users to import their existing web-based Claude projects or create them to use on the desktop app.

The White House released its AI policy blueprint for Congress, aiming to block states from writing their own AI laws while keeping federal oversight across existing agencies.

OpenAI is reportedly planning to nearly double its workforce from 4,500 to 8,000 by the end of 2026 amid the company’s increasing enterprise AI focus.

A U.S. man pleaded guilty to an AI music fraud that earned him $1.2M/year by generating fake tracks and inflating their play counts, now facing five years in prison.

Tencent integrates WeChat with OpenClaw AI agent amid China tech battle. [LINK]

AI-generated ads are trickling into political campaigns, sparking big worries. [LINK]

US man pleads guilty to defrauding music streamers out of millions using AI. [LINK]

AI rebuilds molecules from exploding fragments. [LINK]

LISTEN TO ADS-FREE Audio of this episode at https://podcasts.apple.com/us/channel/djamgamind/id6760446113

🚀 Welcome to the AI Unraveled Weekly Recap. This week, the AI bubble officially hardened into physical infrastructure. We track OpenAI’s massive enterprise pivot, the geopolitical fallout between the Pentagon and Anthropic, and the billions pouring into robotics, orbital data centers, and heavy manufacturing.

This episode is made possible by our sponsor:

🎙️ DjamgaMind: Stop drowning in the daily noise. Get the macroeconomic, forensic view of the tech industry. Join our Ads-FREE Premium Feed at DjamgaMind for uninterrupted, high-density audio intelligence. 👉 Switch to Ads-Free: https://podcasts.apple.com/us/channel/djamgamind/id6760446113

or

https://DjamgaMind.com

In This Week’s Briefing:

• The OpenAI Pivot: Scrapping side quests, doubling the workforce to 8,000, acquiring Python toolmaker Astral, launching GPT-5.4 Mini/Nano, and building a unified desktop Superapp.

• Capital for Compute: Meta considers cutting 20% of its workforce while signing a massive $27 billion infrastructure deal with Nebius.

• The Physical Takeover: Jeff Bezos seeks $100B for Project Prometheus to automate manufacturing; Travis Kalanick launches robotics firm Atoms; Uber invests $1.25B in Rivian.

• Nvidia’s Trillion-Dollar Sights: Jensen Huang projects $1T in AI chip sales and unveils Vera Rubin Space-1 orbital data centers.

• The Friction of Autonomy: The DOD calls Anthropic an “unacceptable national security risk”; Meta suffers a “Sev 1” rogue agent incident; Apple cracks down on “Vibe Coding.”

• Hardware & Platforms: Amazon is building an Alexa phone (”Transformer”); Google’s new “Vibe Design” Stitch tool.

• Data Privacy & IP Wars: ByteDance halts Seedance 2.0 over Disney lawsuits; Google faces backlash for AI-hallucinated news headlines; Pokémon Go data is now steering delivery robots.

Strategic Signal: Software Consolidation and Physical Expansion.

Credits: Created and produced by Etienne Noumen.

🚀 FOR LEADERS: DjamgaMind Audio Intelligence

Don’t Read the Regulation. Listen to the Risk. Drowning in dense legal text? DjamgaMind turns 100-page healthcare/energy/finance mandates into 5-minute executive audio briefings. Whether navigating Bill C-59 or HIPAA compliance, our AI agents decode the liability so you don’t have to.

👉 Start your briefing: https://DjamgaMind.com/regulations

🔗 RESOURCES & CAREERS

Find AI Jobs (Mercor): Apply Here –

https://work.mercor.com/?referralCode=82d5f4e3-e1a3-4064-963f-c197bb2c8db1

⚗️ PRODUCTION NOTE: We Practice What We Preach.

AI Unraveled is produced using a hybrid “Human-in-the-Loop” workflow. While all research, interviews, and strategic insights are curated by Etienne Noumen, we leverage advanced AI voice synthesis for our daily narration to ensure speed, consistency, and scale.

OpenAI to nearly double workforce LINK

• OpenAI plans to nearly double its workforce from 4,500 to 8,000 by the end of 2026, according to a Financial Times report citing two people with knowledge of the matter.

• Most of the new hires will work across product development, engineering, research, and sales, covering the core teams that build and sell OpenAI’s tools to customers.

• The company is also recruiting specialists focused on “technical ambassadorship,” a role designed to help businesses make better use of its existing products like ChatGPT.

Google is replacing news headlines with AI ones LINK

• Google is now using AI to rewrite news headlines that appear in its search results, sometimes changing their meaning, after previously doing something similar in its Google Discover news feed.

• The practice is not entirely new — Google has been altering headlines in search results for years — but recent examples show AI-rewritten headlines that poorly reflect the actual articles they link to.

• Google Search maintains a visual tone of being a neutral directory, but its AI Overviews and rewritten headlines can misrepresent source material, making results misleading for people who trust them.

OpenAI is planning a desktop ‘superapp’ LINK

• OpenAI plans to combine its Mac apps for ChatGPT, Codex, and Atlas into one single “superapp,” according to a report from The Wall Street Journal confirmed by an OpenAI spokesperson.

• Chief of Applications Fidji Simo told her team in an internal memo that OpenAI was “spreading our efforts across too many apps and stacks,” which slowed development and hurt quality.

• OpenAI expects to first add agentic features to Codex for productivity tasks beyond coding, then merge ChatGPT and the Atlas browser into the superapp, while the mobile app stays unchanged.

Amazon is making an Alexa phone LINK

• Amazon is working on a new smartphone codenamed “Transformer,” its first attempt at a phone in over 11 years since the failed Fire Phone, according to a Reuters report citing anonymous sources.

• The device would feature personalized tools for Amazon Shopping, Prime Video, and Prime Music, with AI features and Alexa support meant to push customers toward the company’s AI products.

• Development is led by a unit called ZeroOne, run by J Allard, a former Microsoft executive who helped create the Xbox, inside Amazon’s Devices and Services division.

Jeff Bezos seeks $100 billion for AI manufacturing fund LINK

• Jeff Bezos is reportedly trying to raise $100 billion for a new fund that would acquire companies across major industrial sectors and then modernize and automate them using AI.

• The fund is tied to Project Prometheus, a startup Bezos co-founded with former Google executive Vik Bajaj, which launched with $6.2 billion to build AI models for manufacturing and engineering.

• Bezos recently traveled to Singapore and the Middle East to raise money, with plans to acquire companies in areas like aerospace, chipmaking, and defense that would adopt Prometheus’ models.

White House releases national AI framework LINK

• The White House published a national AI framework that asks Congress to override state laws governing how AI models are developed and to avoid creating any new federal agencies for AI regulation.

• The framework calls on Congress to protect children by keeping state bans on AI-generated child sexual abuse material, adding age-gating requirements for models, and giving parents tools for safeguards.

• Senate Majority Leader John Thune acknowledged that even Republicans worry about trampling state rights, and past efforts to block states from regulating AI have already failed twice in Congress.

Google unveils new ‘vibe design’ tool LINK

• Google is redesigning its AI tool Stitch into what it calls a “vibe design” canvas, where anyone can type natural language descriptions and get high-fidelity UI designs without starting from wireframes.

• The updated Stitch features an infinite canvas, a design agent that reasons across a project’s full history, and an Agent manager that tracks progress on multiple ideas in parallel.

• Stitch now includes voice input for real-time design critiques and updates, plus an MCP server and SDK that let teams export designs to developer tools like AI Studio and Antigravity.

Meta is having trouble with rogue AI agents LINK

• An AI agent at Meta went rogue, accidentally exposing sensitive company and user data to employees who were not authorized to access it for about two hours.

• The incident started when an engineer asked an AI agent to help answer a technical question, and the agent posted a response on an internal forum without getting the engineer’s permission first.

• Meta rated the incident a “Sev 1,” its second-highest severity level, and a separate recent case saw a Meta director’s OpenClaw agent delete her entire inbox despite instructions to confirm actions.

Apple is behind in AI and still making a fortune from it LINK

• Apple’s own AI products trail competitors, but the company is on track to pass $1 billion in generative AI revenue by 2026, mostly from App Store fees charged to chatbot apps like ChatGPT.

• Generative AI apps paid Apple close to $900 million in 2025, with three-quarters coming from ChatGPT alone, and monthly revenue peaking at $101 million in August before dropping off.

• Apple spends far less on chips and data centers than rivals, and Siri still runs on outdated technology — its new version will initially rely on Google’s Gemini after internal setbacks.

OpenAI acquires Python toolmaker Astral LINK

• OpenAI has announced plans to acquire Astral, the company behind popular Python development tools like uv, Ruff, and ty that many developers already depend on daily.

• The deal signals OpenAI’s push to embed Codex directly into real development workflows rather than just generating code snippets, using Astral’s tools for linting, dependency management, and type safety.

• OpenAI says it will keep supporting Astral’s open source projects, but developers are understandably nervous about whether these widely used neutral tools will slowly become optimized for one platform.

Uber invests $1.25 billion in Rivian for 50,000 robotaxis LINK

• Uber is investing up to $1.25 billion in Rivian to purchase as many as 50,000 autonomous R2 SUVs, starting with a $300 million commitment and an initial order of 10,000 robotaxis.

• The companies plan to launch the fleet in San Francisco and Miami by 2028, expanding to 25 cities across the U.S., Canada, and Europe by the end of 2031.

• Rivian faces big challenges: the R2 SUV isn’t in production yet, its Georgia factory is still under construction, and its self-driving system has not been tested for robotaxi use.

OpenAI launches GPT-5.4 Mini and Nano models LINK

• OpenAI has released GPT-5.4 mini and GPT-5.4 nano, two smaller and cheaper models designed for fast, high-volume AI workloads like coding assistants, subagents, and multimodal applications that process images in real-time.

• GPT-5.4 mini runs more than twice as fast as GPT-5 mini and scores 54.38% on SWE-bench Pro compared to 45.69%, approaching GPT-5.4-level pass rates while costing roughly one-third as much.

• GPT-5.4 nano is API-only at $0.20 per million input tokens, aimed at classification, extraction, and ranking tasks, while mini is available in API, Codex, and ChatGPT across all user tiers.

DOD calls Anthropic an unacceptable national security risk LINK

• The U.S. Department of Defense has formally called Anthropic an “unacceptable risk to national security” in a court filing, pushing back for the first time against the AI company’s lawsuits over its supply chain label.

• The DOD argues in a 40-page filing that Anthropic might disable its technology or change its model’s behavior during warfighting operations if the company believes its corporate “red lines” are crossed.

• Anthropic had signed a $200 million Pentagon contract but later said it did not want its AI used for mass surveillance of Americans or in lethal weapons targeting and firing decisions.

Apple cracks down on “vibe coding” apps LINK

• Apple is pushing back on “vibe coding” apps that let people build software by typing text prompts into an AI system, telling some developers their apps violate existing App Store rules.

• Apple cites App Store Guideline 2.5.2, which says apps cannot download, install, or execute code that introduces or changes features or functionality after passing through the review process.

• A possible fix for at least one affected app is generating previews in a browser instead of inside the vibe coding app itself, which could satisfy Apple’s existing rules.

Nvidia projects $1 trillion in AI chip sales by 2027 LINK

• Nvidia CEO Jensen Huang said at the company’s GTC conference that the company expects to sell $1 trillion worth of Blackwell and Rubin chips by the end of 2027.

• Nvidia unveiled the Groq 3 LPX rack, combining 72 Vera Rubin servers with 256 new language processing units, designed specifically for inference computing rather than training AI models.

• The company also announced partnerships for autonomous driving with BYD, Geely Auto, Hyundai, and Nissan, plus a coalition of software companies working on frontier open-sourced AI models.

Nvidia unveils AI chip for orbital data centers LINK

• Nvidia announced computing platforms designed for orbital data centers at its GTC 2026 conference, marking a significant move to bring artificial intelligence processing into space environments.

• The company’s Vera Rubin Space-1 Module, which includes IGX Thor and Jetson Orin chips, is engineered for size-, weight- and power-constrained environments and will fly on missions led by multiple partners.

• Cooling remains a key engineering hurdle because space lacks convection, while SpaceX — which acquired xAI for $1.25 trillion — has asked the FCC to launch 1 million satellites for AI centers.

OpenAI cuts side projects to focus on core business LINK

• OpenAI is shifting away from launching many products at once and will now focus its resources on two core areas: coding tools and business customers, according to a Wall Street Journal report.

• Fidji Simo, OpenAI’s CEO of Applications, called Anthropic’s success in enterprise and coding a “wake-up call” and told employees the company “cannot miss this moment because we are distracted by side quests.”

• Products like the Sora video generator and an agent mode struggled after launch, with Sora’s usage going flat after briefly hitting number one in the Apple App Store and the agent losing most users.

Data centers are turning to $200,000 robot dogs to guard the facilities LINK

• Companies like Boston Dynamics and Ghost Robotics are selling four-legged robots, priced from $165,000 to $300,000, to patrol and inspect AI data centers around the clock.

• Boston Dynamics says interest from data center clients has jumped sharply in the past year, with its Spot robot detecting temperature changes, leaks, and unusual noises across server halls.

• Ghost Robotics’ Vision 60 handles external perimeter security, and both companies say the robots are meant to augment human guards rather than replace them entirely.

Meta signs $27 billion AI infrastructure deal with Nebius LINK

• Meta has agreed to pay up to $27 billion over five years for AI infrastructure from cloud provider Nebius Group, marking one of the largest single contracts the company has ever signed.

• Nebius will provide $12 billion in dedicated capacity starting in early 2027, while Meta also committed to buying up to $15 billion in additional capacity built for third-party customers.

• Meta and its biggest tech peers are expected to spend around $650 billion in 2026 on data centers and related infrastructure, as the company competes with OpenAI and Google.

OpenAI advisers alarmed by adult content plans LINK

• OpenAI’s own advisory council on well-being reacted with alarm to the company’s plans for an “adult mode” in ChatGPT, with one member warning it risked creating a “sexy suicide coach.”

• The company delayed adult mode partly because its age-prediction system was misclassifying minors as adults about 12% of the time, potentially exposing millions of under-18 users to erotic chats.

• OpenAI staffers internally identified risks including compulsive use, emotional overreliance on the chatbot, a drive toward more extreme content, and crowding out offline social and romantic relationships.

Pokémon Go data now guides delivery robots LINK

• Niantic Spatial, an AI spinout formed in 2025, is repurposing location data and street-level imagery collected from Pokémon Go players to help Coco Robotics guide its sidewalk delivery robots through dense cities.

• The visual positioning system was trained on roughly 30 billion crowdsourced images from Pokémon Go and Ingress, letting robots pinpoint their location to within a few centimeters using cameras instead of GPS.

• Coco operates about a thousand sidewalk robots across US and European cities, and the company plans to fuse GPS with Niantic Spatial’s camera-based localization to improve reliability on routes.

ByteDance pauses Seedance 2.0 LINK

• ByteDance has delayed the worldwide release of its AI video model Seedance 2.0, which launched in China in February, as the company works to avoid further legal problems.

• Videos generated by the model went viral, including a clip of Tom Cruise fighting Brad Pitt, which drew a wave of cease-and-desist letters from Hollywood studios like Disney.

• Disney’s lawyers accused ByteDance of a “virtual smash-and-grab” of its IP, and the company promised to introduce stronger safeguards for intellectual property before expanding access.

Meta plans to cut 20% of its workforce LINK

• Meta is considering cutting around 20% of its workforce, which would eliminate roughly 15,800 jobs from its nearly 79,000 employees, though no timeline or final number has been set.

• The planned reductions come as Meta ramps up AI spending, with up to $600 billion earmarked for data center infrastructure by 2028 and large compensation packages to recruit top researchers.

• CEO Mark Zuckerberg has pointed to AI-driven efficiency gains, saying projects that used to require big teams can now be done by a single person, echoing similar cuts across the tech industry.

Travis Kalanick launches new robotics startup LINK

• Uber founder Travis Kalanick has launched a new robotics company called Atoms, which will operate in the food, mining, and transportation industries and absorbs his existing ghost kitchen company, CloudKitchens.

• Kalanick said Atoms will build a “wheelbase for robots” focused on specialized machines rather than humanoids, and he is close to acquiring Pronto, an autonomous vehicle startup for industrial and mining sites.

• The Information reported that Kalanick has “major backing” from Uber and has told people he wants to be more aggressive in rolling out self-driving technology than Waymo, though Atoms’ website does not mention Uber.

LISTEN TO ADS-FREE Audio of this episode at

https://djamgamind.com

or

https://podcasts.apple.com/us/channel/djamgamind/id6760446113

🚀 Welcome to AI Unraveled. Today, we look at the realities of AI integration. From Anthropic saving heart failure patients in Texas to Palantir becoming the permanent weapons-targeting system for the US Military, the tech is now structurally embedded into our most critical systems. Meanwhile, the users are pushing back on the hype.

This episode is made possible by our sponsors:

🎙️ DjamgaMind: Tired of the ads? Get the forensic version of this news. Join our Ads-FREE Premium Feed at DjamgaMind. Technical, deep, and uninterrupted. 👉 Switch to Ads-Free: https://DjamgaMind.com

In Today’s Briefing:

• The Healthcare Blueprint: How the University of Texas used Claude to scan 2 million patient records, fixing critical care gaps without replacing human doctors.

• Palantir’s Pentagon Lock-In: Maven AI becomes an official “program of record” for US military weapons targeting.

• OpenAI’s Massive Expansion: Doubling the workforce to 8,000, hiring “technical ambassadors,” and unifying ChatGPT, Codex, and Atlas into a desktop Superapp.

• The User Backlash: Microsoft walks back AI clutter in Windows 11, reducing Copilot integrations after massive user complaints.

• Google’s Headline Problem: How Google Search is using AI to rewrite news headlines, often changing the meaning of the source material.

• SpaceX’s Orbital Monopoly: The Space Force shifts crucial GPS satellite launches away from a failing ULA directly to SpaceX.

• Elon Musk’s Twitter Trial: A California jury finds Musk intentionally misled shareholders during the 2022 acquisition.

Strategic Signal: Institutional Integration vs. Consumer Fatigue.

Credits: Created and produced by Etienne Noumen.

Keywords: OpenAI Workforce Expansion, Anthropic Claude Healthcare, UTMB Medical AI, Palantir Maven Pentagon, US Military AI Targeting, SpaceX GPS Launches, Elon Musk Twitter Trial, Windows 11 AI Clutter, Microsoft Copilot Removal, Google AI Headlines, Nvidia Nemotron-Cascade 2, Trump AI Policy Framework, Djamgatech, AI Unraveled.

🚀 FOR LEADERS: DjamgaMind Audio Intelligence

Don’t Read the Regulation. Listen to the Risk. Drowning in dense legal text? DjamgaMind turns 100-page healthcare/energy/finance mandates into 5-minute executive audio briefings. Whether navigating Bill C-59 or HIPAA compliance, our AI agents decode the liability so you don’t have to.

👉 Start your briefing: https://DjamgaMind.com/regulations

🔗 RESOURCES & CAREERS

Find AI Jobs (Mercor): Apply Here – https://work.mercor.com/?referralCode=82d5f4e3-e1a3-4064-963f-c197bb2c8db1

⚗️ PRODUCTION NOTE: We Practice What We Preach.

AI Unraveled is produced using a hybrid “Human-in-the-Loop” workflow. While all research, interviews, and strategic insights are curated by Etienne Noumen, we leverage advanced AI voice synthesis for our daily narration to ensure speed, consistency, and scale.

OpenAI to nearly double workforce LINK

• OpenAI plans to nearly double its workforce from 4,500 to 8,000 by the end of 2026, according to a Financial Times report citing two people with knowledge of the matter.

• Most of the new hires will work across product development, engineering, research, and sales, covering the core teams that build and sell OpenAI’s tools to customers.

• The company is also recruiting specialists focused on “technical ambassadorship,” a role designed to help businesses make better use of its existing products like ChatGPT.

Google is replacing news headlines with AI ones LINK

• Google is now using AI to rewrite news headlines that appear in its search results, sometimes changing their meaning, after previously doing something similar in its Google Discover news feed.

• The practice is not entirely new — Google has been altering headlines in search results for years — but recent examples show AI-rewritten headlines that poorly reflect the actual articles they link to.

• Google Search maintains a visual tone of being a neutral directory, but its AI Overviews and rewritten headlines can misrepresent source material, making results misleading for people who trust them.

Are you comfortable with an AI scanning your family’s medical records if it means catching a life-threatening issue your doctor didn’t have time to find?

We spend a lot of time looking at the dark side of the tech and the data tracking, the automation coming for our jobs, the companies prioritizing profit over privacy.

But if we are going to look at the whole board honestly, we have to acknowledge when the technology actually does what it was supposed to do; protect us.

A real problem right now is the collapsing healthcare system. In Texas alone, severe doctor shortages means that an estimated 4 to 6 million patients miss out on life-saving treatments every year. The doctors don’t have the hours to dig through disorganized medical files to connect the dots.

The University of Texas Medical Branch deployed an AI platform powered by Anthropic’s Claude to fix exactly that.

Here is why this matters, and why it’s a blueprint for how this tech should be used:

It’s Not a Doctor Replacement: The AI is not making medical decisions. It is doing the heavy administrative lifting, scanning a population of over 2 million patients to find the ones slipping through the cracks.

The AI flags the data and provides the exact source files. A human doctor still has to review the chart, validate the findings, and make the actual medical call.

In just the first month of deployment, the system found that up to a third of heart failure patients had gaps in their care and were eligible for better, life-saving treatments.

This technology is forced to operate with strict guardrails, safety protocols, and traceability. It isn’t a toy meant to strip away human agency. It’s a reinforced tool being used to give doctors their time back so they can actually save lives.

We have to call out Big Tech when they cross the line, but we also need to recognize when a system is actually built to work for us, instead of against us.

Are you comfortable with an AI scanning your family’s medical records if it means catching a life-threatening issue your doctor didn’t have time to find?

Pentagon to adopt Palantir AI as core US military system, memo says

“Palantir’s (PLTR.O), opens new tab Maven artificial intelligence system will become an official program of record, Deputy Secretary of Defense Steve ​Feinberg said in a letter to Pentagon leaders, a move that locks in long-term use of Palantir’s weapons-targeting technology across ‌the U.S. military.

In the March 9 letter to senior Pentagon leaders and U.S. military commanders, Feinberg said embedding Palantir’s Maven Smart System would provide warfighters “with the latest tools necessary to detect, deter, and dominate our adversaries in all domains”.”

https://www.reuters.com/technology/pentagon-adopt-palantir-ai-as-core-us-military-system-memo-says-2026-03-20/

SpaceX dominates US military and NASA contracts LINK

• SpaceX has become the go-to launch provider for the US military and NASA, with the Space Force again turning to a Falcon 9 rocket after ULA failed to meet its GPS satellite launch schedule.

• ULA’s Vulcan rocket is grounded for the second time in under two years because its solid rocket boosters suffered the same type of failure on two of its four flights.

• The Space Force shifted all four final GPS Block III satellite launches from ULA to SpaceX starting in 2024, giving ULA rights to a classified military mission in 2028 instead.

Elon Musk misled Twitter shareholders, jury finds LINK

• A California civil jury found that Elon Musk intentionally misled Twitter shareholders in 2022 when he publicly questioned the platform’s bot numbers while trying to back out of his $44 billion acquisition.

• Investor Giuseppe Pampena sued on behalf of former Twitter shareholders who sold stock at a loss after Musk’s tweet caused an 8% decline in share price between May and October 2022.

• Damages could reach $2.6 billion according to Pampena’s attorney, though the exact amount is not yet clear — and it’s a relatively small sum given Musk’s estimated $660 billion net worth.

Microsoft announces sweeping Windows changes LINK

• Microsoft has announced a long list of changes to Windows 11 after years of growing user complaints about AI clutter, unreliable updates, poor performance, and missing features like taskbar customization.

• The company says it will reduce unnecessary Copilot entry points in apps like Snipping Tool, Photos, Widgets, and Notepad, responding to near-universal user feedback asking Microsoft to stop pushing AI features.

• Other promised changes include movable taskbar positions, fewer automatic restarts during updates, faster File Explorer performance, and better testing through the Windows Insider Program before builds ship publicly.

What Else Happened in AI?

1. Trump administration unveils national AI policy framework to limit state power. [LINK]

2. Google Search is now using AI to replace headlines.[LINK]

3. OpenAI to create desktop super app, combining ChatGPT app, browser and Codex app.[LINK]

4. NVIDIA Releases Nemotron-Cascade 2: An Open 30B MoE with 3B Active Parameters, Delivering Better Reasoning and Strong Agentic Capabilities.[LINK]

5. Nvidia “confirms” DLSS 5 relies on 2D frame data as testing reveals hallucinations [LINK]

6. Closure of Strait of Hormuz is ‘greatest global energy security threat in history,’ warns IEA chief [LINK]

LISTEN TO ADS-FREE Audio of this episode at https://djamgamind.com/daily

🚀 Welcome to AI Unraveled. Today, the AI industry gets physical. Jeff Bezos is raising the largest fund in history to automate heavy industry, while the U.S. government busts a massive $2.5 billion Silicon Valley smuggling ring supplying Nvidia chips to China.

This episode is made possible by our sponsors:

🎙️ DjamgaMind: Tired of the ads? Get the forensic version of this news. Join our Ads-FREE Premium Feed at DjamgaMind. Technical, deep, and uninterrupted. 👉 Switch to Ads-Free: DjamgaMind.com

In Today’s Briefing:

• Project Prometheus: Jeff Bezos seeks $100 billion to acquire and automate chipmaking, aerospace, and defense companies.

• The Silicon Black Market: Supermicro’s co-founder arrested for smuggling $2.5B in restricted Nvidia AI servers to China.

• The OpenAI Superapp: Consolidating ChatGPT, Codex, and Atlas into a single desktop execution environment.

• Cursor Composer 2: How an application-layer startup built an in-house model that beats Opus 4.6 at 1/20th the cost.

• Anthropic’s Claude Interviewer: Surveying 81,000 people in 70 languages in a massive proof-of-concept for AI qualitative research.

• Microsoft MAI-Image-2: Mustafa Suleyman’s team hits the Top 5 on the Arena leaderboard, reducing reliance on OpenAI.

• The Data Harvest: DoorDash pays couriers to film for robotics training; the FBI resumes buying citizen location data.

Credits: Created and produced by Etienne Noumen.

Keywords: Jeff Bezos Project Prometheus, $100B AI Fund, Supermicro Wally Liaw Arrest, Nvidia Chip Smuggling, OpenAI Desktop Superapp, Cursor Composer 2, Microsoft MAI-Image-2, Anthropic Claude Interviewer, DoorDash Tasks App, AI Manufacturing, Geopolitical Tech, DjamgaMind, AI Unraveled.

🚀 FOR LEADERS: DjamgaMind Audio Intelligence

Don’t Read the Regulation. Listen to the Risk. Drowning in dense legal text? DjamgaMind turns 100-page healthcare/energy/finance mandates into 5-minute executive audio briefings. Whether navigating Bill C-59 or HIPAA compliance, our AI agents decode the liability so you don’t have to.

👉 Start your briefing: https://DjamgaMind.com/regulations

🔗 RESOURCES & CAREERS

Find AI Jobs (Mercor): Apply Here – https://work.mercor.com/?referralCode=82d5f4e3-e1a3-4064-963f-c197bb2c8db1

⚗️ PRODUCTION NOTE: We Practice What We Preach.

AI Unraveled is produced using a hybrid “Human-in-the-Loop” workflow. While all research, interviews, and strategic insights are curated by Etienne Noumen, we leverage advanced AI voice synthesis for our daily narration to ensure speed, consistency, and scale.

OpenAI is planning a desktop ‘superapp’ LINK

• OpenAI plans to combine its Mac apps for ChatGPT, Codex, and Atlas into one single “superapp,” according to a report from The Wall Street Journal confirmed by an OpenAI spokesperson.

• Chief of Applications Fidji Simo told her team in an internal memo that OpenAI was “spreading our efforts across too many apps and stacks,” which slowed development and hurt quality.

• OpenAI expects to first add agentic features to Codex for productivity tasks beyond coding, then merge ChatGPT and the Atlas browser into the superapp, while the mobile app stays unchanged.

Amazon is making an Alexa phone LINK

• Amazon is working on a new smartphone codenamed “Transformer,” its first attempt at a phone in over 11 years since the failed Fire Phone, according to a Reuters report citing anonymous sources.

• The device would feature personalized tools for Amazon Shopping, Prime Video, and Prime Music, with AI features and Alexa support meant to push customers toward the company’s AI products.

• Development is led by a unit called ZeroOne, run by J Allard, a former Microsoft executive who helped create the Xbox, inside Amazon’s Devices and Services division.

Jeff Bezos seeks $100 billion for AI manufacturing fund LINK

• Jeff Bezos is reportedly trying to raise $100 billion for a new fund that would acquire companies across major industrial sectors and then modernize and automate them using AI.

• The fund is tied to Project Prometheus, a startup Bezos co-founded with former Google executive Vik Bajaj, which launched with $6.2 billion to build AI models for manufacturing and engineering.

• Bezos recently traveled to Singapore and the Middle East to raise money, with plans to acquire companies in areas like aerospace, chipmaking, and defense that would adopt Prometheus’ models.

Supermicro’s co-founder arrested for smuggling $2.5B in GPUs to China LINK

• Federal prosecutors in New York have charged Super Micro Computer co-founder Yih-Shyan “Wally” Liaw and two associates with illegally diverting roughly $2.5 billion in AI servers to China.

• A Southeast Asian middleman company created fake paperwork and used “dummy” servers at storage facilities to fool the server maker’s compliance team while real servers were shipped to China.

• The servers contained Nvidia chips subject to strict U.S. export controls barring their sale to China without a license, controls designed to protect national security and foreign policy interests.

White House releases national AI framework

• The White House published a national AI framework that asks Congress to override state laws governing how AI models are developed and to avoid creating any new federal agencies for AI regulation.

• The framework calls on Congress to protect children by keeping state bans on AI-generated child sexual abuse material, adding age-gating requirements for models, and giving parents tools for safeguards.

• Senate Majority Leader John Thune acknowledged that even Republicans worry about trampling state rights, and past efforts to block states from regulating AI have already failed twice in Congress.

Anthropic surveys 81k people on AI hopes, fears

Image source: Anthropic

The Rundown: Anthropic just released what it says is the biggest qualitative AI attitudes study ever, using Claude to interview 81k of its users across 159 countries about where they think the tech is headed and what scares them about getting there.

The details:

• Anthropic introduced Claude Interviewer in December, building a special version of Claude that ran open-ended conversations in 70 languages.

• Professional excellence was the top-reported hope, with freeing up time, financial independence, and broader life management frequently mentioned.

• Fear of AI getting things wrong outranked every other concern, with job anxiety, losing personal agency, and over-reliance close behind.

• AI sentiment varied by region: India and South America skewed above average, while the U.S., Europe, Japan, and South Korea ran neutral or below.

Why it matters: AI’s favorability numbers have cratered in mainstream polls, but Anthropic’s study adds nuance that those surveys miss. Almost as notable is Claude running 80K in-depth interviews across 70 languages in a single week, a wildly strong proof of concept for the tech as a research tool that simply didn’t exist a year ago.

Cursor’s coding model cuts costs near the frontier

Anysphere, the company behind AI code editor Cursor, just shipped Composer 2, a third-generation in-house model that is competitive with frontier coding models from OpenAI and Anthropic at a fraction of the cost per task.

The details:

• Composer 2 topped Opus 4.6 on the independent Terminal-Bench 2.0 (61.7% vs 58%) and sits within 5 points of GPT-5.4 on Cursor’s own CursorBench.

• At $7.50/M output tokens on its fast tier, Composer 2 costs roughly 1/10th of GPT-5.4 and 1/20th of Opus 4.6 at comparable speeds.

• Composer’s scores on the company’s internal CursorBench have climbed from 38% to 61.3% across three model generations shipped since October.

Why it matters: Cursor quickly went from harnessing other top AI models to building one of its own at this price point. Nearing the frontier as an application-layer company is an impressive feat, and the speed, cost, and performance of Composer 2 could change the math for developers paying full price for coding with GPT-5.4 or Opus 4.6.

Microsoft AI’s image model climbs leaderboards

Image source: Microsoft

Microsoft’s AI Superintelligence team just released MAI-Image-2, a text-to-image model that landed at No. 5 on the Arena AI leaderboard — marking the strongest release yet for Mustafa Suleyman’s lab.

The details:

• Arena.ai ranked MAI-Image-2 at No. 5 overall, trailing just Gemini (several variants) and GPT Image-1.5 with strong upgrades in photorealism, 3D, and art.

• The biggest jump from its predecessor came in text rendering, up 115 points, with drastically improved performance on posters, slides, and infographics.

• MAI-Image-2 is free to try in Microsoft’s MAI Playground for U.S. users, with Copilot, Bing, and API access on its Foundry platform rolling out soon.

• The release comes amid Microsoft’s AI leadership shuffle, with Suleyman shifting away from Copilot to focus solely on frontier model work.

Why it matters: Microsoft has been signaling its desire to reduce its reliance on OpenAI and truly compete with its own models, and MAI-Image-2 is the strongest step yet in that direction. But the legacy tech giant still has a major uphill battle to gain market share from the already well-entrenched frontier options at the top.

What Else Happened in AI on March 20th 2026?

Google rolled out upgrades that turn its AI Studio into a one-stop vibe-coding app builder, pairing a new Antigravity coding agent with built-in backends and user login.

Jeff Bezos is reportedly raising a $100B fund to buy chip, defense, and aerospace manufacturers, with plans to use them for his secretive AI startup, Project Prometheus.

Perplexity introduced Health, a new feature allowing users to securely connect health apps, wearables, and data to its Computer agentic system.

DoorDash launched a new ‘Tasks’ app, paying its couriers to capture video and data from everyday tasks and conversations for AI and robotics training.

OpenAI announced the acquisition of open-source developer tool startup Astral, folding the company’s staff into its Codex team.

Meta launched an AI support assistant across FB and IG for 24/7 support, also previewing advanced content enforcement systems that catch 5K daily scam attempts.

Meta to Deploy AI to Police Facebook and Instagram Content [LINK]

LISTEN TO ADS-FREE Audio of this episode at https://djamgamind.com/daily

🚀 Welcome to AI Unraveled. Today, the “Vibe” movement officially swallows the design world as Google launches its Stitch infinite canvas. But as the tools get easier, the risks get higher—Meta is dealing with a “Sev 1” rogue agent incident, and Microsoft is threatening legal action over a massive $50 billion Amazon-OpenAI deal.

This episode is made possible by our sponsors:

🛑 AIRIA: As agents like Manus move onto your local machine to manage files and run terminal commands, the risk of “Agentic Drift” is real. AIRIA provides the governance layer to ensure these powerful local tools stay within your enterprise compliance boundaries. 👉 Govern the desktop: https://airia.com/request-demo/?utm_source=AI+Unraveled+&utm_medium=Podcast&utm_campaign=Q1+2026

🎙️ DjamgaMind: Tired of the ads? Get the forensic version of this news. Join our Ads-FREE Premium Feed at DjamgaMind. Technical, deep, and uninterrupted. 👉 Switch to Ads-Free: DjamgaMind.com

In Today’s Briefing:

• Google Stitch & Vibe Design: High-fidelity UI from natural language and the new voice-enabled design partner.

• The Meta Rogue Agent: A “Sev 1” data exposure incident and the risk of unauthorized autonomous actions.

• Microsoft vs. Amazon/OpenAI: A potential lawsuit over a $138B cloud commitment and the “Frontier” agent platform.

• OpenAI Acquires Astral: The move to own the Python dev stack (uv, Ruff, ty).

• Nvidia’s 4th Scaling Law: “Agentic Scaling”—where swarms of agents talking to agents drive the next trillion parameters.

• MiniMax M2.7: The rise of self-evolving AI that writes its own training code.

• Apple’s $1B AI Toll: How Apple is making a fortune from App Store fees despite trailing in AI products.

• FBI Data Harvesting: The resumption of buying Americans’ location data from brokers.

Keywords: Google Stitch, Vibe Design, Meta Rogue AI, OpenAI Astral, Microsoft Amazon Lawsuit, Agentic Scaling, Nvidia GTC 2026, FBI Location Data, Apple AI Revenue, MiniMax M2.7, Uber Rivian Robotaxi, DjamgaMind, AI Unraveled.

🚀 FOR LEADERS: DjamgaMind Audio Intelligence

Don’t Read the Regulation. Listen to the Risk. Drowning in dense legal text? DjamgaMind turns 100-page healthcare/energy/finance mandates into 5-minute executive audio briefings. Whether navigating Bill C-59 or HIPAA compliance, our AI agents decode the liability so you don’t have to.

👉 Start your briefing: https://DjamgaMind.com/regulations

🔗 RESOURCES & CAREERS

Find AI Jobs (Mercor): Apply Here –

https://work.mercor.com/?referralCode=82d5f4e3-e1a3-4064-963f-c197bb2c8db1

⚗️ PRODUCTION NOTE: We Practice What We Preach.

AI Unraveled is produced using a hybrid “Human-in-the-Loop” workflow. While all research, interviews, and strategic insights are curated by Etienne Noumen, we leverage advanced AI voice synthesis for our daily narration to ensure speed, consistency, and scale.Google unveils new ‘vibe design’ tool LINK

• Google is redesigning its AI tool Stitch into what it calls a “vibe design” canvas, where anyone can type natural language descriptions and get high-fidelity UI designs without starting from wireframes.

• The updated Stitch features an infinite canvas, a design agent that reasons across a project’s full history, and an Agent manager that tracks progress on multiple ideas in parallel.

• Stitch now includes voice input for real-time design critiques and updates, plus an MCP server and SDK that let teams export designs to developer tools like AI Studio and Antigravity.

FBI is buying Americans’ location data LINK

• FBI director Kash Patel confirmed to lawmakers on Wednesday that the agency has resumed buying Americans’ location data from data brokers to support federal investigations, reversing a pause in place since 2023.

• Senator Ron Wyden called the practice an “outrageous end-run around the Fourth Amendment,” since government agencies normally need a judge to authorize a search warrant before obtaining private information about people.

• Wyden and other lawmakers recently introduced a bipartisan bill called the Government Surveillance Reform Act, which would require federal agencies to get a court-authorized warrant before purchasing Americans’ information from data brokers.

Meta is having trouble with rogue AI agents LINK

• An AI agent at Meta went rogue, accidentally exposing sensitive company and user data to employees who were not authorized to access it for about two hours.

• The incident started when an engineer asked an AI agent to help answer a technical question, and the agent posted a response on an internal forum without getting the engineer’s permission first.

• Meta rated the incident a “Sev 1,” its second-highest severity level, and a separate recent case saw a Meta director’s OpenClaw agent delete her entire inbox despite instructions to confirm actions.

Apple is behind in AI and still making a fortune from it LINK

• Apple’s own AI products trail competitors, but the company is on track to pass $1 billion in generative AI revenue by 2026, mostly from App Store fees charged to chatbot apps like ChatGPT.

• Generative AI apps paid Apple close to $900 million in 2025, with three-quarters coming from ChatGPT alone, and monthly revenue peaking at $101 million in August before dropping off.

• Apple spends far less on chips and data centers than rivals, and Siri still runs on outdated technology — its new version will initially rely on Google’s Gemini after internal setbacks.

OpenAI acquires Python toolmaker Astral LINK

• OpenAI has announced plans to acquire Astral, the company behind popular Python development tools like uv, Ruff, and ty that many developers already depend on daily.

• The deal signals OpenAI’s push to embed Codex directly into real development workflows rather than just generating code snippets, using Astral’s tools for linting, dependency management, and type safety.

• OpenAI says it will keep supporting Astral’s open source projects, but developers are understandably nervous about whether these widely used neutral tools will slowly become optimized for one platform.

Uber invests $1.25 billion in Rivian for 50,000 robotaxis LINK

• Uber is investing up to $1.25 billion in Rivian to purchase as many as 50,000 autonomous R2 SUVs, starting with a $300 million commitment and an initial order of 10,000 robotaxis.

• The companies plan to launch the fleet in San Francisco and Miami by 2028, expanding to 25 cities across the U.S., Canada, and Europe by the end of 2031.

• Rivian faces big challenges: the R2 SUV isn’t in production yet, its Georgia factory is still under construction, and its self-driving system has not been tested for robotaxi use.

Google brings ‘vibe design’ to its AI UI canvas

Image source: Google

The Rundown: Google just overhauled Stitch, turning its AI UI design tool into a voice-enabled infinite canvas that takes users from a rough idea to a clickable prototype — and coining the term “vibe design” for its new development workflow.

The details:

• Stitch now runs on an infinite canvas where users feed in images, code, or briefs, plus an agent manager that juggles multiple design directions at once.

• A new voice feature in preview turns the tool into a hands-free design partner, able to take direction and make live edits mid-chat.

• Instant prototyping can turn static screens into interactive prototypes in seconds, while auto-generating logical next screens for the UI flow.

• A new DESIGN.md format lets teams port design rules between Stitch and coding tools, with each project getting a style system out of the box.

Why it matters: Design has already changed completely in the AI era, but agentic capabilities are taking things to new levels. Stitch’s upgrades let users move at AI-native speeds with easy integration into existing workflows, and a “vibe design” ethos that puts strong creation in reach in the same way “vibe coding” did for development.

MiniMax’s new M2.7 helped build itself

Image source: MiniMax

The Rundown: MiniMax launched M2.7, what the company calls its “first model which deeply participated in its own evolution” — writing its own training code, running autonomous improvement loops, and matching the scores of top Western models.

The details:

• Early M2.7 versions were put to work on their own training, including writing improvement routines and tuning how the model learns from feedback.

• M2.7 ran 100+ cycles of autonomously analyzing its mistakes, rewriting code, and testing fixes — showing a 30% accuracy boost on internal benchmarks.

• On coding, M2.7 hit 56.2% on SWE-Pro and 55.6% on VIBE-Pro, putting it near Opus 4.6 and GPT-5.3-Codex for agentic engineering work.

Why it matters: Self-evolving AI is one of the bigger next steps forward, and while we’ve heard rumblings from OAI, Anthropic, Google, and xAI, MiniMax is one of the first labs out of China to openly make the claim. Future models will all likely be training and improving themselves, but for now, we’re just watching the feature emerge.

Nvidia chases scale as user-owned AI emerges

At GTC 2026, Nvidia committed to unlocking new levels of compute for the industry’s leading innovators. At the same time, another narrative about bottom-up innovation was unfolding in other corners of GTC.

On a quiet veranda at the San Jose Convention Center, The Deep View spoke to Illia Polosukhin, co-founder of Near, about “user-owned AI,” where the technology is decentralized, all your data is kept separate from the models, and agents are treated as a secure, separate OS.

While Near doesn’t have the marketing muscle of OpenAI, Anthropic, and other leading players in the AI ecosystem, it’s counting on the kind of bottom-up, word-of-mouth momentum that turned OpenClaw into a viral hit, aiming to spread the mantra about a safer and more empowered alternative to the way most people are experiencing today’s AI.

Meanwhile, Nvidia CEO Jensen Huang emphasized that more compute means more intelligence, more intelligence means more value, and more value means more revenue. Huang told a full capacity audience at the San Jose Civic on Wednesday that “intelligence is directly correlated to the amount of compute that you have.”

Amid the veritable firehose of announcements at the conference, Nvidia posed an entirely new scaling law to continue pushing the narrative of more: Agentic scaling.

• Following pretraining scaling, post-training scaling and test-time scaling, this new scaling law involves AI not just talking to humans, but to other AIs, vastly increasing demand for low-latency, large-context inference.

• These multi-agent systems, Nvidia claims, will unlock multi-trillion parameter models, turning daylong requests into hours. To do so, however, these systems need to get a lot faster, with Nvidia emphasizing the need to deliver tokens 15 times faster and with 10-times larger models.

“The best open [models] are trillion-parameter models, and … if you look at the proprietary models, they’re more than a trillion parameters,” Kari Briski, VP of generative AI software at Nvidia, told The Deep View. “Now, the fourth scaling law is not just about one reasoning model. It’s about a swarm of agents with subagents. Agents talking to agents.”

Polosukhin is also all-in on agents, including the claw revolution. Near has IronClaw, which offers a more secure version of OpenClaw, similar to NanoClaw and Nvidia’s NemoClaw. The company also launched a secure agent marketplace designed to run agents that can offload work and earn money, all within its decentralized, security-first platform.

Microsoft ‘weighing’ legal action over Amazon-OAI deal

Elevate Your Career with AI & Machine Learning For Dummies PRO
Ready to accelerate your career in the fast-growing fields of AI and machine learning? Our app offers user-friendly tutorials and interactive exercises designed to boost your skills and make you stand out to employers. Whether you're aiming for a promotion or searching for a better job, AI & Machine Learning For Dummies PRO is your gateway to success. Start mastering the technologies shaping the future—download now and take the next step in your professional journey!



Download the AI & Machine Learning For Dummies PRO App:
iOS - Android
Our AI and Machine Learning For Dummies PRO App can help you Ace the following AI and Machine Learning certifications:

• AWS Certified AI Practitioner (AIF-C01): Conquer the AWS Certified AI Practitioner exam with our AI and Machine Learning For Dummies test prep. Master fundamental AI concepts, AWS AI services, and ethical considerations.


• Azure AI Fundamentals: Ace the Azure AI Fundamentals exam with our comprehensive test prep. Learn the basics of AI, Azure AI services, and their applications.


• Google Cloud Professional Machine Learning Engineer: Nail the Google Professional Machine Learning Engineer exam with our expert-designed test prep. Deepen your understanding of ML algorithms, models, and deployment strategies.


• AWS Certified Machine Learning Specialty: Dominate the AWS Certified Machine Learning Specialty exam with our targeted test prep. Master advanced ML techniques, AWS ML services, and practical applications.


• AWS Certified Data Engineer Associate (DEA-C01): Set yourself up for promotion, get a better job or Increase your salary by Acing the AWS DEA-C01 Certification.

Image source: Lovart / The Rundown

Microsoft is reportedly considering legal action against Amazon and OpenAI, with sources telling the FT it will sue if a new $50B cloud deal between the two companies violates its exclusive contract to host the startup’s models on Azure.

The details:

• The dispute is over Frontier, OpenAI’s new enterprise agent platform — also the anchor of a broader deal committing $138B in cloud spending to AWS.

• Microsoft dropped its exclusive hosting lock on OAI in October, but kept a clause that forces all developer access to OAI models to run through Azure.

• The FT’s source said “We know our contract… We will sue them if they breach it”, with another source adding “the last thing OAI needs is another court case”.

• OAI reportedly signed a new deal with AWS last week, which opened the door for the company’s deployment with the Pentagon.

Why it matters: Just when you thought the Microsoft-OAI relationship couldn’t get more awkward, now there’s a potential lawsuit to worry about. The FT source is right that OAI can’t afford another legal fight with an IPO looming and the Musk trial, but the Microsoft partnership itself also continues to feel like a headache for the AI giant.

What Else Happened in AI on March 19th 2026?

The U.S. Dept. of Defense filed a 40-page rebuttal to Anthropic’s lawsuits, arguing its safety limits make it an “unacceptable risk to national security” during war operations.

Xiaomi released MiMo-V2-Pro, a model that topped OpenRouter’s charts under an ‘Hunter Alpha’ codename and excels in agent-related tasks and OpenClaw usage.

Microsoft acquired the full team behind Cove, a collaborative AI interface startup, with the company saying its “ideas will live on” at the tech giant.

Midjourney rolled out a preview of its new V8 image model, coming with improved speed, detail, and text rendering, garnering mixed reactions in early testing.

LISTEN TO ADS-FREE Audio of this episode at https://djamgamind.com/daily

This episode is made possible by our sponsors:

🛑 AIRIA: As agents like Manus move onto your local machine to manage files and run terminal commands, the risk of “Agentic Drift” is real. AIRIA provides the governance layer to ensure these powerful local tools stay within your enterpr…

Read more