Listen to to Full Audio here

🚀 Welcome to AI Unraveled (February 6th, 2026): Your strategic briefing on the business, technology, and policy reshaping artificial intelligence.

Today is arguably the biggest day in AI development this year. OpenAI and Anthropic have simultaneously released their flagship coding and reasoning models, with OpenAI’s new model explicitly used to “build itself.” Meanwhile, Alphabet and Amazon have committed a combined $385 billion to AI infrastructure in 2026 alone.

Strategic Pillars & Key Topics:

⚔️ The Model Wars: Dueling Releases

• OpenAI GPT-5.3-Codex: A self-improving model that helped debug its own training runs. It scores nearly double its predecessor on desktop control benchmarks and is rated “High” risk for cybersecurity.

• Anthropic Opus 4.6: Features “Agent Teams” that split projects among multiple AI agents, a 1-million-token context window, and direct integration into PowerPoint and Excel.

• The “Self-Improvement” Era: Both labs now confirm their AI models are helping design their successors.

💰 The Trillion-Dollar Infrastructure Race

• Alphabet’s $185B Bet: Google’s parent company announces a CapEx plan of up to $185 billion for 2026 to support data centers and servers.

• Amazon’s $200B Plan: CEO Andy Jassy commits $200 billion to AI, chips, and robotics in 2026, causing the stock to dip on missed profit expectations.

🤖 The Enterprise Agent Layer

• OpenAI “Frontier”: A new platform that lets enterprises manage “AI Coworkers” like new hires—complete with onboarding, performance reviews, and permissions.

• Amazon Alexa+: Reports suggest Amazon is in talks to integrate ChatGPT into Alexa+, combining OpenAI’s brains with Amazon’s voice reach.

⚖️ Culture & Conflict

• Super Bowl Drama: Sam Altman calls Anthropic “dishonest” after their Super Bowl ad mocks ChatGPT for inserting ads into conversations.

• Meta’s Comeback: Internal memos reveal Meta’s new “Avocado” model is outperforming rivals, signaling a recovery after a rocky 2025.

Credits: This podcast is created and produced by Etienne Noumen, Senior Software Engineer and passionate Soccer dad from Canada.

Keywords: GPT-5.3-Codex, Claude Opus 4.6, Alphabet CapEx, Amazon AI Spending, OpenAI Frontier, Agent Teams, Meta Avocado, Sam Altman Super Bowl Ad, Self-Improving AI, Alexa+ ChatGPT

🚀 Reach the Architects of the AI Revolution

Want to reach 60,000+ Enterprise Architects and C-Suite leaders? Download our 2026 Media Kit and see how we simulate your product for the technical buyer: https://djamgamind.com/ai

Connect with the host Etienne Noumen, Senior Software Engineer and passionate Soccer dad from Canada

LinkedIn: https://www.linkedin.com/in/enoumen/

Youtube: https://youtube.com/@enoumen

X: https://twitter.com/enoumen

OpenAI’s GPT-5.3-Codex helps build itself

Image source: OpenAI

OpenAI just rolled out GPT-5.3-Codex, a new flagship coding model that merges its best programming and reasoning capabilities into one faster package — while also serving as a key tool in its own training and deployment process.

The details:

• OpenAI said early versions of 5.3-Codex were used to find bugs in its own training runs, manage its rollout, and analyze evaluation results.

• Codex tops agentic coding benchmarks like SWE-Bench Pro and Terminal-Bench 2.0, topping Opus 4.6 by 12% on the latter just minutes after its release.

• On OSWorld, a benchmark testing AI control of desktop computers, the model scored 64.7% — nearly double the 38.2% from the prior Codex version.

• OpenAI flagged the model as its first “High” cybersecurity risk rating and committed $10M in API credits to fund defensive security research

Why it matters: The self-improvement angle here is the headline, with Anthropic’s Dario Amodei also recently saying Claude is helping design its own successor. Yesterday’s bickering over ads now looks childish compared to the true fight on the model frontier, with a big day of dueling releases out of both labs.

Anthropic’s Opus 4.6 with ‘agent teams’, 1M context

Image source: Anthropic

Anthropic released Claude Opus 4.6, the company’s new most powerful model — featuring multi-agent collaboration in Claude Code, a massive context window, and new Office integrations that put the AI directly inside PowerPoint.

The details:

• A new “agent teams” feature in Claude Code lets multiple AI agents split a single project and work simultaneously instead of handling steps one at a time

• Opus 4.6 brings a 1M token context window to Anthropic’s Opus tier for the first time, matching what Sonnet offers for heavy document and code work.

• New Excel and PowerPoint sidebars let Claude read users’ existing templates and build models or decks natively without copying and pasting between tools.

• 4.6 topped most agentic benchmarks, including a leap on ARC-AGI-2 to nearly 70% — though OAI’s Codex 5.3 reclaimed agentic coding highs minutes later.

Why it matters: It’s a big day for devs, with both Codex 5.3 and Opus 4.6 releases bringing major capability increases across the board. With time between upgrades getting shorter and the length of tasks models can take on continuing to move up the curve, the “AI is hitting a wall” crowd seems pretty quiet these days.

OpenAI’s Frontier to manage ‘AI coworkers’

Image source: OpenAI

OpenAI just launched Frontier, a new platform for enterprises to deploy and manage AI agents like new hires — complete with onboarding, permissions, and performance reviews across a company’s existing tech stack.

The details:

• Frontier connects to existing enterprise systems like CRMs and ticketing tools, letting agents pull context from across the business without migrations.

• Built-in eval and feedback loops let agents learn via experience, with OAI comparing it to onboarding a new employee with reviews and boundaries.

• Every agent operates under its own profile with scoped access and hard limits on what it can touch for enterprise and regulated control.

• HP, Oracle, State Farm, and Uber are among the first adopters, with OAI embedding engineers on-site to help teams get agents into production.

Why it matters: Anthropic and OAI have been battling over models and coding tools, but Frontier shows the fight is also bleeding into who controls the enterprise agent layer underneath. Model capabilities are making AI coworkers a reality in the near future, and the system that ultimately orchestrates them will be valuable real estate.

Alphabet plans record AI spending after strong earnings

• Alphabet announced it will spend between $175 billion and $185 billion on AI infrastructure this year, far exceeding the $119.5 billion analysts predicted and more than its combined spending over the past three years.

• CEO Sundar Pichai said the AI investments are necessary, stating the company is seeing AI investments and infrastructure drive revenue and growth across search, cloud, and other products.

• About 40% of technical infrastructure investments go toward data centers and networking equipment, while 60% go toward servers, as Alphabet races to compete against Amazon, Microsoft, and OpenAI.

Sam Altman clashes with Anthropic over Super Bowl ads

• OpenAI CEO Sam Altman and Anthropic got into a public fight after Anthropic released Super Bowl ads that mocked ChatGPT’s upcoming ad-supported free tier by showing chatbots inserting awkward product promotions into personal conversations.

• Altman admitted the ads were funny but called Anthropic “dishonest” and “authoritarian,” saying OpenAI would never insert ads into conversations, though the company has said it plans to show ads based on chat content.

• Altman claimed Anthropic only serves rich people, but both companies offer similar free and paid tiers, and both enforce content policies, making his accusations seem like an overreaction to a cheeky commercial.

OpenAI unveils Frontier, a platform to build ‘AI-coworkers’

• OpenAI launched Frontier, an enterprise platform that lets companies build, deploy, and manage AI agents that can run other software like Salesforce and Workday.

• The platform connects databases, business tools, and internal applications, letting AI agents and human employees work together with the same data, tools, and security controls.

• Fortune 500 companies including Intuit, State Farm, Thermo Fisher, and Uber have signed up as initial customers, while investors worry Frontier could threaten traditional SaaS providers.

Meta’s internal memo signals AI comeback after rocky year

• An internal memo reveals that Meta has completed pretraining on its new AI model codenamed “Avocado,” marking a potential recovery after the company’s troubled 2025 with its Llama 4 release.

• Avocado beats the best freely available base models and competes with leading fully trained models in knowledge, visual perception, and multilingual performance, despite not yet going through post-training.

• The new model is ten times more efficient than Maverick and a hundred times more efficient than Behemoth, with Meta planning capital spending of $115 to $135 billion for 2026.

Alexa+ could integrate ChatGPT

• Amazon is reportedly in talks to integrate ChatGPT into Alexa+, which could mean the voice assistant would be powered by both Anthropic’s Claude and OpenAI’s technology after a potential $50 billion investment.

• The timing is notable because Amazon just opened Alexa+ to all US users after nearly a year of limited testing, and Apple is about to launch an upgraded Siri powered by Google’s Gemini.

• The deal could include other elements beyond Alexa+, such as access to OpenAI’s Codex, getting OpenAI to use Amazon’s Trainium chips, or opening Amazon’s store to OpenAI’s agents.

Anthropic launches Claude Opus 4.6

• Anthropic released Claude Opus 4.6, its newest model, featuring a 1M token context window in beta — a first for the company’s Opus-class models — along with stronger coding and agentic task performance.

• Opus 4.6 scores highest on benchmarks like Terminal-Bench 2.0, Humanity’s Last Exam, and BrowseComp, and beats OpenAI’s GPT-5.2 by around 144 Elo points on the GDPval-AA knowledge work evaluation.

• New developer tools include adaptive thinking, four effort levels, context compaction for longer-running agents, and agent teams in Claude Code that let multiple agents work in parallel on tasks.

Amazon plans $200B in AI spending

• Amazon plans to spend $200 billion in capital expenditures in 2026, up from the $125 billion it had estimated for 2025 and well ahead of analyst expectations.

• CEO Andy Jassy tied the $200 billion AI spending plan to strong demand for existing offerings and opportunities in AI, custom chips, robotics, and low earth orbit satellites.

• Amazon’s stock sunk more than 10% in after-hours trading after the $200 billion spending plan was revealed alongside missed Wall Street profit expectations in its Q4 earnings.

Apple scales back AI health coach plans

• Apple has scaled back Project Mulberry, its planned AI-powered health coach for the Health app, after leadership changes in the company’s health and AI organizations prompted a reassessment of the initiative.

• Services chief Eddy Cue, who now oversees the health team, told colleagues that Apple’s plan wasn’t compelling enough compared to rivals like Oura Health and Whoop and their iPhone apps.

• Some Project Mulberry features may still ship individually over time, including one that would use the iPhone’s camera to analyze how a person walks, plus Siri health queries in iOS 27.

OpenAI launches GPT-5.3-Codex its self-building coding model

• OpenAI has released GPT-5.3-Codex, a new coding model designed for developers that the company says set record scores on SWE-bench Pro and Terminal-Bench while running 25% faster than its predecessor.

• The model showed a big jump on Terminal-Bench 2.0, scoring 77.3% compared to 64.0% for GPT-5.2-Codex, and on OSWorld-Verified it reached 64.7% versus the prior model’s 38.2%.

• Developers can now steer and interact with GPT-5.3-Codex mid-task without losing context, and the model provides frequent progress updates so users can ask questions and guide it in real time.

Meta tests a standalone app for its AI-generated ‘Vibes’ videos

• Meta confirmed it is testing a standalone app for Vibes, its AI-generated short-form video feature that lets people create, share, and browse AI videos in a dedicated feed.

• Vibes previously lived inside the Meta AI app, but spinning it out positions the product as a more direct competitor to Sora, OpenAI’s AI video and social app.

• Meta also plans to explore freemium access and paid subscriptions for Vibes video creation, letting users unlock additional monthly video creation opportunities in the coming months.

What Else Happened in AI on February 06 2026?

Perplexity launched Model Council, a new feature that runs queries through multiple AI models at the same time and synthesizes outputs into a single answer.

Roblox introduced 4D generation via its Cube AI foundation model, letting creators generate fully functional, interactive objects from text prompts.

Lotus Health raised $35M in Series A funding for its free AI-powered primary care platform, providing diagnosis, prescriptions, and referrals across 50 states.

Meta is rolling out a standalone app for its Vibes AI video platform, which was previously only available via the Meta app.

AI evaluation firm METR released new analysis for GPT-5.2 (high), finding it can now handle tasks that would take a human engineer over 6 hours to complete.

OpenAI introduces Frontier, a new platform that helps enterprises build, deploy, and manage AI agents that can do real work.

‘In the end, you feel blank’: India’s female workers watching hours of abusive content to train AI.

Rentahuman.ai Turns Humans Into On-Demand Labor For AI Agents.

OpenAI says new coding model helped build itself.