In partnership with
In Today’s Issue:
🏗️ A new enterprise "operating system" debuted alongside GPT-5.3
🧬 GPT-5 achieved a 40% reduction in protein production costs
📈 GPT-5.3-Codex marks a milestone as the first frontier model
🏛️ Researcher Dean W. Ball warns that frontier labs are moving toward "agentic workforces" of hundreds of thousands
✨ And more AI goodness…
Dear Readers,
Twenty minutes separated two rival AI giants on February 5th, and in that sliver of time, the entire industry shifted gears. Anthropic launched Claude Opus 4.6 with a million-token memory and collaborative agent teams; OpenAI countered with GPT-5.3-Codex, faster, sharper, and the first model to debug its own development.
While these two titans trade blows with Super Bowl ads and head-to-head releases, the real story is unfolding in your inbox today: AI agents slashing protein synthesis costs by 40% in autonomous cloud labs, OpenAI's new Frontier platform turning models into actual coworkers driving 5% revenue growth, and a brutal January jobs report—108,435 layoffs, the worst since 2009, where AI gets both the credit for productivity gains and the blame for disappearing roles.
The gap between what AI can do and what we're ready for has never felt narrower or more unsettling, so let's dig into what happened, what it means, and where this acceleration actually takes us.
All the best,
🤖 On Recursive Self-Improvement (Part I) by Dean W. Ball
Frontier AI labs are rapidly automating research and engineering, potentially expanding their effective workforces from thousands to hundreds of thousands of AI agents within just a couple of years. This shift could dramatically accelerate AI progress in 2026, either as a faster version of today’s scaling trends or as a true paradigm leap driven by algorithmic efficiency gains, massive new compute, and AI-assisted experimentation. The big takeaway for policymakers: this isn’t sci-fi anymore, but it demands calm, technocratic oversight, not panic, as the dynamics of AI competition and governance quietly change behind closed doors.
I highly recommend you read his blog.
🧬 AI Slashes Protein Production Costs
GPT-5, paired with an autonomous cloud lab run by Ginkgo Bioworks, cut cell-free protein synthesis costs by 40% after just six closed-loop experimental rounds spanning 36,000+ reactions. By directly designing, running, and learning from robotic lab experiments, the system found novel, low-cost reaction recipes that outperform prior state-of-the-art methods, showing how AI-driven iteration can unlock faster, cheaper biology. The bigger signal: autonomous labs remove the iteration bottleneck, potentially transforming how quickly biological breakthroughs move from idea to impact.
🚀 OpenAI Launches Frontier Platform
OpenAI has unveiled Frontier, a new enterprise platform designed to turn AI agents into real “AI coworkers” that can plan, act, learn, and operate securely across business systems. Early adopters report massive gains, cutting workflows from weeks to days, freeing up 90% more sales time, and driving up to 5% revenue growth, by giving agents shared context, permissions, and feedback loops. Frontier aims to close the growing gap between what AI models can do and what companies can actually deploy, helping enterprises scale AI safely, faster, and with real ROI.
Sam Altman Responds to Anthropic’s Attack Ads, Live on TBPN
OpenAI frontier agents
The Takeaway
👉 Anthropic and OpenAI released their most powerful models within 20 minutes of each other on February 5th — the closest head-to-head launch in AI history, signaling an unprecedented escalation in competition.
👉 Opus 4.6 dominates in reasoning and professional knowledge work with a 1M-token context window, while GPT-5.3-Codex leads in raw coding speed and terminal tasks — neither model is a clear overall winner.
👉 GPT-5.3-Codex is the first major model that actively contributed to its own development, marking a symbolic milestone toward recursive AI self-improvement.
👉 Enterprise AI spending is surging to $11.6M per company in 2026, and both companies are racing to become the default AI colleague — not just a chatbot, but an autonomous worker embedded in daily workflows.
Twenty minutes. That's all it took to turn February 5th into one of the most dramatic evenings in AI history. At 6:40 PM (Berlin time, since I’m German), Anthropic dropped Claude Opus 4.6, its most powerful model ever. By 7:00 PM, OpenAI fired back with GPT-5.3-Codex. Two titans, two flagship models, one unmistakable signal: the battle for AI supremacy has reached a whole new level.
Both models claim the crown in different arenas. Opus 4.6 brings a massive 1-million-token context window; it’s like feeding it three entire novels and it remembers almost every word. It crushed GPT-5.2 on real-world professional tasks by 144 Elo points and introduced agent teams, where multiple AI instances collaborate like a dev squad. GPT-5.3-Codex countered with raw speed, 25% faster than its predecessor, and dominated Terminal-Bench 2.0 at 77.3% versus Opus 4.6's 65.4%. The wildest part: GPT-5.3-Codex literally helped build itself, with early versions debugging their own training.
Neither company blinked. Neither backed down. And with competing Super Bowl ads lined up for Sunday, this rivalry is just warming up. The real question: when AI models start improving themselves, where does this race actually end?
Why it matters: This simultaneous launch marks a turning point, AI is graduating from a tool you consult to a colleague you delegate to. Enterprise AI spending is projected to hit $11.6 million per company in 2026, and the winner of this race will shape how millions of professionals work every day.
Sources:
🔗 https://www.anthropic.com/news/claude-opus-4-6
🔗 https://openai.com/index/introducing-gpt-5-3-codex/
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
“The past year has seen an explosion in coding productivity” via Financial Times
Worst January Since the Financial Crisis
January 2026 just delivered the ugliest jobs number in 17 years, and AI is sitting right at the center of it. U.S. employers announced 108,435 layoffs last month, up 118% from a year ago and a staggering 205% from December. It's the worst January since 2009, when the economy was still reeling from the financial crisis. Hiring? Just 5,306 new positions announced, the lowest January figure ever recorded.
The timing is hard to ignore. While Anthropic and OpenAI race to build AI agents that can do entire jobs autonomously, companies like Amazon, Pinterest, and Dow are openly citing AI as the reason for cutting thousands of workers. In 2025 alone, over 55,000 layoffs were directly attributed to artificial intelligence, twelve times more than two years earlier. Meanwhile, a $285 billion stock market rout hammered software companies this week after investors realized AI isn't just disrupting chatbots, it's coming for entire industries. Salesforce down 26% year-to-date. Intuit down 34%. The market is voting with its feet.
However, researchers at Oxford Economics and Forrester say many of these "AI layoffs" aren't really about AI at all. Companies are using the technology as a convenient excuse to trim payrolls bloated by pandemic-era over-hiring. Forrester predicts that half of AI-attributed layoffs will be quietly rehired, just offshore and at lower salaries. The real disruption isn't the robots replacing us. It's the uncertainty about what comes next.
The gap between AI's promise and its actual deployment is creating a dangerous limbo for millions of workers; jobs are disappearing today based on technology that isn't fully ready yet. How companies and policymakers navigate this transition will define whether AI becomes an engine of shared prosperity or a widening inequality machine.