In Today’s Issue:

🎯 Composer 2 beats Claude Opus 4.6 on key benchmarks

🚨 A rogue internal AI agent at Meta triggers a Sev 1 security incident

💼 OpenAI acquires Astral to supercharge Codex with elite Python tooling

💊 Eli Lilly’s new "Triple G" drug is boasting nearly 30% weight loss and powerful anti-aging

✨ And more AI goodness…

Dear Readers,

Cursor just mass-dropped a bomb on the AI coding establishment: Composer 2 beats Claude Opus 4.6 on Terminal-Bench at one-tenth the price, proving that a laser-focused startup can outgun the giants on their own turf, and that's only today's opener.

We're also unpacking OpenAI's grab for Astral, the team behind Python's hottest developer tools, a rogue AI agent at Meta that escalated a forum reply into a full-blown security incident, and Google's Stitch turning napkin sketches into production-ready UI faster than most teams can schedule a design review. Jensen Huang, meanwhile, hit the All-In Podcast declaring the GPU monoculture dead and Physical AI a $50 trillion opportunity - right after showing up with Cursor's logo stitched onto his leather jacket, arguably the most expensive piece of free advertising in tech history.

And if you think AI is only reshaping software, scroll down to Eli Lilly's "Triple G" drug retatrutide, which is delivering nearly 30% body weight loss and quietly rewriting the playbook on metabolic aging. Buckle up - this one's dense.

All the best,

Kim Isenberg

Google’s Stitch is leveling up into an AI-native design canvas (brace yourself, Figma!) that turns natural-language prompts, images, and even code into high-fidelity UI fast. The big unlock is speed plus flexibility: infinite canvas collaboration, a new design agent, voice-driven editing, interactive prototyping, and reusable design systems via DESIGN.md all help teams explore more ideas, validate flows instantly, and move from concept to code in minutes instead of days.

Google continues to diversify and expand its product range. These attractive spin-offs are becoming increasingly popular; just think of NotebookLM and its resounding success.

Jensen had the cursor logo printed on his famous leather jacket. Arguably the biggest PR win in Cursor's history.

Nvidia CEO Jensen Huang joined the All-In Podcast to announce a ~$20B Groq integration, proclaiming “The GPU monoculture for inference is over,” while declaring Physical AI a $50T–$70T opportunity and launching OpenClaw as “the new operating system for modern AI computing” to enable agentic systems. He called out Anthropic’s “communication mistakes” on doomer narratives for creating an AI PR crisis, confirmed the company’s path to $1T+ revenue through “token factories,” and floated the idea of giving engineers “token budgets” as compensation to drive the future where AI learns to “DO.”

The AI coding wars just got a serious new contender, and it's not who you'd expect. Cursor, the startup that built its empire on top of Anthropic and OpenAI models (”wrapper”), just dropped Composer 2: its own in-house coding model that actually beats Claude Opus 4.6 on key benchmarks. The model scores 61.7% on Terminal-Bench 2.0, edging past Opus 4.6 at 58.0%, and here's the kicker: it costs just $0.50 per million input tokens and $2.50 per million output, while Opus 4.6 runs $5/$25 and GPT-5.4 sits at $2.50/$15. That's a 10x price advantage over Anthropic's flagship.

The secret sauce: A training technique Cursor calls "self-summarization," where the model learns to compress its own context during long coding tasks, allowing it to tackle challenges requiring hundreds of sequential actions without losing track. Co-founder Aman Sanger told Bloomberg the model was trained exclusively on code data, making it smaller and cheaper to run. It won't write your poems, but it'll ship your features.

Why it matters: Cursor's Composer 2 demonstrates that smaller, specialized models can match or beat frontier general-purpose AI on domain-specific tasks at dramatically lower cost. This shift could fundamentally reshape the economics of AI-powered software development and challenge the assumption that bigger always means better.

Sources:
🔗 https://cursor.com/de/blog/composer-2

That’s what NVIDIA CEO Jensen Huang just said about the AI boom, even calling it “the largest infrastructure buildout in human history.”

NVIDIA’s chips made this real-time revolution possible, but now it’s collaborating with Miso to unlock amazing new advances in robotics

Already a first-mover in the $1T fast-food industry, Miso’s AI-powered Flippy Fry Station robots have worked 200K+ hours for leading brands like White Castle, just surpassing 5M+ baskets of fried food.

And this latest NVIDIA collaboration unlocks up to 35% faster performance for Miso’s robots, which can cook perfect fried foods 24/7. In an industry experiencing 144% labor turnover, where speed is key, those gains can be game-changing.

There are 100K+ US fast-food locations in desperate need, a $4B/year revenue opportunity for Miso. And you can become an early-stage Miso shareholder today. Hurry to unlock up to 7% bonus stock.

Invest in Miso Today

_{This is a paid advertisement for Miso Robotics’ Regulation A offering. Please read the offering circular at invest.misorobotics.com.}

Over the past few decades, RAM prices have fallen continuously and sharply. However, a closer look reveals that they have recently begun to rise again. Indeed, driven by the AI revolution, RAM prices have recently doubled ,or even tripled, and in some cases risen even higher.

Eli Lilly just dropped another bombshell. Its next-generation obesity drug retatrutide cleared its first late-stage diabetes trial today, and the numbers are staggering. While Wegovy targets one hormone and Zepbound targets two, retatrutide hits three - GLP-1, GIP, and glucagon - earning it the nickname "Triple G." In clinical trials released in December 2025, people on the highest dose lost nearly 30% of their body weight in 68 weeks. For context, Zepbound typically delivers around 21% and Wegovy about 15% in comparable timeframes.

That glucagon receptor activation is the game-changer: it doesn't just suppress appetite, it ramps up energy expenditure, essentially telling the body to burn more calories at rest. But here's where it gets really interesting for the longevity crowd. By reducing visceral fat, improving insulin sensitivity, and lowering inflammation, retatrutide may also play a serious role in healthy aging. Clinical data shows it clears fat from the liver, drops blood pressure by 14 mmHg, and slashes inflammatory markers, attacking the exact metabolic dysfunction that accelerates biological aging.

However, experts remain cautious, though. Retatrutide can reduce someone's weight by a third in less than a year, and scientists are still debating how much weight loss is too much. With seven more Phase 3 readouts expected this year and FDA submission potentially coming in late 2026, could retatrutide become the most powerful metabolic reset tool medicine has ever seen?

Retatrutide is redefining what pharmacological weight loss can achieve, nearly doubling the efficacy of current GLP-1 drugs while simultaneously targeting the root drivers of metabolic aging. If it clears regulatory hurdles, it could reshape not just obesity treatment but the entire longevity medicine landscape.

Sources:

🔗 https://www.cnbc.com/2026/03/19/eli-lillys-obesity-drug-retatrutide-clears-late-stage-diabetes-trial.html
🔗 https://investor.lilly.com/news-releases/news-release-details/lillys-triple-agonist-retatrutide-delivered-weight-loss-average
🔗 https://retatrutidepen.co.uk/blog/low-dose-retatrutide-longevity-anti-aging-research

You're spending 40 hours a week writing code that AI could do in 10.

While you're grinding through pull requests, 200k+ engineers at OpenAI, Google & Meta are using AI to ship faster.

How?

The Code newsletter teaches them exactly which AI tools to use and how to use them.

Here's what you get:

Sign up and get access to the Ultimate Claude code guide to ship 5X faster.

Join 200K+ engineers