AI Mornings with Andreas Vig
Your daily AI news briefing in under 10 minutes. New models, product launches, research breakthroughs, and industry shifts, explained clearly, no hype.
AI Mornings with Andreas Vig
Anthropic's $400M Biotech Bet & OpenAI's Executive Shuffle
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Hey, welcome to AI Mornings with Andreas Vig. It's the 6th of April 2026. Anthropic just made its first major acquisition outside the AI model space. The company has purchased a stealth biotech startup called Coefficient Bio in a$400 million stock deal. The company was only eight months old, founded by two former Genentech researchers who were using AI to accelerate drug discovery. The entire team, about 10 people, is joining Anthropic's Health and Life Sciences Division. This is a strategic move beyond just building better models. Anthropic launched Claude for Life Sciences back in October, and this acquisition signals they're serious about building domain-specific products for scientific research. Sticking with Anthropic, there's news that affects anyone using Claude Code with third-party tools. Starting April 4th, Claude Code subscribers can no longer use their subscription limits with OpenClaw and similar third-party harnesses. Instead, they'll need to pay separately through a pay as you go option. The timing here is interesting. OpenClaw's creator, Peter Steinberger, recently announced he's joining OpenAI, with OpenClaw continuing as an open source project. Steinberger publicly criticized Anthropic, saying they copied popular features into their closed harness and then locked out open source. Anthropic's head of Clawed Code, Boris Cherney, framed it as an engineering constraint. Subscriptions weren't designed for how these third-party tools consume API calls. Over at OpenAI, there's been a significant executive shuffle. COO Brad Lightcap is transitioning to a new role leading special projects, which involves complex deals and investments. He'll report directly to Sam Altman. Fidji Simo, who runs AGI development, is taking medical leave for several weeks due to a neuroimmune condition. Greg Brockman will handle product responsibilities while she's away. The company's marketing head, Kate Rauch, is stepping down to focus on cancer recovery, though she plans to return to a narrower role later. Denise Dresser, the former Slack CEO who recently joined OpenAI as chief revenue officer, is taking over some of Lightcap's commercial duties. OpenAI says it's still focused on advancing research, growing to nearly a billion users and winning enterprise customers. Microsoft is dealing with an awkward optics problem around copilot. The terms of use for copilot, last updated in October, state that the AI assistant is for entertainment purposes only, and that users should not rely on it for important advice. The company is actively pitching copilot to enterprises, so this language has drawn attention. A Microsoft spokesperson told PC Mag this is legacy language that will be updated soon. To be fair, similar disclaimers exist at OpenAI and XAI. OpenAI tells users not to treat outputs as factual truth. But it's a reminder that even as AI companies market these tools for serious work, their legal teams are still hedging. Japan is emerging as a leader in what's being called physical AI robots and autonomous systems operating in the real world. The driver here isn't efficiency, it's survival. Japan's working age population has declined for 14 straight years and is projected to shrink by another 15 million people over the next two decades. The government under Prime Minister Sanae Takaichi has committed$6.3 billion to build domestic physical AI capabilities with a goal of capturing 30% of the global market by 2040. Japanese companies already hold 70% of the industrial robotics market globally. Softbank is deploying robots with vision language models in factories and warehouses. The signal here is real deployment companies are buying robots to run full shifts, not just running trials. SpaceX has reportedly filed confidential paperwork for an IPO that would raise$75 billion at a$1.75 trillion valuation, part of Elon Musk's pitch, orbital data centers. He's not alone. Jeff Bezos's Blue Origin and the recently funded StarCloud startup are all pursuing similar concepts. The idea is that space could avoid the permitting battles and community opposition that ground-based data centers face across the US. But analysts note that orbital compute would remain a tiny fraction of what's built on Earth for the foreseeable future. This is more about expanding SpaceX's Starlink satellite business and creating new revenue streams for launches. A fun one from the developer community. A new Claude Code skill called Caveman has gone viral on GitHub with over 1,800 stars in just a few days. The idea is simple: make the AI talk like a caveman to cut token usage by about 75% while keeping all the technical accuracy. Instead of saying the reason your React component is re-rendering is likely because you're creating a new object reference on each render cycle. It just says new object ref each render, inline object prop equals new ref equals re-render, wrap in use memo. What's interesting is there's actual research behind this. A March paper found that constraining models to brief responses improved accuracy by 26 percentage points on certain benchmarks. The tool offers three intensity levels from professional but concise to full telegraphic caveman. It's a clever hack that highlights how much filler language models typically produce. That's all for today. See you tomorrow.