Good morning. Biohackers tried to connect an AI agent to a lobster. The idea came out of Biopunk House in San Francisco, where two tech founders saw OpenClaw’s lobster logo and treated it like an invite. The goal was to put a remote control cockroach kit inside the lobster so OpenClaw could steer a living creature. But wait, aren’t lobster and cockroach brains different? Wouldn’t this require actual neurosurgical skill? Don’t worry, if vibe coding has taught us anything, it’s that biology just needs API access.
Zuck’s dream team still can’t make AI agents work. Meta is projected to spend as much as $145 billion this year on AI infrastructure while poaching top talent from rival labs for its agent push. At an internal town hall, Zuckerberg told staff agent development has moved slower than expected, and the benefits from Meta’s new AI structure still have not shown up. After the comments leaked, Alexandr Wang tried to cover for Zuckerberg, saying he meant the whole industry’s progress on AI agents, not just Meta’s. Incredible damage control. If everyone is struggling with agents, that only makes the story bigger. If elite talent + unlimited money can’t figure out agents, maybe agents are the problem. (Reuters, Barrons)
Palantir’s CEO says frontier AI is stealing your alpha. Alex Karp went on CNBC to talk about Palantir’s Nvidia deal, then torched the token economy instead. Karp says CEOs are furious because they’re paying huge AI token bills while giving OpenAI and Anthropic a close look at the secrets that make their companies valuable. He calls the AI industry “effing insane” and frames it as a national security fight. Why should businesses pay Silicon Valley labs to rent back their own intelligence? His answer: open models. And that’s where the story gets dangerous. Palantir’s CEO set out to defend American business, yet he handed China’s AI labs the best sales pitch they never had to write. (The Street, Silicon Angle)
Wispr Flow works inside every Al tool you use. Speak detailed prompts 4x faster than typing.
- Flow handles camelCase, snake_case, and acronyms
- Better input, better output
- Used by millions of developers and Al professionals
👀 closer look
Nvidia is financing its own demand. It launched a new revenue sharing program to help AI startups get compute, giving them access to Nvidia infrastructure without forcing them to bring all the cash upfront. Agents are proving harder to build than the demos suggested, and real products burn compute every time they run. Nvidia gets a cut of cloud revenue whenever that compute gets used, turning scarcity into a very useful loop. Critics call it circular financing because Nvidia is helping fund the customers who help prove demand for Nvidia is real. The agent boom is so inevitable that Nvidia has to help finance it. (Quartz)
AI startups are doing more with fewer people. A Harvard Business School study looked at nearly 50,000 venture backed startups and found AI native firms run with 25% fewer employees than traditional startups. These companies are built around small, senior technical teams that rely heavily on AI and automation. The surprising part is they still raise about the same amount of money as much larger companies. That makes valuation per employee a serious signal for investors watching the next wave of startups. The old startup flex was hiring a big team. The new one is getting more done with less. (Forbes)
fun stats
🔥 $200. Tesla’s new weekly AI spending cap per employee after engineers burned through thousands of dollars in tokens each week.
👔 3%. Companies fully back in the office 5 days a week. Nearly 1 in 4 Americans still worked from home at least part time in 2025. So much for the grand return.
⚽ $1 billion+. Fed investment in AI surveillance and drones to secure the 2026 World Cup. No word on how long the biometric data will be stored or how it will be used later.