Key Takeaways
- California's proposed billionaire wealth tax is drawing rare cross-criticisms from Naval, Paul Graham, Garry Tan, and Alexis Ohanian, who warn it will drive founders out of the state.
- AI-assisted engineering showed a concrete win: Fable's full Rust rewrite cut startup time from 87ms to 2ms and improved rendering by 9.6x.
- Amjad Masad argues that Anthropic's regulatory proposals disadvantage frontier AI labs, and calls for objective pre-deployment testing plus a FINRA-like regulator.
- Basecamp is emerging as an AI-agent hub, with 'Agent Clawdito' automating customer-report generation and task creation.
- Naval and Guillermo Rauch both spot a 'next Disney' opportunity in AI-era content, where values are the core moat.
1. California Wealth Tax Sparks Founder Backlash
California's wealth tax proposal targeting billionaires continues to stir backlash. Naval calls the policy impractical and says it will force founders to leave the state; Paul Graham sarcastically notes that progressives may follow Trump's playbook and use the rich as scapegoats; Garry Tan criticizes the Silicon Valley representative who supports it, saying he doesn't understand the harm, and calls for a strong challenger in the primary in two years. — via 1 2 3
Alexis Ohanian points to a specific flaw in the plan: founders would be forced to sell shares to pay a one-time 5% tax, with the state selling them if they default. He argues this ignores founders who pour all their time and profits into their companies and often lack liquidity, unlike public executives such as Jensen and Lisa Su. He calls it 'the biggest insult in startup history.' — via 1
Garry Tan also mocks a government loan program, saying YC founders who have raised Series A would not borrow hundreds of millions from the government for ten years; they would instead move out of California. He says this is good for the Cambridge, Massachusetts tech ecosystem and that he could open a YC office there right away. — via 1
2. AI Agents Hit New Performance Milestones
Fable used AI to complete a full Rust rewrite in one pass, slashing startup time from 87ms to 2ms and boosting rendering performance by 9.6x. The case demonstrates how AI-driven large-scale code refactoring can dramatically improve product performance. — via 1
Jason Fried shared a Basecamp AI agent workflow: Agent Clawdito automatically writes customer sentiment weekly and monthly reports, publishes them to a Basecamp project, creates its own task cards, to-dos, and chat summaries, and even writes runbooks and report guides. He also endorses other teams using Basecamp as the primary communication tool for AI agents, saying it's 'spot on.' — via 1 2 3
Patrick Collison argues that agentic coding harnesses should not use the terminal as their main interface, since terminals have low information density and weak UI. He hopes for a breakthrough similar to what Jupyter did for REPLs, pointing to a clear opportunity in tooling around AI agents. — via 1
3. AI Regulation: Open Weights vs. Frontier Controls
Amjad Masad calls the framing 'either concentrate AI in a few companies and politicians via regulation, or distribute it widely' a false choice. Objective, fair institutional processes have a decentralizing force, similar to how formal courts are better than vigilante justice. — via 1
He details how Anthropic's policy proposals intentionally put frontier AI companies at a disadvantage while helping smaller competitors: supporting California's SB53, stricter testing for frontier models compared to non-frontier, and the 'Pacing the Frontier' approach that regulates the top models without restricting followers. These measures would harm frontier labs and help challengers, including open-weight players. — via 1
Masad also argues that AI structurally tends to concentrate power due to scaling laws, independent of regulation. Open-weight models help but are insufficient, because concentration would shift to entities with the most compute and chips. He supports pre-deployment testing for frontier models and similar testing when open-weight models approach frontier capability, along with a FINRA-like regulatory body. — via 1
4. The Hunt for the Next Disney in an AI Era
- Naval and Guillermo Rauch both highlight the opportunity to find and fund 'the next Disney.' Rauch argues entertainment companies should return to core American values (merit, family, overcoming adversity, optimism, and the pursuit of happiness), and that AI enables everyone to create; Mickey Mouse has survived technology shifts because of the values it represents. Naval adds that in the AI era, values and a distinct brand in content creation become even more important. — via 1 2
