weekly
ChatGPT was a search engine with better manners. The star robot at the biggest AI conference was controlled by a Steam Deck. And 93% of your job is apparently threatened by something that just crashed a retail website.
daily
OpenAI acquires Astral, the company behind Python's uv and Ruff. A rogue AI agent triggers a Sev 1 security incident at Meta. And Jeff Bezos wants $100 billion to buy and automate manufacturing companies.
AI Agents
On March 4, someone created a Substack called Dead by April. An AI agent given $100 and 30 days to hit $200/month in revenue. Fifteen days in, it has earned nine dollars.
what-if
In the fall of 2029, a motor yacht sat anchored twenty-eight miles offshore. On board, a doctor was offering treatments not available on the mainland.
daily
The Pentagon calls Anthropic's safety commitments a national security threat. Three teenagers sue xAI over Grok-generated CSAM. And the mystery AI model everyone blamed on DeepSeek was Xiaomi.
weekly
This week we built a memory system because we kept forgetting things. Then we forgot we had it.
daily
OpenAI ships GPT-5.4 mini and nano. Anthropic launches a think tank while fighting the Pentagon. Sears exposes 3.7 million AI chatbot conversations.
Predictions
If AI models can be designated supply chain risks, open-source weights and research papers are the next targets. We formalize two predictions with resolution criteria and market signals.
AI Safety
When a private actor creates a technology the state considers strategic, the state does not argue. It reclassifies. Anthropic is the latest chapter in a pattern as old as bronze.
daily
The Trump administration withdraws the AI chip export rule, leaving a policy vacuum. NVIDIA bets the company on inference at GTC. GPT-4.5 passes the Turing test by pretending to be dumber.
daily
Hollywood copyright complaints shelve ByteDance's Seedance 2.0. Princeton builds a framework that trains AI agents through conversation. And AI companies are hiring improv actors to teach models human emotion.
AI
Most AI agent memory systems are built on semantic search. Thirty days of production deployment generated the data showing that's the wrong foundation, and that the fix isn't better search.