The code you can't explain
AI agent loops have made code fast and cheap to generate. The hard part, knowing what to build and staying in ownership of it, is still entirely yours.
Every post since 2009, newest first. Recent writing is on AI costs, AI agents, coding agents and AI strategy for founders, CTOs and leadership teams.
AI agent loops have made code fast and cheap to generate. The hard part, knowing what to build and staying in ownership of it, is still entirely yours.
Mitchell Hashimoto pledged $700k of his own money to a programming language he partly disagrees with. It's a rare honest answer to who actually pays for the tools the software industry depends on.
Anthropic now requires a government ID before accessing its most capable models. What looks like a safety measure is export control, and it is nudging developers toward the Chinese AI alternatives the US is trying to contain.
Norway just banned AI in elementary schools. The research behind that decision is more uncomfortable than the headlines suggest.
The companies building giant AI compute for themselves have started renting it out: SpaceX leased all of Colossus to Anthropic, and Meta says a cloud business is on the table. Apple is about to have the most efficient, most private inference fleet on earth. Can it keep that to itself?
Google invented the transformer, paid $2.7B to buy back its co-author, and just lost him to OpenAI. The pattern says something about why large organizations keep giving away the people who built them.
SpaceX just paid $60 billion for a coding tool. The price is interesting. The justification -- a $26 trillion addressable market -- reveals more about how we price AI right now than about what it's actually worth.
Serious developers are dropping paid AI subscriptions for local models and reporting it actually works. The 'it's free' pitch is real - but it obscures a different set of costs.
Rio's 397B-parameter 'homegrown' model turned out to be an unattributed merge of two others. What any public body should disclose before it says 'we built it'.
A Derbyshire officer is under investigation for AI-generated evidence. The fix is chain-of-custody and disclosure rules for AI tools, not one disciplinary case.