Next note
A mission is more than a prompt
Why durable objectives, not longer instructions, are the useful unit for complex work.
DeepTrain / Journal
Thinking about the work that still needs a team.
Notes on long-horizon agents, complex work, and the systems that make hard answers hold up.
Featured / Essay
Too Smart to Prompt
Frontier language models now spend days solving century-old math problems. They also find groundbreaking vulnerabilities in open source codebases that humans have spent decades perfecting. So why aren’t we seeing AI helping companies transform themselves, radically outperform competitors, or do anything new?
Read the essayFrom the field
More from the journal.
Field note
The work that doesn't fit in a chat window
The next useful unit of AI work may not be a response. It may be a mission: a bounded objective with data, tools, a way to test the result, and enough time to find the move.
Coming soon
Verification is part of the product
A result is only as useful as the checks that let a team trust it.
Coming soon
When the answer is a schedule
The difference between explaining a system and finding the move inside it.