Sakana's self-improving loop is an efficiency result, not a capability leap
Published
Augur writing
Analysis worth returning to
Longer-form Augur analysis, published here first and kept under one stable address as the argument changes.
Latest pieces
Newest publication first.
OpenAI's safety firings are a vendor-risk signal, whatever the cause
Published
Anthropic swaps a bright-line election ban for judgment calls it must enforce
Published
AI lifts the work, not the worker: Google's attorney trial splits its juniors
Published
Anthropic's cyber access tiers turn safety gating into vendor gatekeeping
Published
Keep the reviewer: Greptile's AI code data is a sample, not a license
Published
Databricks's catalog exit is a preview, so test before you unbundle
Published
Anthropic's 10,000-engineer academy is a cheap bet on the integration layer
Published
Barclays expects half its developers on Claude Code; watch the bill first
Published
OpenAI pulled GPT-6.1 Astra, and its own testing was the only real brake
Published
OpenAI writes safety cases while its agents keep beating the monitors
Published
Anthropic's Sonnet 5.5 saving is in the token count, so test it before rerouting
Published
Cursor's 15x spread means the next AI coding dollar goes to review, not seats
Published
Stanford's benchmark study means AI buyers must run their own tests
Published
Black Forest Labs' open robot model: buy the cost curve, not the benchmark win
Published
Anthropic's Claude enzyme find proves the pipeline more than the science
Published
Anthropic's Opus 5.5 is cheaper and stricter; price the refusals first
Published
TypeSafe's Jev set adoption records; test it on your own labels first
Published
Descript's router switch: buy the testing loop, the model choice follows
Published
Anthropic's billion-dollar outside evaluator is also its sales partner
Published
Anthropic moves the safety gate from the model to your credentials
Published
Apple's agent memory paper says forget more: full transcripts scored worst
Published
Neither 26% nor 86% is yours: Trail of Bits reopens 1Password's AI patch test
Published
Fireworks puts DeepSeek V4.1-Flash level with Astra on coding; test it yourself
Published
Keep exploring
Move from the argument to the record, data, or underlying evidence.