The view from above the noise
Three things, three minutes, every two weeks: what changed in AI and what it means for you, one problem we solved for a client, and a look at how we work.
Sign up for our email newsletter
One issue every two weeks. Unsubscribe any time. See our Privacy Policy.
Past editions
The AI labs are using AI to build the next AI.
Anthropic says AI is speeding up its own research, a company that cut a daily task from 11 minutes to 2 without replacing its systems, and an app we're building so our people can rehearse hard conversations.
Your slowest step isn’t the work. It’s the wait before it.
The factory-AI startup that raised $232 million to cut the time it takes to get ready for new work, an 86-page plan set an estimator reviews before anything gets priced, and a question our engineers ask first.
Washington's AI honeymoon may be over
The FTC opens an investigation into OpenAI, Anthropic, and METR, a freight quote the AI drafts and a person approves, and why we never start with a platform.
A new AI model works for pennies, if you're asking it the right kind of question
What changed in AI and what it means for you, one client problem solved, and why every AI idea has to clear the same bar before it gets built.
Nearly a third of companies just said no to buying software
The model that can drive your software also got much harder to trick, why our verification agent is built to never make the final call, and the internal habit that turns AI's own answers into rules.
Running AI got cheaper again, unless you picked the wrong supplier
What changed in AI and what it means for you, one client problem solved, and why we run every build past a second AI reviewer.