Articles & threads
The weekly How I AI digest leads with Claire Vo spending a week inside Jev: nine cents to make sense of two years of ChatPRD pull requests, about four dollars for 200,000 classifications, and a working answer to which tasks still deserve a frontier model.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
Eight takeaways accompany the episode, several with numbers worth stealing: her AI usage fell from nearly all engineering in January to under 40% by September as agents took over, and pairing Jev's cheap sorting with GPT-6 Astra reasoning kept a 1,100-signal insights graph near four dollars. The post also links ChatPRD's written walkthroughs of each workflow.
Start by collecting past project updates or work logs from your team. A developer can write code to pair these notes together. Your app then asks Jev if each pair shares the same focus, grouping related jobs into piles without needing a preset list of folders.
Once Jev finishes sorting, have a regular text bot read each pile and give it a helpful name. Jev only picks choices or scores instead of writing sentences, so generating human labels still requires a standard language tool.