Uncensored AI for lab notebook summaries
The month of bench work, turned into the paragraph the meeting wants
Bench science generates notes nobody has time to read, including you. The monthly update, the thesis chapter, the handover note: all are the notebook, compressed, and the compression is the work you keep deferring.
Why the mainstream APIs fail here
Uploading notebooks to a hosted service trains on your unpublished results, which every supervisor and tech-transfer office has opinions about. Doing it by hand costs the evening, so the update gets written from memory, which is how experiments get remembered as working.
How Wild West API is used for this
Paste the raw notes; ask for what was done, what worked, what failed and what the next decision is, in that order. 1M context holds months of notes at once, so patterns across weeks surface.
A working loop
- Fails first: a summary that hides failures is a lie with formatting.
- Ask for the next decision, not just the activities; the meeting needs it.
- Keep one thread per project so the summary sees the history.
Details are on the privacy page. Per-model prices are on /models and the pricing model is on /pricing.