It starts with the questions I expect to get, marked as starter FAQs. After launch, new entries are questions that the record could not answer. Every answer becomes part of the record, so the ask box can find it next time.
answeredAIStarter FAQPublished 2026-10-15
Why run models locally instead of the best cloud model?
I use both. Frontier models like Claude do the hardest work; local models handle private work at no cost per question. The price of local is checking their work harder, and that checking turned out to be the interesting part.
answeredWorkStarter FAQPublished 2026-10-15
Do you take speaking engagements?
Yes. Email me at ted@tedsmoot.com with the event, the audience and the date.
answeredAIStarter FAQPublished 2026-10-15
Which local models do you actually trust?
For auditing, one so far: Qwen3.6-35B-A3B passed my graduation test. The rest still run, but a script checks their work before I believe it.