Working with AI
Verify and trust
What AI gets wrong, how to check its output, and what is safe to share at work.
0/2 complete
01
How to verify AI output before you trust it
AI is fluent enough to make wrong answers sound right. A practical checklist for catching hallucinations, broken facts, and silent drift — for code, copy, research, and analysis.
7 min read
02
Evaluating an indie AI tool before you commit
The 15-minute pre-commit checklist for any AI tool — wrapper-risk, team signal, changelog test, community signal, lock-in audit. Plus the LTD-specific risks (acquisition, feature gating, support cliff) and when indie genuinely beats incumbent.
9 min read
When you are done here, go to
How AI works
Now go one level deeper — the mechanism behind tokens, embeddings, attention, agents. For when "it just works" stops being enough.