1
Don't classify, hallucinate
A useful pattern for LLM classification at scale is to let it hallucinate plausible, fake entities. Then resolve to real ones.
Doug Turnbull
- Popular posts
- 11
- Total points
- 2,526
- HN comments
- 1,986
- Rank
- #608 personal
Visit softwaredoug.com → RSS feedOn Hacker News Nov 2021 – Aug 2026
Software Engineering 7Career & Management 4AI & Machine Learning 3Life, Learning & Writing 2Databases & Data 2Tech Industry & Business 2
2
Are we at peak vector database?
Seriously - why do we need all these vector databases? Do we need dozens of them?
3
Coping with dumb LLMs using classic ML
Taking the output of many dumb LLM search relevance judges and feeding the output to a decision tree to improve precision