"Cached web" for training is old news — pretraining has done it all along
adonis_singh · x · 2026-10-24
Reacting to a claim that a lab built a cached copy of the web just to train its models, adonissingh quips that this is nothing new — pretraining has effectively always been that.
More from Fun
- 'Quota-exhaustion syndrome': the shared affliction of heavy AI users — ezshine · 2026-10-24
- Grad students should work resident shifts to keep up with AI slop, jokes researcher — SoloGen · 2026-10-24
- Claude tasked with building an office for Tavus PALs recreates The Office set — anthara_ai · 2026-10-24
- Old users can still call it Twitter: insider quip on platform naming and belonging — MatthewChang · 2026-10-24
- Claude built a Tavus PAL office — and it looked like a video game — anthara_ai · 2026-10-24
- Building with AI, drowning in research: when does optimising the process become procrastination? — SJW_Shadow_Monarch · 2026-10-24