News for hackers. Stuff that matters.

A comment on Gemini 4 Argon. Parent comment

Re:Gemini 4 Argon by FranciscoCarlos on Wednesday October 14, 2026 @09:36AM (#50087586)
Yeah, it can be worked around with in-context instructions but they still tend to scope, delegate and architect without awareness of the fact that agents exist and work. It's an actual big deal IMHO, opus 5.5 is the most updated AFAIK.
[ Parent | on HN ]
Re:Gemini 4 Argon by piperswe on Wednesday October 14, 2026 @11:27AM (#50089140)
Nowadays the scraped training data is only one part of the training process. The RL phase of training can instill a knowledge of tactics like sub-agent delegation without documentation about them being present in the training corpus. Context isn't the only way for a model with an early knowledge cutoff to know about it.
[ Parent | on HN ]
Re:Gemini 4 Argon by FranciscoCarlos on Thursday October 15, 2026 @08:45AM (#50102849)
Still a crutch around incomplete pre-training data, a ton of useful knowledge wont be properly represented on the weights.
[ Parent | on HN ]