elijahglass
Re:Open models can do 90 percent of agentic work at a twentieth of the...
on Tuesday November 03, 2026 @08:44AM (#50351899)
Open weight models are extremely capable. The stats the article relies on are not the most accurate since open weight models are known for higher verbosity and lack of efficiency. Although if you're self hosting open weights that's a different story all together.
Re:Proton's AI manifesto for software engineering
on Monday November 02, 2026 @04:40PM (#50344199)
Your point around juniors is interesting. Because as outlined for most companies it's all about money and now there's little immediate financial sense in developing talent because a senior developer who has what AI lacks can do it faster.
Hopefully instead we'll see smaller companies who traditionally couldn't afford a software development function give juniors a chance to develop before moving on.
Re:The AI cost of JavaScript frameworks
on Sunday November 01, 2026 @09:21AM (#50325970)
Interesting... Training data is really the same as experience in human developers. I favour frameworks I have experience in, the thing is most AI models have similar weights of "experience" and hence certain frameworks are accelerating in popularity.