They've been talking about the recurrent layers... but my suspicion is that this "innovation" is to obfuscate "thinking" tokens/content instead of actually improving the model in any sort of useful way.
this post was submitted on 04 Sep 2026
47 points (91.2% liked)
TechTakes
2689 readers
81 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 3 years ago
MODERATORS
Someone pointed out that on some tests, Astra is actually a regression from previous models. Totally AGI guys
Benchmaxxing for one set of benchmarks can actually degrade performance on other benchmarks. They've plateaued for a while, all they can do is scale inference compute up and down (at logarithmically poor rates of exchange) and trade performance in one area for performance in others
It must have been too good before. They needed it to be more general. /s
Yeah sloppers I know went from “see it proved new math! that totally clears us of plagiarism accusations we get when we repost that someone ai generated a frogger and a WoW clone” to “tried to use it for coding and it didn’t do a good job”.
The math stuff is mostly using Lean proof verifier and all that, by the way, the only use of LLM that they found that is actually legit because it doesn’t matter if its slopping, and you want your attempts randomized.
It's so annoying, every time a new model comes out, the "thought leaders" on my team are always gushing about how they finally achieved AGI this time.
This has been happening since GPT3. Every time. Right on schedule. They just regurgitate the ad copy.
Of course none of them have any idea what the terms they're using mean
The finances must be really really bad.
Also didn't they define AGI as 'it should be able to make x00 billion dollars'?
if it was that they would not be selling it!!