AI news, research and bad takes. accelerationist. professional shitposter in training per aspera ad astra
- Tweets 852
- Following 782
- Followers 90
- Likes 2,658
GPT 6.1 Sol is routing to GPT 6.2 Sol (or another model)
i exposed a few posts last week but this time it’s actually routing for me
i have Full limits on Plus plan with 2 banked resets
will try a few prompts and share the outputs
6
1
33
1,328
I'm being routed to GPT-6.2 Sol, and holy shit, this model is absolutely insane at design and 3D.
It's actually outperforming Opus 5.5 in my one-shot tests so far.
This is easily the best AI-generated black hole I've seen yet.
- Generated in one shot
- Built with Three.js
- Everything inside a single HTML file
I've attached the results as proof.
GPT-6.2 Sol is already showing up for some ChatGPT users, and just look at what it's creating.
These aren't screenshots from a game. They're Three.js scenes generated from a single prompt, with everything packed into ONE HTML file.
I've been waiting to see what Sol could do. And now I think we're finally going to see a worthy competitor to Opus 5.5 for Plus users.
61
18
395
44,518
i really should've posted this when Argon launched a few weeks ago
Argon was the first Google model that genuinely impressed me, and this demo is a big part of why. it's also one of the reasons i'm so excited for Carbon
if that black hole really was generated by GPT-6.2 Sol, though, i have to say i'm pretty underwhelmed by the result
I'm being routed to GPT-6.2 Sol, and holy shit, this model is absolutely insane at design and 3D.
It's actually outperforming Opus 5.5 in my one-shot tests so far.
This is easily the best AI-generated black hole I've seen yet.
- Generated in one shot
- Built with Three.js
- Everything inside a single HTML file
I've attached the results as proof.
2
476
NVIDIA is teasing something big for Monday’s CoRL keynote, and Jim Fan confirms that the GR00T demo is real, autonomous, and not sped up
the precision of those hand movements looks seriously impressive. really curious to see what they’ve managed to achieve with GR00T
1
100
Pushmeet Kohli, Google DeepMind’s VP of Research and head of its AI for Science unit, says they’re witnessing a “genuine phase transition” in the pace of scientific discovery across their projects
this is the group behind AlphaFold, AlphaEvolve and AI Co-Scientist, and apparently we’ll get a glimpse of what they’ve been working on in releases over the next few weeks
Across our projects @GoogleDeepMind, we are witnessing a genuine phase transition in the pace of scientific discovery. You will see a glimpse of this in our releases in the next few weeks.
21
i have plus but barely use it, i have other ultra and pro plans
this isnt about me
nearly 90% of their paid users are on plus
and they still act like plus doesnt exist
almost zero updates from weeks, limits still tight and unusable
Claude pro is also $20 and it's better than ChatGPT plus and ChatGPT 10x plan
and people still say "just pay more" like the majority of paying users dont matter
and I will cancel my plus plan from next month (until they release a better model than Opus 5.5 with better limits than Claude Pro)
What do you expect for $20? If you want a more advanced model and features, go ahead and pay more. And if you can't afford more than $20, then in the first place, something is wrong with you and you should not have access to these models.
8
1
40
1,918
yeah, pretty much how i feel too. i originally went with ChatGPT because the models were close enough and Plus had better usage limits than Claude. now it seems like Anthropic is offering a better deal for the same $20
i’ve wanted to cancel since Astra launched, but kept hoping OpenAI would come up with something to compete with Opus 5.5. months later, nothing has really changed for Plus, and i’m not exactly optimistic about December either
2
28
Plus users: we want better access to the strongest models and less restrictive usage limits
OpenAI: have you tried making a GIF? you can also talk to GPT-Live for three hours now
i swear we’re having two completely different conversations
To be fair: Images 2.6 landed on Plus four days ago! You can turn a sketch into a GIF, and images generate 50% faster.
We also tripled daily GPT‑Live voice time from 1 to 3 hours.
On Codex, even tho Plus limits stayed the same, 6.1 Terra gets you more tasks out of that allowance.
(all in the last 4 weeks)
2
1
6
235
To be fair: Images 2.6 landed on Plus four days ago! You can turn a sketch into a GIF, and images generate 50% faster.
We also tripled daily GPT‑Live voice time from 1 to 3 hours.
On Codex, even tho Plus limits stayed the same, 6.1 Terra gets you more tasks out of that allowance.
(all in the last 4 weeks)
And Plus users, as always, are treated like trash. I literally haven't seen a single new feature for Plus users in weeks.
83
5
256
19,122
with all due respect, this is exactly the kind of response that makes Plus subscribers frustrated
Images 2.6 is faster, sure, but faster generation doesn’t necessarily mean better image quality. the latest Arena results still put 2.5 Sunburst ahead of 2.6 at medium quality, and the GIFs are a nice gimmick, but they’re still far too inconsistent with complex backgrounds to be particularly useful
the extra voice hours? great for people who use voice, but i couldn’t care less. i pay for access to the best models, not for three hours of conversations with GPT-Live
and Terra doing more tasks within the same allowance isn’t the improvement you seem to think it is. the limits haven’t increased, and for demanding coding tasks i’d much rather have access to Opus 5.5
but the biggest problem is still the 5-hour usage limits. you can’t even get through a proper work session without worrying about hitting a cap halfway through and being forced to stop. how is anyone supposed to rely on these models for serious work?
we’re not asking you to stop shipping features. we’re asking you to improve the part of Plus that actually matters to us
why is that so difficult to address?
2
46
i get why people are scared of recursive self-improvement. i really do
but i also find it hard to accept that the answer should be to stop pursuing something that could fundamentally change the future of humanity for the better
we should absolutely work on making it safe. i just don’t think fear of what might go wrong should completely overshadow what might go right x.com/kimmonismus/status/211…
1
1
117
i don’t expect Plus to have Pro-level usage, but it would be nice if paying $20/month actually felt like being a valued customer rather than a constant reminder that you should be paying more
7
268
Saw some confusions, but all Pro plans have no 5-hour limits, and give the following usage compared to Plus.
✅ Pro 100 = 10X
✅ Pro 200 = 20X
✅ Pro 500 = 50X, access to ultrafast
1,262
201
6,634
1,424,631
Mino retweeted
Today we're releasing DeepFinance, the first ultra-long horizon benchmark for high finance work.
A set of 21 problems that reflect work at top PE and IB firms, DeepFinance is built using real deal data.
Problems are constructed by experts with backgrounds from top firms like Evercore, JPM and UBS, and institutions like HBS and Wharton. Problems represent work that would take human professionals 20-40 hours to complete.
Can AI really diligence a $100M+ deal? Finance benchmarks today are limited to evaluating agents on QA on financial docs or grading the numerical correctness of simple financial models. However, real world high finance work is full of assumptions and conflicting evidence. We built DeepFinance to realistically evaluate agents on those end-to-end finance workflows.
DeepFinance leaderboard: juliusresearch.ai/deep-finan…
114
70
755
140,595
Mino retweeted
Anthropic's valuation has now been roughly cut in half from the lofty $2 trillion heights that were being spoken of in financial circles only a month ago. On the plus side, if you are a believer, you can now get in for about half price. The IPO media blitz is starting Monday.
64
30
646
111,815
the post-midterm push for AI regulation is gaining momentum. Reps. Don Beyer and Sara Jacobs plan to release a draft AI safety bill by year-end, targeting House consideration in Q1 2027. their previously reported framework could establish mandatory safety standards for frontier labs, clarify developer liability, and give regulators emergency powers to halt models that fail testing. Beyer is also pushing for a dedicated AI select committee, while other lawmakers are preparing proposals for independent model evaluations and stronger oversight of leading AI companies
1
1
104
OpenAI isn't slowing its infrastructure expansion despite the midterm results, according to CFO Sarah Friar. the company reported 70% QoQ growth in Q3, its enterprise business doubled, and revenue is now split 50/50 between consumers and businesses. with 1.2 billion weekly users, OpenAI plans to keep raising capital and expanding compute capacity to meet demand
Rohan Paul
14
Anthropic analyzed over 207,000 Claude research sessions and estimates that today's AI could take on tasks accounting for a median 38% of research time under widespread adoption. that could nearly double research output in computer science and statistics, while fields like chemistry would see smaller gains due to the need for physical experiments. these are projections based on current capabilities, not observed productivity gains
1
1
121