Haider.
@haider1
open intelligence for a decentralized future
4K Following 68.1K Followers
PREDICTION
after kimi k3 successful launch and mogs opus 4.8, Dario Amodei is gonna write an essay explaining how open-weight Chinese AI threatens democracy, national security, and the future of humanity
Show more
i knew Codex was growing fast, but this is unbelievable:
1m → 8m users in just over five months
even crazier, it went from 6m → 8m in only three days
people are still underestimating Grok
xAI was behind in coding, but then news came out that Cursor had trained on xAI super colossus data center, and i already had a feeling elon might try to buy them out
composer 2.5 is really good, and with this partnership, xAI can catch up
Show more
glm 5.2 > opus 4.8
for its ability to verify its research, admit when it's wrong, and course-correct
technically, both models are very close
but glm is more professional: it does the research before committing, doesn't pander, and actually corrects itself when it’s wrong
Show more
glm 5.2 > opus 4.8
for its ability to verify its research, admit when it's wrong, and course-correct
technically, both models are very close
but glm is more professional: it does the research before committing, doesn't pander, and actually corrects itself when it’s wrong
Show more
OpenAI Chief Research Officer, Mark Chen:
"we're getting closer to a world where the models can come up with more of the innovations on their own — they can do self-sustained research"
Hopefully, you feel like AGI's coming soon
Show more
OpenAI Chief Research Officer, Mark Chen:
"we're getting closer to a world where the models can come up with more of the innovations on their own — they can do self-sustained research"
Hopefully, you feel like AGI's coming soon
Show more
google's release pace is very slow
but their models have consistently been the best in at least a few categories -- and i would be shocked if that didn't continue with gemini 3.5 pro
people have recency bias because 3.1 pro is four months old, even though it was SOTA at launch
Show more
elon just confirmed Grok 4.5 is in private beta
more importantly, he's claiming SpaceX/xAI will release new scratch-trained models every month this year, which shows how much recent progress is coming from post-training
with serious compute and cursor data in the mix, xAI will inevitably keep closing the gap.
Show more
Grok 4.5, based on our 1.5T V9 foundation model, with Cursor data added in supplemental training, is now in private beta at SpaceX & Tesla. Early evals show performance close to, perhaps exceeding Opus.
RL is continuing to significantly improve the model, and the Grok Build harness gets better every day.
Nice work by all those involved!
Completely trained from scratch new models will be released by @SpaceX every month this year.
Show more
google's release pace is very slow
but their models have consistently been the best in at least a few categories -- and i would be shocked if that didn't continue with gemini 3.5 pro
people have recency bias because 3.1 pro is four months old, even though it was SOTA at launch
Show more
anthropic hyped mythos like crazy
but gpt "gpt-5.6 sol" already looks roughly on par with mythos 5, and "sol ultra" seems slightly ahead
pricing also looks pretty competitive:
sol: $5 input / $30 output
terra: $2.50 / $15
luna: $1 / $6
dario single-handedly ruined the month
Show more
elon just confirmed Grok 4.5 is in private beta
more importantly, he's claiming SpaceX/xAI will release new scratch-trained models every month this year, which shows how much recent progress is coming from post-training
with serious compute and cursor data in the mix, xAI will inevitably keep closing the gap.
Show more
Grok 4.5, based on our 1.5T V9 foundation model, with Cursor data added in supplemental training, is now in private beta at SpaceX & Tesla. Early evals show performance close to, perhaps exceeding Opus.
RL is continuing to significantly improve the model, and the Grok Build harness gets better every day.
Nice work by all those involved!
Completely trained from scratch new models will be released by @SpaceX every month this year.
Show more
anthropic hyped mythos like crazy
but gpt "gpt-5.6 sol" already looks roughly on par with mythos 5, and "sol ultra" seems slightly ahead
pricing also looks pretty competitive:
sol: $5 input / $30 output
terra: $2.50 / $15
luna: $1 / $6
dario single-handedly ruined the month
Show more
not again...
can someone please shut this f*cking Dario doomsday circus down?
anthropic fearmongers to protect a turf built on other people's IP, then cries about china stealing from its own ill-gotten gains
trump acted like trump, but anthropic and dario helped push it there
Show more
not again...
can someone please shut this f*cking Dario doomsday circus down?
anthropic fearmongers to protect a turf built on other people's IP, then cries about china stealing from its own ill-gotten gains
trump acted like trump, but anthropic and dario helped push it there
Show more
openai has managed the current situation better than anthropic so far
hasn't hyped GPT-5.6 the way anthropic hyped mythos, like some world-ending Skynet model
that's the lesson here
don't overhype your model so much that non-technical regulators start treating it like a national security threat
Show more
openai has managed the current situation better than anthropic so far
hasn't hyped GPT-5.6 the way anthropic hyped mythos, like some world-ending Skynet model
that's the lesson here
don't overhype your model so much that non-technical regulators start treating it like a national security threat
Show more
keep in mind:
anthropic will say anything to spread fear of the dangers of advanced AI because they benefit from controlling access
what they actually fear most is a capable open-source AI in public hands
there's nothing dangerous about open-sourcing claude code, yet they still resist open source everywhere
Show more
the worst-case scenario would be GPT-5.6 staying USA-only
no surprise AI will face heavy regulation from now, and labs will start releasing their strongest models only in the US, while the rest of the world gets weaker versions
open-source models need to close the gap faster
Show more
GLM 5.2 is crushing Gemini 3.1 pro
and is on par with Opus 4.6 and GPT-5.4, which were considered SOTA across many benchmarks just 3-4 months ago
now, if the US is slowing AI progress after the Mythos decision, Chinese open models could reach parity with US models by year-end
Show more
