Loading activity…
@dorialexander
building open ai infrastructure @pleiasfr — χαλεπὰ τὰ καλά
AI
The reported MATH-500 score rose from 7.9% to 48.1% with the same sampling settings.
AI
Giotto's website lists its own Alto and Solo benchmark comparisons against other AI models.
AI
A user links to @pieterdelobelle's teaser and mentions having thoughts about the instrument.
AI
An announcement says Pleias will attend NVIDIA GTC in Berlin, where its October 21 mixer will focus on synthetic data.
AI
One poster asks whether people want general intelligence; another jokes about researchers modeling consciousness at a DH conference.
AI
A user compares that potential shift to writing and the printing press, which spread intellectual output widely.
AI
A user says Baguettotron's feature connections differ sharply from Gemma-270m's, perhaps more because of synthetic pretraining than deep architecture.
AI
Goodfire researcher Nick Cammarata argues that modern tools reveal most neurons as legible in models up to trillions of parameters.
AI
A post links to a photo without offering any explanation beyond 'hmm.'
AI
A post says theoretical physicist Suvrat Raju couldn't sleep for a few nights after pondering the results' implications.
AI
A user monitoring the models says download counting may have changed slightly, but the decline has continued.
AI
One post questions how much revenue OpenAI keeps after partners take a cut and argues token growth matters more.
AI
A user shared the statement; another criticized people they said were attacking it while defending OpenAI.
AI
OpenAI says it drew on an independent mathematics advisory group’s advice about how to release the results.
AI
Reflection says Beam uses 23B active parameters and plans to publish its full weights later in October 2026.
AI
Aleph Alpha says the bilingual mixture-of-experts model has 78.1 billion parameters and a context option up to 1 million tokens.
AI
Posts on X congratulated Nathan Lambert and Trillium Labs after a launch, with several users describing the effort as open, science-oriented or focused on open post-training. Lambert later said some criticism came from Substack and more general-audience AI followers, not the technical AI community.
AI
The author claims their synthetic environment generalizes across benchmarks, including in telecommunications and healthcare, with results close to Qwen 0.6B.
AI
A user linked to an earlier post of their own, writing, 're-upping for no reason.'
AI
A user says ‘dots’ is a much better name and links to a photo.
