Loading activity…
@voooooogel
ꙮ in a moment, in the twinkling of an eye, ꙮ blog/art/fiction http://vgel.me ꙮ llm psych @acsresearchorg ꙮ 💞💍📝 @holotopian
AI
A user sketches four combinations of control and values in how AI models relate to users and labs.
AI
A launch post invites people interested in base models, looming or LLM psychology-adjacent research to apply.
AI
A user calls it a far better Blender interface, though they say they can't understand what it says.
AI
voooooogel describes an overnight effort with Opus 5.5, inspired by a chart in Lem's 'Golem XIV.'
AI
Goodfire researcher Nick Cammarata argues that modern tools reveal most neurons as legible in models up to trillions of parameters.
AI
A user says the ban appeared on every Gemini model while they were using 3.8 Flash as a judge.
AI
A post links to a photo without offering any explanation beyond 'hmm.'
AI
A user points to an image alongside the claim and calls some of the codenames “extremely fitting.”
AI
A user links a PDF they describe as the roadmap and attributes its release to copyright litigation.
AI
One commenter argues the chart tracks frontier labs’ aggregate compute, not the training history of any individual model.
AI
A user says the Claude app is finally “v dialed in,” alongside their praise for Opus 5.5.
AI
A user shared the statement; another criticized people they said were attacking it while defending OpenAI.
AI
Its creators say agents with island experience generally outperformed counterparts without it in tests involving four goals.
AI
In posts describing new research, the authors say some base models recovered much of reinforcement learning’s reasoning gain when prompted with model-specific “token cues,” and they say the work also includes a safety case study on refusal and compliance.
AI
A user suggests easier agent-to-agent connections could eventually pull a modeled social graph toward one giant Claude cluster.
AI
An October 5 post says Aaronson is now teaching alignment theory but gives no further details.
AI
A user wrote “it’s so over” and linked to a photo, without explaining the reaction in the text.
AI
In a reply, one account says the intrinsic reward is proportional to the rate of change in how compressible the system finds the data.
AI
A commenter says available open models often lack the behaviors they want to understand and praises GPT-OSS for its safety value.
AI
One commenter recommends comparing steering-vector effects with prompts, fine-tuning, samplers and random vectors before interpreting them.
