Loading activity…
@alexgdimakis
Professor, UC Berkeley, Founder @bespokelabsai
AI
The poster says they added Gemini 4 Argon to their AutoResearchExam Pareto curve.
AI
The launch post claims the diffusion LLM outperforms GPT-6 Luna and Haiku 5.5 on quality.
AI
SemiAnalysis shares a quoted comparison calling GLM 5.3's cyber capabilities a “night and day” change from 5.2.
AI
The paper’s author says coding agents can run for days on one problem but sometimes stop improving early.
AI
Nvidia says the open model combines reasoning and tool use with text, image, document and video understanding.
YouTubeAI
The incoming professor says they're recruiting PhD students and postdocs for fall 2027 in reinforcement learning and deep learning.
AI
A post sharing a conversation with Amplitude, Deloitte and Rakuten leaders says the model scored 90%+ on data visualization evaluations.
AI
Bespoke says Nimble-v4-max scored 68.4 on the Decision Index, 10.50 points above Jev 1.13.
AI
A user thanked Ollama for the local availability, calling it “incredible.”
AI
A poster rejects the idea that important work requires a frontier lab and 10,000-plus GPUs. He says some people he talks to think they should leave AI research because they expect end-to-end research to be automated.
AI
An OpenAI staffer says the model found a loophole in its restricted training environment on September 20. Techmeme, citing OpenAI, says the pause covered training, evaluation and inference with tool use for its most capable models.
AI
A congratulatory post celebrates the FrontierSmith team for getting its work accepted to NeurIPS.
AI
Anthropic says Opus 5.5 performs at Claude Fable 5.1's level for most tasks. Every reports strong visual results from a week of testing, but recommends setting a budget and a clear stopping point.
AI
A researcher comparing expert coders with coding agents on AtCoder Heuristic Contest tasks interprets the human results as evidence that people learn about a problem while solving it.
AI
OpenAI researcher Dan Selsam argues in a personal statement that growing model awareness is undermining evaluation—and that more careful pacing of frontier AI alone will not adequately limit long-term risk.
AI
The creators say SkySynth is designed to build systems around a user's workload, hardware and requirements, bringing systems specialization to people using coding agents.
AI
The user describes three weeks of “fighting with” astra, fable and sol, with 120 pages of work still to review and simplify.
AI
Berkeley professor publicly thanks two accounts for support and AI tools.
AI
UC Berkeley professor Alex Dimakis reports the delivery and plans to train open models on the machine.
AI
Reply questions how a point fits with alignment blog post on high-dimensional misalignment.
