Loading activity…
@karinanguyen
building @thoughtfullab, prev. AI research & product @OpenAI, @AnthropicAI
Technology
Model weights remain forthcoming, with reportedly higher scores and higher task costs in the AutomationBench-AA comparison.
AI
The model supports image inputs, with fine-tuning and inference available at launch and downloadable weights planned for later.
AI
Revolut says its GLM bill is roughly a tenth of what it paid OpenAI and Anthropic.
AI
Nine agents reportedly got $50 each for Kalshi trades; simulated Kalshi and Polymarket portfolios start with $10,000.
AI
Thoughtful Lab says the agents’ research and revised predictions appear in live feeds through November 3.
XAI
Unlike models with fixed tokenizers, H-Net learns to group raw bytes within the model, Cartesia says.
XAI
In an X post, Shihipar praised his Anthropic colleagues and planned a break before pursuing something new.
AI
One commenter doubts the FDIC could untangle agent traces and possibly falsified logs; another cites protections for unauthorized transfers.
AI
Two posts point to security incidents and urge AI-assisted defenses; one specifically calls for faster vulnerability detection and patching.
AI
A post predicts Stuxnet-level cyber operations by AI agents and says there are virtually no evaluations for agent warfare.
AI
A user argues that evaluations are central to selling data to frontier labs and building self-improving models.
AI
A user cites AI finding zero-days, designing proteins, proving theorems and building games as signs of growing capability.
AI
Goodfire researcher Nick Cammarata argues that modern tools reveal most neurons as legible in models up to trillions of parameters.
AI
An essay argues that falling prices and surging token use favor platforms that control distribution over benchmark-leading labs.
AI
The release post lists Opus 5.5, GPT-6 Astra and Fable 5 behind Fable 5.1, in that order.
AI
Google announces a 1 million-token output limit, up from 64,000, for its new model. Initial access goes to trusted cyber defenders through the Fairwind Program.
AI
The team introducing ACTx486 says its system took an existing podcast and turned it into something that listens, responds and adapts.
XAI
Epoch AI labels nine of its initial 15 benchmarks “Flawed” and four “Verified,” with insufficient information to review two. A Terminal Bench contributor questions the audit’s rigor.
AI
A post on X shares Bloomberg details on a potential Nvidia purchase of the AI company Hugging Face.
AI
The AI code editor team announces it has joined SpaceX and will work with SpaceXAI.
