Loading activity…
@omarsar0
Building and measuring frontier intelligence @dair_ai • Prev: Meta AI | Galactica LLM | Papers with Code | Elastic | PhD
AI
Rohan Paul says Emma’s face, voice and conversational reactions come from recordings made with an actress.
AI
His early-access testing favored one agent for coding projects, but clear goals and human oversight still mattered.
AI
The reported workflow pairs brokerage explanations with text messages asking users to approve suggested trades.
AI
Steve is reportedly designed to record what went wrong and simulate the outcome of a decision before acting.
AI
Private datasets now carry 45% of the index’s weight, up from 25% in the first version.
AI
Drex reportedly assigns probabilities to fixed-choice answers instead of generating text.
AI
The benchmark tests agents' research methods and software use; every model reportedly scored zero on its end-to-end research track.
AI
Aina says the ring uses context from devices and routines to suggest actions users can accept or dismiss.
AI
Atmee v2 uses one image and 10 seconds of audio to create a character for live video conversations.
AI
TwIL-8B's open-source release drew praise from a commentator who sees locally run models aiding scientific discovery.
AI
Aether's extra credits cost $0.10 for 12 minutes on a small machine, with idle time free, a user says.
AI
Nvidia's tests put Opus 4.5 at 70.2 with compaction after every action versus 70.0 with full history on SWE-bench Verified.
AI
The Stacks Project spans hundreds of chapters and tens of thousands of results; AxiomMathAI's version is reportedly open and machine-checked.
AI
Living Brain is described as keeping context in a separate store that agents can access through MCP or an API.
AI
Skill++'s authors report that it selected every required skill for 76.3% of queries, versus 36.4% with original metadata.
AI
The managed inference product is pitched for spiky search, OCR and agentic workloads, with a $500 launch-credit offer.
XAI
Traversal says its software finds the root cause of production incidents and helps prevent future ones.
XAI
Its reported matched-compute comparisons favor Locus over selected baselines and AI agents, with detail on property prediction.
AI
In the authors' tests, changing opening tokens mainly helped models add or restyle information, not consistently leave it out.
AI
Revolut says its GLM bill is roughly a tenth of what it paid OpenAI and Anthropic.
