Loading activity…
@wolfejosh
co-founder + partner @ Lux Capital | Trustee @SfiScience Santa Fe Inst | Founding Chair @CiPrep (Brooklyn) | Co-Founder of Carson, Quinn & Bodhi w/ @ltwolfe
AI
Cognition says Devin Fusion beats Claude Code with Opus 5.5 on performance and price, at 31% lower cost per task.
AI
Runway's MCP page tells Grok Bot users to add its plugin and sign in.
AI
Modal says the gap emerged under heavy sandbox load, while Vera was not always faster on a quiet host.
AI
Together says developers can define training loops using reinforcement learning or supervised fine-tuning while it handles the infrastructure.
AI
A post calls them the “best company” and “best city” but does not name either.
AI
Runway says the tool generates full soundscapes from video, synced to every frame while keeping dialogue intact.
AI
Anduril says the US Air Force pilot served as an airborne “quarterback” during the October flight.
AI
A post says ANALYST handles investigations, multidisciplinary reviews and checkouts, with a human auditing each step.
AI
Nominal poses the question in a post linking to an image, without giving an answer in the text.
AI
Benchling says agent teams handle tasks and coordinate handoffs, while scientists still make the important decisions.
AI
Sakana AI says coding agents test hypotheses against preset standards, share results and check which parts of successful programs matter.
AI
Sakana AI says My Fish includes a timeline of memory activity and a map showing how memories connect.
AI
The creator says they made two papercut images with ChatGPT, then used Claude for the animation and audio mashup.
AI
Its creators also report 10–21× faster unconditional generation on three datasets, with comparable or better generative perplexity.
AI
Tenzai says GreenfieldVC curated the class at TechCrunch Disrupt.
AI
A post describes LatchBio as a Lux family company and connects OncologyBench to AI and cancer.
AI
Runway says users can animate charts or make explainers in Claude Motion, then generate videos and images in Runway.
AI
The @appliedcompute announcement calls Spork an efficient post-training recipe and reports a 100% rise in its chess-puzzle solve rate.
AI
SemiAnalysis says RL practitioners burned by trainer–inference mismatch tend to avoid it.
AI
Cognition says Argon scored 60.5% on FrontierCode 1.1, making it the highest-performing Gemini model on that leaderboard.
