AI News

Reasoning models achieve strong performance through long chains of thought (CoT), but often spend more tokens than needed to solve a problem. Existing ways to shorten the CoT require additional training for each model, add a fixed steering vector to the model's hidden states, or use decoding heuristics such as stopping early or banning words like "wait", which can reduce accuracy. We propose DoubtLess, a training-free decoding method in which a much smaller reasoning model steers the large one towards brevity. The small model reads the large model's CoT so far twice, once prompted to be brief (the brief thinker) and once prompted to be verbose (the verbose thinker), and the difference between its two logit vectors, scaled by a strength, is added to the large model's logits at every step of the CoT. On seven benchmarks of competition mathematics, graduate-level science and competitive programming, DoubtLess shortens the CoT of Qwen3-32B by 31% with unchanged accuracy, outperforming five inference-time baselines. We also show that DoubtLess reduces thinking tokens per turn by 20% in a self-debugging coding agent with unchanged task performance. It also generalizes to two other model families with unchanged accuracy. When combined with training-based compression, it further shortens the CoT of a model already trained for brevity. Ablations and behavioral analysis show that DoubtLess reduces repeated checking, with most token savings coming after the correct answer first appears in the CoT. Our code is available at https://github.com/leoluolol/DoubtLess.

Off-policy data in reinforcement learning with verifiable rewards (RLVR) can arise from asynchronous generation, rollout-training mismatch, and response reuse. Token-wise importance weighting avoids multiplying likelihood ratios across a response but generally biases policy gradients. We prove that, under coverage and local expressiveness conditions, full future KL credit makes the exact proximal optimizer the update's unique stationary policy, even for mixtures of older policies, without full-response weights or a fitted critic. We further establish convergence of continuous-time tabular population updates under persistent coverage and time-varying behavior policies, and a finite-sample reward guarantee for linear logistic models under additional assumptions. These results motivate Proximal KL, a practical off-policy RLVR method that separates the proximal reference from the sampling policy and replaces PPO clipping with bounded future KL credit. Across all seven matched off-policy comparisons on Qwen3-1.7B and Qwen3-4B, spanning two learning rates, response reuse, and rollout staleness, Proximal KL outperforms GRPO in micro-averaged accuracy over eight math benchmarks by up to 2.58%. A separate on-policy comparison yields comparable aggregate accuracy. In periodic-refresh experiments, Proximal KL avoids the collapses observed in PG and CISPO, exceeds GRPO's training accuracy, and reaches higher accuracy in fewer policy updates. Ablations show that token-wise correction outperforms sequence-level and unweighted controls. With token-wise correction retained, future KL feedback reduces sensitivity to the tested baseline perturbation relative to PPO clipping. A finite future horizon improves AIME24 mean@16 and pass@16 over the full tail, while fixed-policy measurements show that truncation reduces correction variance over the tested horizons.

Nvidia CEO Jensen Huang said he plans to unveil $2 billion in new European investments next week, as the continent’s economic prospects get a boost from AI.

New on our Engineering blog: Creating in-house source control.

In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.

Here, we share how we…

One of the core promises of AI is that the more it knows about a person, the better it will be able to assist them with tasks and make appropriate suggestions, which, if you'll believe the tech companies, requires apps and devices to listen, watch, or record users' activity all the time and gather all possible context.

Led by Headline, the funding will support the London-based company's expansion into the US and Asia-Pacific as it tackles lengthy certification processes for AI-powered medical devices.

ExcaliClaw is a collaborative workspace for hand-drawn diagrams. Organize them in collections, search instantly, share by link or by name, and draw together in real time. Its MCP server lets Claude, Cursor or VS Code create and edit diagrams for you. Generous free plan: 200 drawings and unlimited AI (MCP) use, no credit card. Paid plans from €9/month. Version history, trash and backup export included. English, Español, Français.

SolidCraft: an open-source parametric 3D CAD application built in pure Rust. Desktop, web, and automation through MCP. - bherbruck/solidcraft

From new Silicon One-powered systems to an open optics architecture developed with Nvidia, Cisco's 2026 Open Compute Project (OCP) Global Summit announcements underscore the company's push to deliver an open, programmable AI infrastructure stack spanning silicon, networking and operations.

Over the last few months, we’ve made significant progress in reducing the effectiveness of distillation attacks against Claude.

In this post, we explain how: https://t.co/uWiRSLBuOS

📊 GPT-6.1 Sol: 52 on the AA Intelligence Index, at $0.72 per task.
GPT-6 Astra: 53, at $3.26 per task.

⏸️ OpenAI also canceled GPT-6.1 Astra’s release after the model showed deception in internal tests.

Full breakdown in The Batch 📬 https://t.co/wgJqCZ7wHW

#DeepLearningAI #AIEngineering #AIAg…

A cross-platform desktop app to manage Agent Skills in one place and sync them to multiple AI coding tools’ global skills directories — “Install once, sync everywhere”. - qufei1993/skills-hub

Explore 29 interactive PyTorch architectures with guided explanations, captured tensor shapes, and runnable source. No installation or Python needed.

Frontier language-model agents can now improve robots on their own: they write control code, test it, and keep what works as skills and memories for the next task. We introduce SEAR (Self-Evolving Agentic Robotics), a framework and benchmark built around two questions. Is the agent really learning, or just trying more often? And does what it learns still help on new tasks, in a new simulator, and on a real robot? The framework runs an agent in three modes: it writes code that becomes the policy (code-as-policy), acts live in the control loop (agent-as-policy), or works as a researcher that fine-tunes a vision-language-action model (auto-research). The benchmark adds controls that separate retained experience from repeated attempts, familiar tasks, privileged simulator state, larger budgets, and luck. Across seven frontier language models, 330 simulated and five physical tasks, and over four years of agent runtime, we highlight three findings. First, the learning curve of a robotic agent depends both on the overall time budget and the task difficulty relative to model capability. Harder tasks make longer sessions worthwhile; on easier tasks, re-rolling short attempts does just as well. Second, control code and shared routines transfer across tasks and simulators, while task-specific solutions do not. On a real robot, we observe a performance gap between agent-designed practice and novel human-designed scenes. Third, different models evolve more effectively in different ways. Some do best acting live, others writing code first. And a text-only model keeps pace with a comparable multimodal one: perception helps with error assessment, but agentic reasoning and tool use can matter more. We release the benchmark and codebase to support future research.

Banger paper from Google DeepMind on co-evolving an agent's harness and its weights.

(bookmark it)

If you already log your harness changes and results, those logs can become training data for the model running inside the harness.

This is a nice paper from Google showing how to do this co-evoluti…

Today's startup funding report shows investors pouring money into the technology needed to bring AI into the real world, including humanoid robots, computer chips, advanced cooling systems, energy storage, and nuclear fusion. Chinese robotics startup Astribot led today's verified funding activity with RMB 1 billion in fresh financing at an RMB 15 billion valuation, while two

Gemini 4 Argon is now globally available.

Built to solve your hardest problems from start to finish, the model brings frontier capabilities across coding, enterprise workflows, and cyber defense.

The German energytech unicorn is scaling its Heartbeat AI platform while considering US expansion and revisiting plans for a potential IPO.

AI that makes sure devs understand what they've vibe coded - bhurghundii/vibecheck.dev

An experimental pattern-first DAW. Contribute to andrewfader/blokkily development by creating an account on GitHub.

NVIDIA OCP Summit 2026 keynote tonight at 5:45 PM PT in San Jose: VP Ian Buck unveils Spectrum-XGS Ethernet, a new networking tier that lets GPU clusters in separate data centers operate as one AI super-factory, alongside Vera Rubin NVL72 production data showing 4.8x token throughput and the Vera

At the 2026 Open Compute Project Global Summit, Meta is sharing the next generation of open networking infrastructure, including:  Our first liquid-cooled network rack (IcePack), and 102.4T switch …

Cisco, Google, Meta, Microsoft and Visionbay are joining many making advances with Vera Rubin.

A collection of free, single-purpose generative design tools. No accounts, no uploads, everything runs in your browser.

Record your screen and the demo looks great on the first take: zooms planned from your clicks, smooth cursor, 3D camera moves, cursor themes, backgrounds and webcam layouts. Then let your AI agent do the edit over MCP, or cut the video yourself by deleting words in the transcript. Windows, macOS and Linux. Free, MIT, no account, no watermark.