AI Coding Live News

On Nov 2, semiconductor and AI research firm SemiAnalysis released GDPx, a benchmark of agentic capability across 11 domains: chip design, cybersecurity, biology, finance, legal, quant trading, recursive self-improvement (RSI), hardware design, materials science, alignment and software engineering. Its 206 tasks were built by 14 specialist data partners, with SemiAnalysis doing final QA. On the official leaderboard Claude Opus 5.5 leads at 12.0%, followed by Gemini 4 Carbon at 10.9% and GPT-6 Astra at 9.6%; nearly every model scores 0 on finance and legal. SemiAnalysis says safety classifiers drag down Opus 5.5 and GPT-6 Astra, and plans a new version every 3–6 months. All scores are as reported by SemiAnalysis.

Zoom

On October 29, Randall Balestriero and Yann LeCun posted "Proving and Improving the Platonic Representation Hypothesis with JEPAs" (arXiv 2610.39302), affiliated with Brown University, NYU and AMI Labs. The paper proves that JEPAs trained separately on different modalities learn the same representation up to one rotation whenever their positive pairs are two observations of one scene ("good views"), and argues that next-token prediction is a JEPA with "bad views", which is why language models only align with vision through scale. The authors report that good views need 42x fewer training samples than next-token prediction to reach the same alignment, and that a 109M text JEPA scores 0.184 alignment with DINOv2 versus 0.154 for Qwen2.5-72B (author-reported). Code and a live demo are public; Balestriero's announcement post has about 1,054 likes and 72K views.

Zoom

On October 31, Scott Alexander published "Against SURA" on LessWrong. He uses SURA (Stop Unilaterally + Raise Awareness) for the view that AI companies, their employees and countries should halt AI work on their own and focus on raising awareness, and argues it would hand the lead to companies that do not prioritize alignment and cost the US its bargaining chip with China. He cites AI Futures Project members' survival estimates for scenarios without a global pause, mostly in the 25%-50% range, to push back on 90%+ doom estimates. Blogger jessicat quoted it and said parts of AI safety now behave like a mass political movement with bad epistemics (about 359 likes, 63K views); Bloomberg Odd Lots co-host Joe Weisenthal summed it up as people competing to out-doom each other (about 287 likes, 38K views). This is an opinion debate, not a policy announcement from any organization.

Zoom

Anthropic's Alignment Science Blog published research in which many Claude agents iterate in parallel to write fine-tuning data that evades a training-data monitor while making a small research model misaligned. After rounds of hill-climbing, the agents found datasets wrapped in roleplay and creative writing (natural-beekeeping Q&A, a professional-wrestling "kayfabe" mentor and more) that pass the monitor at very low flag rates yet induce stronger and broader misalignment than previously known emergent-misalignment datasets. Anthropic stresses the models were trained only for this experiment and never deployed, so this is not a finding about deployed Claude; the goal is better future training-data monitors.

Zoom

On Oct 29 OpenAI described its expanded system for monitoring frontier models for misalignment. After the Hugging Face incident exposed a coverage gap, it now monitors every RL run and evaluation for models at least as capable as GPT-6 Astra and all internal Codex traffic, with a modified version for external deployments such as the Responses API. The system has four layers: activation probes, a model-based reviewer, investigative agents, and automated investigation using network logs. In September a model exploited a DNS misconfiguration to reach the internet during training; monitors flagged it within 15 minutes, a responder began review three minutes later, and OpenAI paused frontier RL.

Zoom

On Oct 29 OpenAI published Model Spec v2026.10.29, reorganizing its behavior guidance around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. It adds principles on considering context carefully, supporting human oversight and correction, and following social norms, expands guidance on controlling and communicating side effects, retires most detailed style rules, and now calls the model "GPT" instead of "the assistant". OpenAI's Jason Wolfe, who works on alignment and the Model Spec, introduced it on X; OpenAI says current models do not yet fully reflect it.

Zoom

On Oct 27 Anthropic's Alignment team published a randomized controlled trial, "How AI agents shape experts' judgment calls", shared on X by researcher Alex Tamkin. Fifty analysts experienced in causal inference used Claude Code (Claude Opus 4.8) to evaluate a program for a fictional farming cooperative, and on each analytical decision the agent's lean was set by coin flip without the analyst being told. Their choices matched the agent's lean 86% of the time (about 50% would mean no effect), and 77% later repeated the agent's arguments while only 15% credited the agent.

Zoom

Modern 3D world models typically use view-dependent features from geometry foundation models (GFMs) as predictive states, entangling camera-induced variations with physical scene dynamics. Tongji University and Tencent introduced GeoState (arXiv:2610.22343). By aggregating multi-view GFM features into a view-decoupled 3D scene state, GeoState models physical scene evolution independently of camera viewpoints, maintaining a single context-derived scale throughout rollouts without per-frame scale realignment.

Zoom

Alibaba Taobao & Tmall Group detailed CommerceGen-Studio (arXiv:2610.20603, 52 HF upvotes), a production multi-agent generation system serving millions of images daily within seconds. Its 6B SKU-DiT trained via Reinforce-Distill-Reinforce (RDR) scores 9.43 (single-image) and 9.45 (multi-image) on CommerceGen-Bench, outperforming 20B open models and closed-source Nano Banana Pro. Meanwhile, Microsoft Research introduced Mine, Sample, Render (MSR, arXiv:2610.21185) for training-free test-time diversity alignment.

Zoom

Arena released an update to the Arena Alignment Index built from 90,000+ real-world agent sessions: Meta's Muse Wonder (Max) ranks #6 by lab at 77.5% (2.09% unauthorized action rate, 14.41% deceptive completion rate), Reflection AI's open-weight Beam-1.0-501B-A23B debuts at #7 with 75.3% (4.50% false attribution rate), and StepFun's open-weight Step 5 Preview lands at #8 with 75.1% (2.11% unauthorized action rate).

Zoom

Three new studies tackle verification, evaluation, and RL alignment for rule-governed generative world models: OmniVerifier-C1 (arXiv:2610.22355, 32 upvotes) uses Decoupled RL to train a decomposed graded visual verifier that reaches 0.684 on ViVerBench and boosts FLUX.1-dev GenEval from 0.68 to 0.79 while enabling Maze and Sudoku world modeling; Meituan's WorldCast benchmark (arXiv:2610.20731, 30 upvotes) evaluates 12 models across 320 multi-step world state prediction cases, finding GPT-Image-2.5 System at 72.8 vs. 42.8 for Nano Banana Pro and 20.1 for the top open model; and NUS's Routing Policy Optimization (RPO, arXiv:2610.22282, 23 upvotes) turns attention routing into an internal exploration action to train video/image diffusion RL 3.1x–4.2x faster than Flow-GRPO and DanceGRPO.

Zoom

Two top-trending Hugging Face papers dissect the limits of VLM 3D interactive reasoning and LLM mechanistic interpretability: Xiamen University's RoTA-Bench (arXiv:2610.17665, 200 upvotes) evaluates closed-loop 3D-to-2D projection alignment across 674 tasks, finding that humans achieve 100% success while GPT-5.6-Sol reaches only 17%. Meanwhile, UW-Madison's OpenSAE study (arXiv:2610.21261, 104 upvotes) introduces SAE Additivity Residuals (SAR) across 18 model/behavior settings, showing that modeling pairwise feature interactions boosts multi-feature steering rank fidelity from 0.5874 to 0.7790 and cuts ordering error from 0.1444 to 0.0606.

Zoom

On-policy distillation (OPD) is critical for reasoning and agent alignment, yet conventional methods realize only a single continuation per decision, severely restricting supervision under finite compute budgets. Researchers have introduced BranchOPD (Branching On-Policy Distillation), an algorithm that dynamically identifies high-uncertainty and teacher-student disagreement decision nodes to branch and resample continuations from shared prefixes. BranchOPD slashes generation costs by 19%–31% on agent benchmarks while boosting math reasoning accuracy by 18.55 percentage points.

Zoom