AI Coding Live News

On 2026-11-05, Hugging Face Harness Optimization researcher Joël Niklaus (@joelniklaus) released the Harness Leaderboard: same model, provider, and reasoning effort with only the harness swapped, covering GLM-5.3-Flash, DeepSeek-V4.1-Flash, and Flint on Terminal-Bench 2.1 and DeepSWE—28,849 trials and about $4,746 of inference. Author-reported averages: mini-swe-agent ~51.4% resolve rate; Pi ~49.4% at about $0.29 per solved task. This is an individual research leaderboard, not a Hugging Face company product announcement.

Zoom

On 2026-11-06 Vercel updated Token Pass so featured-model usage through AI Gateway draws credit automatically with a standard API key, OIDC token, or fx login. When credit runs out, those auth modes fall back to team AI Gateway credits; dedicated Token Pass keys still stop until the next billing period.

Zoom

On 2026-11-05 Dwarkesh Patel released Clocktower Bench: ~$20k API for 386 social-deception games, open harness. Author board: GPT-6 Astra Elo 1859, ~82% win rate vs Opus 5.5.

Zoom

On 2026-11-05 Hugging Face released the free Ultimate MoE Training Handbook. Co-founder Thomas Wolf cited ~800k GPU-hours, ~70k words and 55 interactive figures; author Nouamane Tazi separately cited ~500k GPU-hours. The Megatron-LM guide covers routing, expert sizing, communication and debugging via the official Space.

Zoom

On 2026-11-05 Nous Research released Hermes Agent v0.23.0 (The Hearth). Your gateway can serve the full Desktop in a browser; CLI/TUI/Desktop/editors/bots/cron share one session authority; the plugin catalog grew from ~600 to 1,050+; plus conversation rewind, inline terminal images, and paid-MCP spend limits. Official: ~1,950 merged PRs and 666 contributors.

Zoom

On November 5 Cursor shipped Build Live Apps: describe a tool to a cloud agent and Cursor builds the web app, adds Postgres if needed, sets up Cursor SSO auth and hosting, runs deterministic and agentic security checks, then publishes to a domain you choose with a shareable link for people, groups or the whole team. Works with Origin and GitHub repos on all paid plans with cloud agents enabled.

Zoom

On November 5, the BenchFlow team publicly launched FrontierPhysics (arXiv:2611.06910): 56 end-to-end physics-research tasks authored by 42 physicists from their own work across nine domains. Agents must do literature review, planning, research/experiments, deterministic artifacts, and a manuscript. Across 22 models in 26 configurations, no setup passed more than 46 of 168 rollouts; the best reported Pass@1 is about 27.4% for Claude Opus 5.5 in Claude Code. Expert trajectory review finds the main gaps are rigorous reasoning and expert intuition, not missing physics facts. Artificial Analysis said it will partner to host the evaluation.

Zoom

On November 5 OpenAI launched GPT-6 Cyber for authorized security research: ExploitGym rose from GPT-5.6 Cyber's 32.1% to 48.5% (about 51% relative), out-of-scope exploitation fell to 0.07% without deployment prompts and 0% with them. The model helped find 81 novel high-severity Firefox bugs and develop exploits for 948 Linux privilege-escalation cases, and OpenAI raised its Preparedness Framework cyber rating from High to Critical. It is initially available via API and Codex to select Daybreak Red customers, with expansion to all Red customers planned.

Zoom

On November 5 Baidu (official blog + @BaiduResearch) open-sourced ERNIE-5.5-Flash-Base: a text-only MoE base checkpoint at end of pre-training (no mid/post-training), 280B total / 14B active per token, 33.5T pretrain tokens, up to 128K train context and 1M for fine-tune/inference. FP8 and BF16 weights are on Hugging Face and ModelScope. Under Baidu's unified eval pipeline vs similar bases: MMLU 93.3, HumanEval+ 82.3, LiveCodeBench 40.1 (vendor-reported). Post-trained ERNIE 5.5 Flash "will follow". License is custom (excludes EU/UK/South Korea; commercial use above $50M related revenue needs prior written authorization).

Zoom

arXiv paper 2611.03233 by Yuqi Jia, Yunjie Tian and five co-authors introduces Axion 1.0, an agent foundation model that is open beyond its weights. It uses just two lightweight SFT stages plus reinforcement learning, with no dedicated mid-training, DPO or on-policy distillation, and about 18K SFT trajectories in total. Training code is hosted under the bytedance GitHub organization, and the training data and an Axion 1.0-35B reproduction checkpoint (Qwen3.6-35B-A3B base, 35B total / 3B active, Apache 2.0) are on Hugging Face; weights for the paper's 122B model are listed as coming soon.

Zoom

Indie developer Pieter Levels (@levelsio) said on X on November 2 that he used Claude Code to decompile the engine of the free Battlefield 1942 demo, port it to C, compile it to WebAssembly and host it at bf1942.pieter.com, where it runs in the browser with online co-op on the Tobruk map. He said the job took about 1 million output tokens, roughly $100 at API prices, while he pays for a $100-a-month Claude subscription. His two main posts drew about 2,200 likes and 340,000 views and reached X trending. These are the author's own claims and community discussion, not independently verified.

Zoom

On November 2, MIT CSAIL PhD student Rikiya Takehi posted "Weight-Space Programming" on X: a team (Takehi, Jayden Teoh, Alex Zhang, Omar Khattab) had GPT-6 Astra read and edit neural network weights directly, with no gradients or training. Their blog says Astra changed only 131,072 weights (0.016%) of Qwen3.5-0.8B, fixed its thinking loops and raised MATH-500 from 7.9% to 48.1%. The post drew about 24K likes and 1.77M views, and Ethan Mollick quoted it saying it is a big deal pending replication. All results are self-reported by the authors and not yet independently reproduced.

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, Google DeepMind said PlanAssist, a Gemini-built planning tool developed with the UK government, Google Cloud and Faculty, will be made available to 12 more local planning authorities in England, bringing the total to 15 for a 12-month Beta on live applications. In early pilots at Barnet, Camden and Dorset, DeepMind says the average officer time on householder assessment tasks fell by 58%, and more than 70% of the drafted reports were approved with no edits. Planning officers keep the final decision.

Zoom