AI Coding Live News

On November 3, Menlo Ventures released its fourth annual enterprise AI report. It estimates enterprise spending on using AI reached $176.8B by September 2026, up 5.4x from $33B in 2025. In enterprise LLM API spend, Anthropic rose from 40% to 55%, OpenAI from 27% to 31%, and Google fell from 21% to 9%. Coding is the largest app category at $34.0B. Figures come from a survey of about 400 US enterprise AI decision-makers plus third-party data, so they are estimates; Menlo is an Anthropic investor.

Zoom

On November 2, Stanford researchers Owen Queen and James Zou published an interactive report, "The shelf life of LLM findings", retesting 50 widely cited claims about LLM behavior on 20 models released from July 2024 to September 2026 by eight developers. Verdict: 22 claims (44%) expired, 17 (34%) weakened, 8 (16%) persist and 3 (6%) strengthened; the median time from publication to expiration for expired claims was 14.7 months. Capability failures such as letter counting, lost-in-the-middle and fabricated citations have largely faded, while preference issues such as sycophancy and judges favoring their own answers persist.

Zoom

Anthropic's Frontier Red Team published a research post on November 2 and open-sourced FlashProve under Apache 2.0. A model provider commits to hidden weights once, then attaches a zero-knowledge proof to each reply, so a user without the weights can check on an ordinary CPU that the reply really came from that model. That targets routing services that quietly swap in cheaper models and tampered token streams. Anthropic reports about 110 tokens/s on one H100 for Llama 4 Maverick (400B MoE, 4-bit), proofs of about 19 MB per 1,024-token request, and roughly 70-95x the GPU time of vLLM inference. It is an unmaintained research prototype limited to 1,024-token contexts and greedy decoding of an integer version of the model. Anthropic engineer Steve Weis shared the release on X.

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

A widely discussed experiment on X: on Oct 31 indie developer div (@divagr1925) showed QwenPU, which compiles a 32-bit RISC-V CPU plus a vector unit directly into the weights of Qwen3.5-4B and runs Andrej Karpathy's 260K-parameter stories260K inside the model's own forward passes, matching a reference implementation token for token. None of the added weights are trained; a compiler computes them all. The post drew about 1,540 likes, 163 reposts and 85K views and became an X trending story. It is a personal, self-described work-in-progress project: the checkpoint is 26.7 GB and one story token takes about 16 seconds on an L4. This is a write-up of the discussion, not an official announcement from any company.

Zoom

HarnessPro (arXiv 2610.26770, submitted Oct 19 and revised Oct 20), from Tsinghua University, Z.AI (Zhipu) and collaborators, is a label-free way to improve an agent harness: specialized agents diagnose execution traces, rewrite instructions and runtime policies, score candidates with LLM-generated tests, and keep Lite, Medium and Heavy harnesses along the performance-cost frontier. Per the paper, with GPT-5.5 and Codex the Heavy mode reaches 80.8% on Terminal-Bench 2, 4.5 points above the prior RHO method, while cost per task drops from $1.09 to $0.72. The code is MIT-licensed and drew wider media coverage this week, including from Pandaily.

Zoom

Cloud security firm Wiz published research on October 30 saying at least nine AI developer platforms disclosed incidents this month: Fireworks AI, Andromeda, Zeabur, Lightning AI, OpenCode, Railway, Warp, CommandCode and Northflank. Attackers hunted stored environment variables, especially keys for paid AI models; Wiz traced two stolen enterprise keys to roughly 800,000 and 1.7 million Amazon Bedrock calls, which it estimates likely ran over $100k at list price in those two cases alone.

Zoom

On Oct 28 Cartesia chief scientist Albert Gu (co-author of Mamba) and team published two research posts on a new generation of hierarchical networks (H-Nets). In a controlled scaling ladder, swapping BPE tokenization for learned dynamic chunking alone gave up to 3-5x compute efficiency. A larger two-stage byte-level H-Net (3.67B parameters), compute-matched per byte against SmolLM3-3B, reached SmolLM3’s final average downstream score after about 1T tokens versus 11T (vendor-reported). The new architecture is not released yet; Cartesia says it plans to release it over time.

Zoom

On Oct 27 OpenBMB (ModelBest) open-sourced the MiniCPM-OCR suite with two models: MiniCPM-OCR 1.0, a 1.3B-parameter document parser that scores 93.68 on Real5-OmniDocBench and 54.7 on OCRBench v2 (vendor-reported), and MiniCPM-OCR Jev, a 2B OCR quality judge that scores any parser's output region by region from 0 to 1 and flags content that needs human review. Code and weights are Apache 2.0, with day-one support for vLLM, SGLang, llama.cpp, Ollama and multiple chip platforms.

Zoom

On Oct 28 DiffuSpace, whose members come from the HKU NLP group behind Dream 7B, released its first model, Clover-35B-A3B-Preview: a mixture-of-experts diffusion language model initialized from Qwen3.6-35B-A3B, with 35B total and about 3B active parameters, a 128K context window and block-parallel denoising. Vendor-reported results include 73.6 on SWE-bench Verified, 51.7 on Terminal-Bench 2.1 and 83.8 on GPQA Diamond, with 680 tokens/s on a single GB300 in agent workloads. Weights are Apache 2.0 on Hugging Face, ModelScope and GitHub, with SGLang and llama.cpp support.

Zoom

On Oct 26, Google's official @googlegemma account said a set of vLLM optimizations built with Gemini 4 Argon lifted gemma-4-E2B-it throughput on a single A100 from about 37,900 to 108,000+ tokens/sec, roughly 2.85x, with 100% token parity. The PRs were filed by Luciano Martins of Google DeepMind, and most are still under review.

Zoom