AI Coding Live News

In his 2027 budget address to the National Assembly on November 2, South Korean President Lee Jae Myung proposed a record 820.9 trillion won budget (+12.8%), with 21.3 trillion won (about $16B) for semiconductors, physical AI, AI data centers and AI, nearly double the 10.8 trillion won in 2026. The Ministry of Planning and Budget document allocates 4.7 trillion won to a sovereign AI model aiming for world-leading performance, including the purchase of 10,000 Vera Rubin GPUs. Parliament still has to pass the budget.

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

On Oct 28, Microsoft reported results for its fiscal Q1 2027, the quarter ended Sep 30: revenue of $92.3B (+19%), Microsoft Cloud revenue of $63.9B (+30%), Azure growth accelerating from 42% last quarter to 48%, and commercial remaining performance obligation of $677B (+73%). CEO Satya Nadella added AI figures on X: Foundry revenue more than tripled year over year, revenue from open source models grew 10x, Copilot daily users grew nearly 50% quarter over quarter, and Agent 365 now holds over 50M agents. Separately, TechCrunch, citing Microsoft's regulatory filing, reports $10B of revenue from OpenAI-related commercial arrangements this quarter.

Zoom

Cognition, the maker of Devin, released FrontierCode 2.0 on Oct 27. It keeps the method of maintainer-written tasks graded on whether the code is actually mergeable, but draws tasks from more repositories, adds security patches and performance work, uses deliberately underspecified prompts and adds many thoroughness criteria. In Cognition's results Claude Opus 5.5 leads at 45.5%, GPT-6 Astra follows at 42.8%, and Grok 4.7 and Gemini 4 Argon drop into the 20s, leaving the benchmark far from saturated.

Zoom

On Oct 26 Springer Nature, one of the world's largest research publishers, announced a research partnership with Google DeepMind to co-design an AI peer review assistant. It uses DeepMind's frontier Gemini models and builds on Google's earlier agentic research tools ScholarPeer and Deep Review to help reviewers navigate manuscripts, analyse large amounts of text, charts and figures, and flag areas that may need closer attention. The system is built with Google's open-source Agent Development Kit (ADK) and runs only inside Springer Nature's private Gemini Enterprise environment, so manuscript content stays in its secure environment. A prototype is being tested; the publisher stresses 'human first, AI second' and has not given a launch date or any performance data.

Zoom

Independent AI evaluation platform Artificial Analysis (@ArtificialAnlys) officially benchmarked DeepSeek's web search API (DeepSeek Search). Augmenting the base model with DeepSeek Search boosted its Search Index score from 33 to 69—an impressive +36 point leap (ranking 17th among 27 tested search APIs). DeepSeek Search established a market-leading price-performance ratio, priced at just $0.75 per 1,000 searches (~$0.032 per completed task), making it the most cost-effective paid search API evaluated. It recorded 70% accuracy on AA-Omniscience and 63% on BrowseComp, highlighting an optimal balance between factual grounding and operational economics.

Zoom

Published in arXiv:2610.25834, FrontierRefactor establishes the industry's first evaluation benchmark targeting autonomous coding agents on multi-scale, performance-oriented codebase refactoring. Unlike traditional benchmarks like SWE-bench that measure only functional pass rates, FrontierRefactor spans seven programming languages and 60 Amdahl-qualified tasks ranging from kernels to full repositories (10^1 to 10^4+ LOC). Initial evaluations reveal that over 41% of agent patches pass functional equivalence suites yet yield zero verified speedup—with 19.3% triggering severe runtime performance regressions, highlighting the vital need for multi-turn profiling feedback.

Zoom

The vLLM project team and Ant Group officially open-sourced Humming (humming-kernels), a specialized high-performance kernel library designed for serving quantized MoE and dense LLMs at massive scale. Humming natively supports WnA16, W8A8, W4A8, and W4A4 across INT1–INT8 and FP4–FP8 formats, spanning NVIDIA architectures from Turing through Blackwell as well as T-Head Zhenwu 810E silicon. In production benchmarks, Humming kernels outpace cuBLAS by 3–4x at small batch sizes and surpass Marlin by 2x on H200 clusters. Serving DeepSeek-V4.1-Flash across 8x H20 GPUs, Humming drives a +43.3% throughput gain under 64 concurrent requests while reducing time-to-first-token (TTFT) by 32%, pre-integrated directly into vLLM v0.29.0+.

Zoom

Researchers introduced EdgeKernelBench and the LEKA agent architecture in arXiv:2610.24648, establishing the first unified framework for evaluating and generating low-level high-performance kernels across heterogeneous edge backends. While LLM-driven kernel generation has demonstrated success on datacenter GPUs, edge deployments remain plagued by hardware diversity, fragmented compiler toolchains (CUDA, OpenCL, TileLang, Triton), and strict memory-compute constraints. EdgeKernelBench bridges this gap through a unified intermediate representation (IR) suite spanning 331 canonical workloads. The companion Layered Edge Kernel Agent (LEKA) hierarchically decomposes optimization decisions, unlocking efficient automated kernel synthesis on edge silicon.

Zoom

Hours after DeepSeek open-sourced DeepInfer, the vLLM project team released official benchmark parity data serving DeepSeek-V4.1-Flash. Utilizing DeepSeek's identical decoding setup (131K context window, nextn=3 speculative decoding), vLLM closed its throughput performance delta to within 6% of DeepInfer across batch sizes 8 through 256 on NVIDIA GB200 systems. On GB300 hardware, vLLM attained up to 1,000,000 tokens/second per GPU and 826 tokens/second per user, generating $38.96 in hourly GPU revenue at DeepSeek off-peak rates. vLLM simultaneously shipped day-0 runtime support for FastH3-Omni and Spark-X2.6-4B.

Zoom

In small language models, vocabulary embedding tables often account for over 60% of total parameters, posing severe bottlenecks for edge deployment and ternary quantization. A new study (arXiv:2610.22428) replaces trainable embeddings with a single frozen 16 MB table of 512-bit binary codes derived via PCA and sign binarization. On Gemma 3 270M, trainable parameters collapse from 268.1M to 101.0M (-62.3%) while matching or exceeding original performance across HellaSwag (30.4%), LAMBADA (23.2%), and PIQA (61.5%).

Zoom

SemiAnalysis revealed the first Google TPUv7 benchmark submission for the 2.8T Kimi K3 MoE model, led end-to-end by Inferact (the creators of vLLM) and Google using the TorchTPU vLLM stack. At a hardware rental cost of just $1.21/chip-hour (vs $2.31 for GB300 and $1.86 for GB200), TPUv7 delivers ~2.2M tokens per $1 TCO in the ultra-high interactivity regime (~390 tok/s/user), outpacing NVIDIA GB200 NVL72, B300, and AMD MI355X on cost-performance.

Zoom

To combat goal drift in long-horizon agent tool-use and multi-agent systems, NUS, USTC, and Alibaba Qwen introduced Steering Agent (arXiv:2610.22316). Jointly trained via pairwise ranking and Group-Relative Policy Optimization (GRPO), the agent operates a low-cost fast path to detect drift and a slow path to generate grounded interventions. It raises Multi-Hop QA benchmark performance from 42.70 to 50.10 while cutting executor turns from 8.8 to 6.5 (-26.1%) without modifying the underlying executor models.

Zoom

Tackling reward hacking and benchmark saturation in coding agent environments, CMU researchers introduce Scalable Weak-to-Strong Verifier Hardening (arXiv:2610.15924)—combining an isolated verifier sandbox with an iterative hacker-fixer loop that slashes exploit success across Terminal-Bench, KernelBench, and SETA, and is now officially adopted by Terminal-Bench. Complementing this, Recursive Environment Scaling (RES, arXiv:2610.16397) recursively mutates execution dependencies across 294 tasks—exposing a 57.14-77.14 point drop on Claw-Eval-Live while boosting equal-budget SFT performance by 7.00-14.61 percentage points across all nine model-benchmark pairs.

Zoom