AI Coding Live News

On Oct 16 China's commerce and science ministries proposed adding 40 technologies, 23 of them in AI, to the export ban and restriction catalogue. The draft restricts core technology of models with at least 10^26 training FLOPs, at least 2 trillion parameters, or a global top-10 ranking, and sets step, concurrency and tool-count thresholds for agents. Open-sourced technology is exempt, and API or SaaS services are not treated as technology exports. Temporary pre-export registration took effect the same day; comments close Nov 15.

Long-horizon LLM agents accumulate expanding interaction histories that saturate GPU KV-cache memory and choke self-attention throughput. While sparse attention algorithms mitigate these bottlenecks mathematically, existing production serving engines (e.g. vLLM) are tightly coupled to dense tensor layouts, preventing heterogeneous sparse mechanisms from sharing infrastructure. Researchers from Harbin Institute of Technology (HIT) release SparseEngine (arXiv:2609.39068), an open-source, sparse-first inference engine built from the ground up for long-horizon agent workloads. SparseEngine establishes a shared lifecycle contract enabling 15 diverse sparse attention methods across four distinct families to control their custom KV representations while unifying memory management. It introduces Chain Cache to resume stateful KV-evicted contexts across multi-turn requests and Controllable Prefix-Cache Pruning to discard stale history without disrupting prefix matching. Empirical evaluations on long-horizon agent benchmarks show that SparseEngine delivers over 10x higher throughput under KV eviction, over 2.5x faster decoding throughput than vLLM at matched concurrency, and more than 2x end-to-end task speedups with zero degradation in model reasoning accuracy.

At HUAWEI CONNECT 2026, Huawei Cloud launched its Agentic Hybrid Cloud architecture, introducing the AI Cluster Service (AICS) and Agentic Model-as-a-Service. The upgraded AgentArts orchestrator provides deterministic concurrency and secure data isolation across 100+ enterprise customers.

Dana (@MatureIntellect) shipped Bench v2.1.0: Claude Code runs under a written constitution with three-stage edit review and a tamper-evident ledger. v2.1 adds prompt caching, per-stage model control, installer/CLI, concurrency locks, weekly cost/latency reports, and 1,036 passing tests.

Dan Kornas (@DanKornas) launched Apache-2.0 open-source Sortie: define tracker, coding agent, and concurrency in WORKFLOW.md beside the repo, then schedule GitHub/GitLab/Gitea/Linear/Jira issues into isolated workspaces. Adapters cover Claude Code, Copilot, OpenCode, Codex, and Kiro, with SQLite run history, retry queues, and stall detection.

LightSeek Foundation released a Kimi K3 draft-model collection (EAGLE-3, DFlash2, DSpark) trained with TorchSpec on 40 NVIDIA GB200 GPUs, with open data recipes. Their blog reports acceptance length across 10 benchmarks and SPEED-Bench end-to-end throughput; DFlash2 averages ~4.02 accepted tokens and leads EAGLE-3 at lower concurrency.

ClaudeDevs announced that standard Claude Code weekly usage limits will rise permanently by 25% on September 14, 2026 for Pro, Max, Team, and seat-based Enterprise plans. Until that date, the current 50% promotional increase remains in effect. Anthropic also stated the new permanent baseline is about a 17% reduction versus the allowance developers are using today. The company said it is working on product changes that should make remaining quota feel more productive, with better visibility and control over usage. For teams running long-horizon coding agents, the practical implication is a two-step calendar: keep the promo window for large refactors and eval loops through mid-September, then resize concurrency, /loop jobs, and default effort levels to the tighter permanent cap. Heavy Claude Code users should treat the dashboard, not older Help Center copy, as the live schedule.

Milvus released an official integration plugin for DeepSeek Harness 0.1, providing high-concurrency hybrid vector retrieval and secure read-only context injection for open-source autonomous agents.

Chroma published Fission, a concurrency-control protocol engineered for multi-agent swarms updating shared Foundation memory concurrently without write collisions or Git-style merge locks.