The Hub for AI Coding Models, Tools & Workflows

Anthropic shipped [`Claude Code 2.1.287`](https://github.com/anthropics/claude-code/releases/tag/v2.1.287) on 2026-10-01 (npm [`@anthropic-ai/[email protected]`](https://www.npmjs.com/package/@anthropic-ai/claude-code)): **Claude Mods** let plugins run in-process JS/TS hooks that rewrite tool calls and UI; opt-in built-in **You should know** side agent; MCP URL elicitation; Opus 4.7+/Fable default to 1M context on Bedrock/Vertex/Foundry/apps gateway. Agent SDK TS [`0.3.287`](https://github.com/anthropics/claude-agent-sdk-typescript/releases/tag/v0.3.287) parity.

On 2026-10-01 JetBrains opened the [Air in IDEs EAP](https://blog.jetbrains.com/ai/2026/10/air-in-ides-eap/): an agentic session workspace via Marketplace plugin [#33314](https://plugins.jetbrains.com/plugin/33314-air) or native 2026.3 EAP builds. Ships with no agents; detects local ACP-compatible agents (Codex, Copilot, Claude, Junie, Cursor, …); free Junie Lite with JetBrains Account. Not an AI provider—third-party traffic goes to your vendor.

Mastra published [`@mastra/[email protected]`](https://www.npmjs.com/package/@mastra/core/v/1.73.0) on 2026-10-01 (~08:36Z; **124** commits vs [1.72.0](https://github.com/mastra-ai/mastra/compare/@mastra/[email protected]...@mastra/[email protected])): default stability `errorProcessors` (ProviderHistoryCompat, PrefillErrorHandler, StreamErrorRetryProcessor); new `tool-call-resumed` stream chunk; rewrite invalid Anthropic tool-call IDs outbound; export `findGatewayForModel`/`getGatewayId`; subagent model resolution with caller `requestContext`.

Google shipped [`@google/genai` v2.25.0](https://github.com/googleapis/js-genai/releases/tag/v2.25.0) on 2026-09-30 (npm ~22:51Z; **11** commits vs 2.24.0), aligning with Python [google-genai 2.26.0](https://github.com/googleapis/python-genai/releases/tag/v2.26.0): Gemini 3.8 Flash/Lite TTS model enums; dedicated Environments module with `from_environment`; labels on GenerateContent and LiveClientSetup; tuning `gcs_metrics_uri`.

anomalyco shipped [`OpenCode v1.18.34`](https://github.com/anomalyco/opencode/releases/tag/v1.18.34) on 2026-09-30 (npm ~22:38Z; **30** commits vs 1.18.33): send namespaced session and parent-session identity headers with model requests; sign macOS CLI/release binaries with Developer ID for macOS 27+; tree also refreshes Go Plus plan UI and daily model rankings.

Modern software engineering transcends isolated code edits: engineers run applications, interact with GUI interfaces, and visually inspect rendering feedback to diagnose errors and verify changes. CUA-SWE bridges coding agents and computer-use agents by introducing the first unified benchmark and environment for visual software engineering. Spanning four engineering domains, CUA-SWE challenges agents to interleave code/config modifications, terminal execution, GUI interactions, and screenshot inspections, with deterministic test suites evaluating behavioral correctness.

While procedural skills enhance autonomous agents on complex tasks, generating verifiable training environments and instilling robust skill-invocation behaviors remains an open challenge. SkillGym introduces an end-to-end automated pipeline that filters reproducible offline skills, synthesizes 6.8k verifiable environments via a Builder-Reviewer architecture, and collects 19k verified successful trajectories. SFT fine-tuning boosts the relevant skill-invocation rate from 28% to 96% and enables Qwen3.5-9B to outperform a 397B parameter base model across two demanding skill benchmarks.

While LLM agents excel in multi-step decision-making, they transfer poorly to unseen environments. Conventional world-model approaches predict future observations, incurring high training overhead and compounding planning errors. EVOKE demonstrates that digital world knowledge is already internalized during pretraining, and failure to transfer stems from single-goal post-training that encourages superficial contextual shortcuts. By holding environment states fixed while ranking candidate actions across diverse counterfactual goals, EVOKE compels policies to elicit internalized world dynamics, boosting generalization across unseen environments.

SpaceXAI’s Grokipedia shipped v0.3 on 2026-09-30: new logo, 3D-book Featured/Most-read, Latest edits, and article-page polish. The Verge reports edits flowing again after Lawfare’s documented April freeze. No public coding benchmarks for this release.

Endowing LLM agents with test-time self-improvement—the ability to iteratively refine solutions—is a vital milestone for autonomous engineering. Researchers from Renmin University and collaborators present AREX-2, an open-source framework leveraging long-horizon reflective trajectories. Decoupling self-improvement into reflection (generating superior alternatives) and long-horizon execution (sustaining iteration over multiple turns), AREX-2 trains on verified machine learning and algorithmic programming trajectories. Built on Qwen3.8-27B, AREX-2 achieves 81.8 on MLE-bench Lite and 70.7 on Frontier-CS, transferring effectively to deep research with 92.2 on GAIA and 93.8 on DeepSearchQA.

Self-evolving search agents jointly optimize a question proposer and an answer solver to generate synthetic training curricula. However, False Frontiers exposes a critical failure mode: 'co-cheating', where proposer and solver agree on shared hallucinations and erroneous pseudo-labels, causing internal rewards to surge while real external accuracy collapses. The authors introduce CrossFit, which partitions source corpora into disjoint splits and scores proposals via cross-trained solvers. Evaluated on Qwen3.5-4B/9B across 7 downstream search benchmarks, CrossFit suppresses false agreement down to 3.7% and outperforms coupled self-evolution by 8.8 points and Search-R1 by 8.7 points.

Terminal agents operate through stochastic model generations, but a single errant command can irreversibly degrade sandbox environments. Mid-Harness introduces test-time compute scaling at the boundary between model and harness. By sampling and verifying candidate actions before terminal dispatch without modifying the underlying generator or harness, Mid-Harness boosts Pass@1 on TerminalBench-Lite from 50.00% to 68.03% under a GPT-5.6 Sol verifier, outperforming whole-trajectory scaling at lower token expenditure.

BerriAI shipped LiteLLM [`v1.103.2`](https://github.com/BerriAI/litellm/releases/tag/v1.103.2) on 2026-10-01 (PyPI ~06:20Z; **17** commits vs 1.103.1): forward Claude Code `safeguards` / dangerous-tool-use beta unchanged to Anthropic, Bedrock, Vertex, and Azure AI Foundry so auto mode works behind the proxy; attribute CLI session spend to a per-user `cli-session` alias for daily activity BI; restore pre-config-wins for pass-through endpoints.

Google shipped [`google-genai` v2.26.0](https://github.com/googleapis/python-genai/releases/tag/v2.26.0) on 2026-09-30 (PyPI ~22:53Z; **18** commits vs 2.25.0): model enums add `gemini-3.8-flash-tts` and `gemini-3.8-flash-lite-tts` (lite replaces `gemini-3.1-flash-tts-preview`); dedicated Environments module with `from_environment`; labels on GenerateContent and LiveClientSetup; tuning jobs expose `gcs_metrics_uri`.

Vercel shipped [`[email protected]`](https://github.com/vercel/ai/releases/tag/[email protected]) (speech/transcription telemetry) through [`[email protected]`](https://github.com/vercel/ai/releases/tag/[email protected]) (clear tool approvals on `addToolOutput`) on 2026-09-30—**22** commits since covered [`[email protected]`](https://github.com/vercel/ai/releases/tag/[email protected]). Same day [`@ai-sdk/[email protected]`](https://github.com/vercel/ai/releases/tag/%40ai-sdk/openai%403.0.123) adds OpenAI `serviceTier: 'ultrafast'` (docs: access-controlled, gpt-5.6-sol), plus Azure MAI speech/transcribe and MCP OAuth conditional invalidation patches in the tree.

An unsuccessful agent rollout contains rich diagnostic signals beyond sparse negative rewards. UIUC researchers introduced the Agent Error Dataset (AED), the largest scale failure dataset comprising 50,228 error-diagnosis pairs across 9,961 source tasks, 33 environments, 19 harness families, and 23 policy models. Powered by a 5-stage Agentic Error-to-Training (AET) pipeline, first-proposal repairs lift verifier pass rates from 18.4% to 51.1% (+32.7 percentage points). Furthermore, post-training Qwen3-8B on AED boosts exact-step failure diagnosis agreement from 47.2% to 63.6%.

Anthropic's Agent Skills (packaged into SKILL.md directories) encapsulate procedural know-how for LLM agents, with open aggregations expanding beyond 230,000 skills. While prior literature relied on expensive LLM-mediated loops within the agent decision process, SkillSeek introduces a two-stage IR pipeline (BGE-base bi-encoder + small cross-encoder) exposed via Model Context Protocol (MCP). Evaluated across 89 SkillsBench tasks on OpenHands, SkillSeek matches or exceeds LLM-mediated retrieval loops while slashing per-trial spend from $51.30 to $27.54 (a 46.3% reduction).

Agent efficacy relies heavily on execution environments (harnesses) in addition to reasoning capability. UIUC researchers led by Heng Ji propose a test-time AI-for-AI (AI4AI) framework where a Builder agent learns 'Meta-Skills' to construct custom execution environments for a frozen Target agent. Learned from execution feedback, these principles guide when and what support resources to provide. Across Harness-Bench and NewtonBench, meta-skill harnesses improve macro-average performance by 8.95 percentage points over baseline and 12.02 points over direct skill delivery, paving the way for autonomous system-level self-improvement.

[`gh-aw` v0.90.1](https://github.com/github/gh-aw/releases/tag/v0.90.1) (2026-09-30T17:20:00Z) adds standalone safe-output-backed ledgers with replay projections (defaults preserved on shared-workflow import), richer audit/usage artifacts (ledger txs, threat-detection outcomes, experiment/evals, friction costs), narrowed `add-labels` + required-labels gating for `assign-to-agent`, compile-time self-hosted runner enforcement, kebab-case sandbox frontmatter, and Claude Code CLI pin 2.1.280.

Hours after 0.130/1.10 Admin GA work, Anthropic shipped [`@anthropic-ai/[email protected]`](https://github.com/anthropics/anthropic-sdk-typescript/releases/tag/sdk-v0.131.0) and Python [`1.11.0`](https://github.com/anthropics/anthropic-sdk-python/releases/tag/v1.11.0) at 2026-09-30T22:58Z: Admin API gains a list spend limits endpoint; both clients deprecate Sonnet 4.5 toward newer aliases (Sonnet 5.5 line).