Cursor joins The X Subscription, with paid plans offering twice the Grok coding usage of equivalent Cursor plans

#1 𝕏

Cursor announced that it is now included in The X Subscription, with paid plans sharing one usage pool across Cursor, Grok Build, and other Grok products and offering twice the coding usage on Grok models versus equivalent Cursor plans. Existing subscribers can keep their plans or switch; Premium, Super, and Ultra include Grok Bot, while third-party Cursor models move to API pricing under the new plans.

#2 𝕏

QwenGyre was announced as an elastic RL system for extreme-long-horizon agent training with two core designs: an Elastic Scheduler that reallocates GPUs between rollout and training, reporting up to 1.85× speedup over Colocate and 1.78× over Async, and a Trajectory Processor that converts branching executions into prefix-shared trajectory trees with bounded, deduplicated training samples. Scaled to Qwen3.8-2.4T-A95B with ~700K tokens per rollout, it increased NL2RepoBench from 52.5% to 58.5% in 48 steps while reducing end-to-end training time by ~44% versus Async.

#3 𝕏

Mustafa Suleyman commented that Microsoft’s Maia 300 silicon delivers 40% more tokens per watt, recapping Satya Nadella’s announcement about Microsoft Maia silicon scaling.

#4 𝕏

Mistral AI shared that Mistral Large 4 zooms in on, inspects, and verifies engineering drawings until the answer is exact.

Also covered by: @Mistral AI

#5 𝕏

Aravind Srinivas says every Perplexity Project is a Git repository, with agents working in short-lived sandboxes and committing to a shared S3 repo; fewer than 2% of submissions collide. For financial document analysis, agents using plain `ls` and `grep` beat semantic search 63% to 37% on answer accuracy.

#6 🐙 vercel-labs/skills

vercel-labs/skills - **vercel-labs/skills** is an MIT-licensed CLI for installing or using agent skills across OpenCode, Claude Code, Codex, Cursor, and 75 more agents. The repository includes reusable skills, CLI source code, scripts, and tests, with support for public, private, Git-based, local, and direct-download sources. Worth trying if you need one workflow for sharing skills across multiple coding agents.

#7 🐙 openai/math

openai/math - **openai/math** is OpenAI’s shared repository of 719 mathematical manuscripts across 372 families, with preprints, proof artifacts, Lean formalizations, and abridged reasoning summaries. About 42% of top-line results are formalized, while some unformalized results may have issues. Consider it if you need a large case study in model-generated mathematics and staged verification.

#8 🐙 tester-army/e2e

tester-army/e2e - **tester-army/e2e** is TesterArmy’s Apache-2.0 end-to-end testing framework for web and mobile apps, with an SDK, runner, CLI, engines, and examples for Next.js, Expo, SwiftUI, and Flutter. On October 12, 2026, tester-army released @e2e-dev/limrun Version 0.2.0. PR #988 added a Limrun device provider for hosted iOS simulators and Android emulators, with one hosted instance per worker, while PR #1066 made agent-device a regular dependency pinned to the mobile engine’s version. With 9,896 GitHub stars and 5,582 weekly trending gains, consider it if you need agent-driven cross-platform testing. Release notes

#9 🐙 mksglu/context-mode

mksglu/context-mode - mksglu/context-mode is an MCP server that limits raw tool output in coding-agent context and preserves session activity for relevant retrieval. It claims its sandbox tools can reduce 315 KB to 5.4 KB, a 98% reduction, and applies its routing paradigm across 17 supported clients. The repo includes plugins, hooks, configurations, reusable skills, source code, benchmarks, and bundled CLI and server files. Worth trying if you need longer, more continuous AI coding sessions.

#10 🐙 paperclipai/paperclip

paperclipai/paperclip - paperclipai/paperclip is an MIT-licensed Node.js server and React UI for orchestrating agent teams, with a CLI, docs, evals, integrations, and adapter code. paperclipai released v2026.1012.0 on October 12, 2026, with 130 commits and an experimental MCP connection that lets Claude Code, Codex, OpenCode, and generic MCP clients act as signed-in users to review and delegate work. Worth trying if you need one dashboard for agent tasks, permissions, costs, and history. Release notes

#11 🐙 heygen-com/hyperframes

heygen-com/hyperframes - heygen-com/hyperframes is an Apache-2.0 framework that turns HTML, CSS, media, and seekable animations into deterministic MP4 videos, with a CLI and 21 published agent skills. heygen-com released HyperFrames v0.8.152 on October 13, 2026, adding per-revision artifacts, edit-target history, and shared-agent safeguards; the repo has 62,230 GitHub stars and 4,482 weekly trending gains. Worth trying if you need agent-assisted, HTML-based video rendering. Release notes

#12 🐙 thedotmack/claude-mem

thedotmack/claude-mem - thedotmack/claude-mem is an Apache-2.0 project that captures agent activity, compresses it with AI, and injects relevant context into later sessions; it includes plugins, a worker, server runtime, MCP search server, web viewer, and search skills. thedotmack released v13.37.0 on 2026-10-13T03:07:09Z, fixing the broken 13.36.1 marketplace update; the repo has 100273 GitHub stars and 3762 weekly trending gains. Worth trying if you need cross-session agent memory. Release notes

#13 🐙 Graphify-Labs/graphify

Graphify-Labs/graphify - Graphify-Labs/graphify turns codebases, docs, SQL schemas, configs, and PDFs into queryable knowledge graphs using local, deterministic AST parsing for code, with explained edges and no vector store. Graphify-Labs released Graphify v0.9.89 on October 13, 2026, incorporating 16 community pull requests; the repo includes a CLI, reusable skill, docs, tests, tools, and Docker support. Worth trying if you need agents to navigate complex codebases through explicit graph relationships. Release notes

#14 🐙 mattpocock/skills

mattpocock/skills - mattpocock/skills is an MIT-licensed collection of small, adaptable skills for guiding coding agents through everyday engineering work. It includes reusable workflows for requirements alignment, shared-language documentation, test-driven development, and disciplined bug diagnosis, with setup instructions for several agents. Worth trying if you need more structured, composable AI development workflows.

#15 🐙 addyosmani/agent-skills

addyosmani/agent-skills - **addyosmani/agent-skills** is an MIT-licensed repository of 25 reusable skills that encode software-development workflows, quality gates, and engineering practices for AI coding agents. It includes 9 lifecycle slash commands and setup instructions for tools supported by a CLI spanning 70+ agents. Worth trying if you need consistent build, test, review, and shipping workflows across coding agents.

#16 🐙 pbakaus/impeccable

pbakaus/impeccable - **pbakaus/impeccable** is an Apache-2.0 GitHub project shared by pbakaus with design guidance for AI coding agents. It reports 80,549 stars and includes 1 reusable skill, 24 commands, live browser iteration, and 59 deterministic detector rules that can run without an LLM or API key. Worth trying if you need repeatable design workflows and frontend checks in supported coding tools.

#17 🐙 promptfoo/promptfoo

promptfoo/promptfoo - promptfoo/promptfoo is an open-source, MIT-licensed CLI and library for evaluating and red-teaming LLM applications. It includes automated model comparisons, vulnerability and code scanning, local evaluations, and CI/CD checks. Worth trying if you need repeatable testing and security reviews for AI products.

#18 🐙 n8n-io/n8n

n8n-io/n8n - **n8n-io/n8n** is a fair-code platform for building AI agents and workflows with a visual canvas, JavaScript, Python, npm packages, and self-hosted or cloud deployment. n8n-io released n8n@2.43.4 on October 12, 2026, fixing duplicate Chat Hub segment handling in queue mode and resuming tool calls for deferred tools loaded by a skill. Worth trying if you need flexible, model-agnostic AI workflow automation. Release notes

#19 🐙 langfuse/langfuse

langfuse/langfuse - **langfuse/langfuse** contains Langfuse, an open-source platform for tracing, evaluating, monitoring, and debugging LLM apps, with prompt management, datasets, a playground, APIs, and cloud or self-hosted deployment options. langfuse released Langfuse v4.58.0 on October 13, 2026, at 07:05:09Z, adding a media upload pool with concurrency and byte limits, per-project trace idle settings, and project LLM connections for Topics. Worth trying if you need one platform for LLM observability and evaluation. Release notes

#20 🐙 mlflow/mlflow

mlflow/mlflow - mlflow/mlflow is an Apache-2.0 open-source platform for tracing, evaluating, monitoring, and optimizing agents, LLM apps, and ML models; it includes source code, docs, quickstarts, integrations, prompt tools, an AI Gateway, and model lifecycle tooling. mlflow released MLflow 3.17.1 at 2026-10-13T09:33:43Z, adding Decision API support in `make_judge()`, Databricks review queues, and tracing and scorer-version fixes. Worth trying if you need unified AI application and model workflows. Release notes