The mjmirza/app-store-compliance repository provides a playbook and pre-submission guard for iOS and Android apps, including checks for privacy manifests and Data safety forms. Its author says the project documents 117 rejection patterns and fixes and tracks 95 regulatory deadlines. The guard can run through AI coding tools and block submission commands when it detects critical risks. The repository uses the OpenRoots Agent License 2.3 for its current release, and earlier releases have different license terms.
Microsoft released Visual Studio Code 1.147 as a stable version, expanding its AI-agent capabilities. The preview Browser Use feature lets agents operate the computer's default Chrome or Edge browser and use existing login sessions, with user authorization required before accessing sites and users responsible for login and sensitive operations. It requires the GitHub Copilot browser extension, Node.js 22 or later, and a local Copilot agent session. The experimental Send Later feature allows one-time agent requests to be scheduled or modified and canceled before submission; it supports only local sessions with automation enabled, and a computer that is asleep or offline will send the request after resuming. The release also adds experimental Model Routers that let a decision model choose among user-configured models according to task type.
In a Latent Space interview with Linear COO Cristina Cordova, the company said agents created 3% of Linear’s work a year ago, while a later measurement said more than half of the issues created in Linear came from agents. The report did not clarify whether the two figures use identical definitions or measurement periods. It also said Linear has shipped an MCP server, a platform for building agents into Linear, integrations with Codex, Cursor and GitHub Copilot, and its own Linear Agent. The report stated that Linear passed $100 million in annual recurring revenue this year, with more than 40,000 paying customers and 177% net revenue retention; these figures were not independently verified.
Google DeepMind released Gemini 4 Neon, the fourth Gemini 4 model in roughly seven weeks, with a 1M-token context window and four effort levels. Latent Space reports pricing of $3 per million input tokens, $0.15 for cached input and $15 for output, with rollout planned for the paid Gemini API, AI Studio, Android Studio, Gemini for business and Antigravity. Preliminary results placed Neon second on Agent Arena and the Vals Index, while Artificial Analysis reported that it matched Carbon’s Intelligence Index score of 60 at high effort.
GitHub’s Octoverse 2026 report ranks TypeScript first by monthly contributors, ahead of Python and JavaScript, while Rust recorded the fastest percentage growth among the languages compared. TechRepublic reported that AI agents produced the first drafts of more than 40% of GitHub pull requests by September 2026, up from 17% in March, while 55% of GitHub Copilot’s monthly active users used agentic features. GitHub’s figures measure participation through repository activity rather than only unique people writing code, and its contributor projections through September 2027 depend on recent trends continuing.
LangChain says LangSmith for Coding Agents is now available in public beta on LangSmith Cloud, linking agent traces and spend to GitHub pull requests and GitLab merge requests. A new overview dashboard aggregates costs by pull request, work type, model, and pull request state, while tracking work that does not produce a pull request as unattributed spend. The release supports Claude Code, OpenAI Codex, Cursor, GitHub Copilot in VS Code, and Deep Agents.
A translated report on a LeadDev story says some experienced engineers are moving from writing code to orchestrating AI agents, reviewing generated code and validating results; the report also describes burnout and departures from software development. It cites an Oregon State University study of 442 developers at 56 organizations that associated greater AI-related job demands and organizational pressure with more burnout. An Atlassian survey of more than 1,500 software professionals found that 87% believed AI helped them deliver faster, while 49% spent at least four hours a week reviewing AI output or refining prompts and 75% reported greater responsibility for validating the quality and safety of AI-generated content.
A Juejin author compares TRAE, GitHub Copilot, Cursor, Windsurf, CodeBuddy, and Tongyi Lingma across pricing, feature completeness, Chinese-language adaptation, and related criteria. The ratings combine the author's personal testing with cited community evaluations, and the author explicitly notes that they do not represent an industry-wide conclusion.
Microsoft announced at Ignite 2026 that its second-generation Maia 200 AI accelerator has entered mass production and is running Microsoft’s largest frontier AI models in data centers. The TSMC 3nm inference chip has FP4 and FP8 tensor cores, 216GB of HBM3e memory, and can scale over Ethernet into clusters of up to 6,144 accelerators. Microsoft claims up to 1.7 times the token throughput per watt of leading GPU infrastructure across built-in features including Teams, Copilot, and Cowork, but this figure has not been independently verified.
Reverse engineering of MissionControlAggregator.dll found that Windows 11 records bitmap indicators showing which coding agents and developer tools were observed, including Codex, Claude, GitHub Copilot, Cursor, Python and .NET. The researcher verified on a US-configured Windows 11 Home ARM64 installation that the report entered DiagTrack's upload queue when optional diagnostic data was enabled, which was the default in that setup. The report contained no prompt text or source code, and its timestamps represented collection windows rather than per-tool usage duration. Behavior in other regions, editions or configurations was not established.
Satya Nadella said in The Pragmatic Engineer interview that Claude Code caught Microsoft off guard by moving coding toward the command line, while Cursor’s agent mode prompted Microsoft to respond. He argued that as AI-generated code becomes cheaper, verification becomes the harder part of software development, and described Microsoft’s “software factories,” slower junior hiring, changing team structures and voluntary exit program. Nadella also said Microsoft and Carnegie Mellon University launched the PRAISE mentoring program for AI-assisted software engineering.
Anthropic Cybersecurity Skills is an independent, community-created project, not affiliated with Anthropic PBC. It is distributed under the Apache 2.0 license and provides structured cybersecurity workflows for AI agents using the agentskills.io standard. The README describes 817 or 818 skills across 29 or 34 domains and maps the collection to MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, MITRE D3FEND, NIST AI RMF and MITRE F3. It also claims compatibility with Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI and other platforms.
GitHub Blog published a practical guide to agent skills, describing them as versioned directories centered on SKILL.md files that capture reusable instructions for repeatable tasks. The guide shows how related skills can share project context, execution paths and approval checkpoints, while emphasizing that tools, credentials and permissions must still be supplied by the environment. It also explains how to package skills for Claude Code, Cursor and GitHub Copilot, then verify each client's setup and required outputs.
GitHub data cited by InfoQ 中文 shows that Coding Agents drafted the first version of more than 40% of Pull Requests by September, up from 17% in March, while agent-initiated PRs grew roughly sixfold in six months. In a Latent Space interview, GitHub head Jay Parikh said personal experiments need not always follow the PR process, although enterprise projects may still require reviews, permissions, and change records. The discussion also covered model routing, session continuity, code review, software supply-chain security, and measuring the business value of AI coding.
A Chinese technology article summarizes Hamel Husain’s guidance for AI product teams, arguing that successful teams focus on measurement and iteration rather than debating which models, databases, or agent frameworks to use. It recommends examining real interactions, conducting error analysis, involving domain experts, using synthetic data when user data is scarce, calibrating automated evaluations, and planning roadmaps around experiments rather than promised features.
掘金 describes GitHub Spec Kit as a specification-driven development toolkit that moves AI coding through constitution, specification, plan, tasks, implementation, and convergence stages. Its clarify stage asks about ambiguous requirements, while converge compares generated code with the specification and appends tasks for uncovered gaps; in one reported CLI todo test, this process found an unhandled corrupted-file case despite 14 passing tests. The toolkit supports more than 20 coding agents, including Claude Code, GitHub Copilot, Codex, Gemini CLI, Cursor, Cline, opencode, Qwen Code, Kiro CLI, and Zed.
The author argues that coding agents and AI assistants use incompatible memory formats, including AGENTS.md, CLAUDE.md, Cursor rules, Anthropic's memory tool, GitHub Copilot memory, and MCP-based stores. Drawing on OpenTelemetry, the article proposes a neutral specification covering memory identity, addressing, lifecycle, provenance, and shared access, while arguing that storage, ranking, embeddings, and interfaces should remain areas of competition. The author, who builds LoreKit, presents its MIT-licensed schema as one input rather than a finished standard.
GitHub published a developer guide to using Copilot across planning, implementation, verification, and production observation. It presents features including GitHub Agentic Workflows, Plan Mode, Copilot Fusion, and Copilot Dreaming, while arguing that coding agents should support the broader software lifecycle. The article also cites a Faros AI study reporting that pull-request review time rose 441% as teams moved from low to high AI adoption, though its efficiency claims are vendor assertions.
OpenAI launched GPT-6.1 Terra, a lower-cost model aimed at workplace automation. API pricing is $1 per million input tokens and $5 per million output tokens. OpenAI describes this as half the cost of comparable models, though that comparison is a vendor claim. Terra is available through OpenAI products and its API, with deployments also rolling out across GitHub Copilot, Amazon Bedrock and Microsoft Foundry.
Huxiu AI reports that the widely circulated claim that 55% of China’s AI tokens go to short drama and video while 58% of US tokens go to programming and R&D lacks authoritative statistical support. China’s figure came from a company investor-relations record without published sample coverage or conversion methods, while the US figure may have been derived from an OpenRouter-a16z sample that excludes video generation and does not represent total US usage. The report also cautions that tokens measure model throughput rather than GPU compute, making direct comparisons between video generation and coding unreliable.
A first-person engineering account from 掘金 · 人工智能 describes a two-month redesign of an LLM Agent Plan system from a dependency-based DAG to a linear checklist. The author attributes the change to verification failures that rolled steps back, inaccurate dependency declarations, difficulty tracking substantive work, and a mismatch between fixed benchmark tasks and dynamically evolving production work. The report concludes that the engine should provide signals while leaving completion judgments to the model, while noting that DAGs may still suit multi-agent or batch workflows with clear dependencies.
The GitHub Copilot CLI repository introduces a terminal-based tool for working locally and synchronously with an AI coding agent that understands code and GitHub context. It supports natural-language workflows for building, editing, debugging, and refactoring code, alongside GitHub repository, issue, and pull-request integration, MCP server extensions, LSP configuration, and model selection. The repository also documents availability on Linux, macOS, and Windows, an active Copilot subscription requirement, and a container image for Copilot SDK clients running as a headless server.
SimpleEnglish is an MIT-licensed Agent Skill that asks LLMs to write plain English by default and can apply stricter ASD-STE100 rules on request. Its documentation lists compatibility with Claude Code, Cursor, VS Code Copilot, OpenAI Codex, Gemini CLI, Goose, OpenCode, and other Agent Skills-compatible tools. The project reports that version 2.0.1 reduced visible formatting defects by 95% across 16 Claude replies and lowered STE linter violations by 81.3% across nine Claude models, but no evaluation run is available for version 2.1.0 or later.
The blader Humanizer repository describes an installable agent skill that rewrites AI-generated text to sound more human without changing what it says. It is built on Wikipedia’s “Signs of AI writing” guide, checks for 26 writing patterns and supports voice matching. The project says it works with Claude Code, Codex and other agents that support skills, including Gemini CLI, GitHub Copilot and Windsurf; these compatibility claims have not been independently verified. The project also reports that judges preferred its rewrites in all 16 blind-test comparisons, a result that has not been independently validated. It says the tool is intended for human readers rather than AI-detector evasion, and that detectors still flag most output.
Apple released Xcode 27.3 beta 1 with GitHub Copilot available in its built-in coding assistant, according to the release notes cited by 9to5Mac. This expands Xcode’s existing AI integration lineup, which has included ChatGPT, Claude, Codex, Claude Agent, and Gemini across earlier versions.
In a first-person account, the author says they moved from engineering on Microsoft’s Windows AI team to research engineering at Salesforce AI Research, with more time spent supporting scientists and rapid prototyping. They now report managing about five to six Claude Code, OpenCode, Codex, or Pi agents for coding, research, writing, and creative work, while drafting a personal AI policy to guide daily use.
In HPE-produced content hosted by MIT Technology Review AI, the authors argue that AI coding agents consume 10 to 50 times more tokens than basic chat prompts, making token usage and blanket spending caps poor measures of value. They recommend tracking fully loaded cost per merged pull request, alongside rework and defects, while routing routine work to owned open-weight capacity and reserving expensive frontier models for difficult planning or novel reasoning. The piece also says HPE routes workloads from more than 150 internal applications to on-premises open-weight models and reports a 30% to 50% reduction in token costs.
A report based on one author's weekend test compares seven AI coding tools in the same code-review project for a six-person team using a Python backend, React frontend and self-hosted GitLab. The test covers project understanding, code changes, pull request submission and handling review comments. The comparison focuses on code-standard consistency, team permissions, private deployment, CI/CD integration and cross-tool migration cost. It recommends testing a free version through a complete development cycle before deciding whether to upgrade, then evaluating results against the team's size and compliance requirements.
Vorssaint is a free, open-source macOS menu bar toolkit published in the vorssaint/vorssaint-utils repository. Its features include system monitoring, window and Dock management, clipboard history and file tools. The project says its Dynamic Island can follow Claude, Codex, OpenCode and GitHub Copilot, displaying information such as plan limits, token usage, models, projects and task status where available. Users can install individual features. The project requires an Apple Silicon Mac running macOS 14 Sonoma or later and is licensed under GPL 3.0 or later.
Matt Pocock has released mattpocock/skills, a public repository of reusable agent skills for software-engineering workflows, including TDD, debugging, code review, domain modeling and codebase design. The repository provides installation instructions for Claude Code, Codex, GitHub Copilot, Gemini CLI and other agents, while its description says the skills are small, adaptable, composable and model-agnostic; this claim has not been independently tested.
What problem it solves
Developers using AI coding CLIs need a way to delegate work among multiple agents and coordinate their execution. CAO addresses this by running provider CLIs in isolated terminal sessions and giving a supervisor tools for coordinating workers.
Key features
Multi-agent delegation — A supervisor delegates work to specialist agents, with parallel and sequential execution suppor…
What problem it solves
OpenPencil lets users describe UI in natural language and see it appear on an infinite canvas, then modify existing designs by selecting elements and chatting. It also lets multiple AI agents work on different sections of complex pages in parallel while keeping `.op` files human-readable, Git-friendly and diffable.
Key features
Prompt-to-canvas generation — Describe a UI in …
What problem it solves
Repository automation often needs reasoning or interpretation for tasks such as issue triage, pull-request review, CI failure investigation, documentation maintenance, dependency analysis, and repository reporting. The project provides a way to define these agentic workflows in Markdown with YAML frontmatter and compile them into standard GitHub Actions workflows while retai…
What problem it solves
Coding tools use different component formats, model names, permissions, and discovery rules. Contributors need a shared way to write reusable plugins, agents, skills, and commands once while producing compatible artifacts for each supported harness.
Key features
Multi-harness component support — One collection of plugins, agents, skills, and commands can be used across Claud…