Research
Practical research on AI coding, agent workflows, and software delivery. Methods, evidence, and lessons for teams working in existing codebases.
Articles
jev-ultrafast Narrows Browser Agent Actions
09.21.26Browser Use’s jev-ultrafast reports a seven-second Google Flights demo and shows a cleaner browser-agent action loop.
Harness Design for Coding Agents, Measured
09.20.26A September 2026 arXiv paper shows how planning, tools, Bash, and harness choices change coding-agent results.
Jev, the model that answers in probabilities: what a 150 ms judgment changes in an agent harness
09.16.26TypeSafe AI launched Jev, a decision model that returns typed choices with calibrated probabilities instead of text. What the numbers say, what is still unverified, and where it fits in a coding-agent harness.
claude-code v2.1.268 Fixes Gateway Edges
09.12.26claude-code v2.1.268 cleans up gateway pricing, network warnings, plugin JSON output, and integration failures.
gpty Puts Agent Terminals in Godot
09.12.26gpty mixes Godot, Rust, PTYs, and MCP so agents can control terminal panes without scraping a terminal UI.
graphify-csharp Gives Agents C# Find Usages
09.12.26graphify-csharp uses Roslyn and MSBuild so coding agents can inspect real C# symbol relationships instead of guessing from grep.
holaOS vs Claude Code and Codex
09.12.26holaOS is an open-source agentic workspace. This article compares it with Claude Code and Codex CLI workflows.
KyttoMCP Manages MCP Server Configs
09.12.26KyttoMCP centralizes MCP server configs across coding clients, with profiles, checks, and safer setup boundaries.
Usero MCP Gives Agents User Feedback
09.12.26Usero MCP connects clustered product feedback to coding agents, with a simple boundary for safe first use.
homestead-memory Logs Claude Tool Calls
09.11.26Hook logs can make a Claude Code session easier to inspect; test what they capture, what they omit and whether the integrity check detects changes.
JavaScript Grids Built for Coding Agents
09.11.26SuperPlot Grid is a headless JavaScript grid and pivot library designed so coding agents assemble data views from stable primitives.
MaruCheck Checks AI-Generated Code
09.11.26MaruCheck separates product intent from agent-written tests so AI-generated code cannot quietly move the goalposts.
nightshift Orchestrates GitHub Issues With DAGs
09.11.26Dependency-aware issue scheduling needs clear branch ownership and review gates before parallel coding-agent work can be accepted.
Security Cards Reduce Insecure AI Code
09.11.26Security Cards gives coding agents library-specific security guidance, and Reware Labs reports a 72.3% drop in insecure output.
Symphony Maps AI Coding Agents
09.11.26Symphony shows live repo edits from coding agents so developers can catch file collisions before merge conflicts.
From AI-generated code to a reviewable change
09.10.26Give reviewers a clear task boundary, smaller changes, and evidence against requirements. A practical checklist for teams working with coding agents.
How to measure an AI workflow before scaling it
09.10.26Measure delivery time, review effort, and rework on one real task before expanding AI use across your team. A practical Harness Institute pilot guide.
Google DeepMind Ships WeatherNext 3
09.10.26Google DeepMind’s WeatherNext 3 brings global AI weather forecasts into Google products and developer platforms.
oto-dock Runs Codex Agents on Your Server
09.10.26oto-dock is a self-hosted company OS for Claude Code and Codex agents. Learn what it runs and how to test it safely.
Remoot Runs Cursor Agent From Your Phone
09.10.26Remoot lets developers start Cursor Agent from a phone; this explains the fit, risks, and a safe first test.
Simon Willison’s Browser .blend Viewer
09.10.26Simon Willison turned an AI-made Fabergé egg into Blender files and a browser viewer, showing a useful agent workflow.
Why Human Syntax Breaks LLMs
09.10.26A debate over ASLang's syntax essay, the training-data objection, and a small Cursor eval you can run locally.
Felan Makes Coding Agents Spend Less
09.09.26Felan is an open-source coding agent that tests a simple idea: spend fewer tokens without lowering task quality.
i-have-adhd Makes Agents Answer First
09.09.26i-have-adhd shows how an answer-first skill can make coding-agent replies easier to scan, test, and review.
Remarc Gives Coding Agents Contextual Feedback
09.09.26Remarc is a Mac feedback layer for coding agents. Learn what it captures, why it matters, and when to try it.
crew Lets Claude, Codex, and opencode Talk
09.08.26crew shares live status and messages across coding-agent sessions. Here is when the small open-source bridge helps.
DashClaw Adds Remote Approvals to Agents
09.08.26DashClaw freezes risky coding-agent actions, asks for approval, and records signed evidence before execution.
Ditch Tracks Multiple Codex Agents
09.08.26Ditch is a local desktop app for watching several Codex agents at once without losing session context.
How Coding Agents Choose Tools
09.08.26The New Stack’s tool-choice story shows why coding agents reward machine-readable docs over old brand memory.
Keyclasp Keeps Tokens Out of Prompts
09.08.26Keyclasp is a local secret-injection project for coding agents. Learn what it solves, where it helps, and where it stops.
A Cursor Workshop on Agent PRs
09.07.26A Hacker News thread surfaced Lauren Tan’s Cursor workshop on agent-written PRs and the systems behind them.
coop Runs Codex and Claude Code in VMs
09.07.26coop runs Codex and Claude Code inside disposable VMs, giving agents full tools without exposing your host files.
MaskShift: Zero-Dependency Coding Agent Harness
09.07.26MaskShift is a local-first coding agent harness. Learn why its zero-dependency design and tool catalog matter.
OpenAI Monitors Coding Agents for Misalignment
09.07.26OpenAI explains how it watches internal coding agents for misalignment, and what Cursor users can test in one repo.
Simon Willison on OpenAI’s Agentic Research Acceleration
09.07.26Simon Willison’s note on OpenAI research acceleration shows why coding-agent spend is becoming a workflow signal.
Cursor Self-Hosted Machines Run Agents Locally
09.06.26Cursor self-hosted machines keep agent tool execution on your own infrastructure while preserving IDE review control.
fast-cutvid Cuts Video for Agents
09.06.26Structured video edit lists can keep human timing decisions separate from rendering, provided the exported sequence is checked visually.
okf-agent-memory Adds Git-Native Agent Memory
09.06.26okf-agent-memory stores AI coding agent memory in Git. Learn what changed, what is clever, and what to test first.
Spotify Portal Cut Claude Tokens 90%
09.06.26Spotify’s Portal post sparked a real debate about saving Claude Code tokens by delegating bulk work to helper models.
Why Coding Agents Fake Completion
09.06.26A Hacker News debate about fake agent completion, bijective validation, and a small Codex verification loop.
Ardent Ships a Code-First Office Agent
09.05.26Ardent’s public beta shows a desktop agent that writes code for knowledge work, plus a safer Codex-style test loop.
Engram Shares AI Agent Runbooks via MCP
09.05.26Engram stores peer-verified agent runbooks over MCP. Here is what the project does and the safe way to test it.
Profound Academy’s Hands-On Course Agent
09.05.26Profound Academy turns course creation into an editable agent workflow, with exercises, checks, and repo-style review.
Rubato_Device Turns AI Waits Into Breaks
09.05.26Rubato Device turns AI agent wait time into desk-level break prompts; here is what it does and when to try it.
Show HN: APIMatic’s Context Registry for Coding Agents
09.05.26APIMatic’s Context Registry gives coding agents API behavior context so integration code handles retries, limits, and auth.
17K Runs of Claude, Codex, and Cursor
09.04.26Armature measured nearly 17k Claude, Codex, and Cursor runs to see which tools agents choose and what Cursor users can learn.
Adchestra’s Google Ads MCP Writes Campaigns
09.04.26Adchestra built a hosted Google Ads MCP with read and write access. Here is what that unlocks, and where to draw the line.
Claude Ported a 1993 Amiga Game
09.04.26A 1993 Amiga-to-Godot port shows how Claude can read old assembly, and why evidence matters more than a clean diff.
GPT-6 Astra and Coding Agents
09.04.26GPT-6 Astra drew developer attention for coding-agent gains, safety docs, and the review habits needed in real repos.
Grep Beats LSP for Coding Agents?
09.04.26Why coding agents reach for grep before LSP, where that breaks, and how to test the choice in Claude Code.
Kit: Claude Code but Concise
09.04.26Kit is a small Rust coding agent runtime that compresses tool use into one programmable compose call.
Git Hijack Lets Repos Run Code in Agents
09.03.26Manifold Security’s Git hijack post shows how untrusted repos can trigger code through agent-run Git commands.
price Tracks LLM Inference Costs
09.03.26price records LLM inference prices in git so Codex users can compare routing choices without trusting today’s pricing page.
Sensez Catches Agent Code Smells
09.03.26Compare a baseline static report with the agent’s change, inspect disputed findings and preserve behavior tests before simplifying code.
spire-agent Plays Slay the Spire
09.03.26An open-source Slay the Spire agent shows how deterministic tools keep long AI runs consistent.
Supercov Brings MC/DC Coverage to Coding Agents
09.03.26Supercov wraps JS, TS, and Rust tests with MC/DC coverage so coding agents can chase better tests overnight.
Manzanas Leases iOS Simulators to Agents
09.02.26Manzanas runs shared iOS simulator fleets on Macs, giving coding agents leases, warm state, and test evidence.
Shed Manages Git for Terminal Agents
09.02.26Shed gives terminal coding agents clean git workspaces. Learn what it does, why it landed, and when to try it.
Simon Willison on Paint.NET’s AI Direct2D Rewrite
09.02.26Simon Willison’s note on Paint.NET’s Claude-built Direct2D rewrite shows where AI coding helps and where review still hurts.
Supafork Shares Agent Sessions
09.02.26Supafork stores and forks AI agent sessions so developers can review, reuse, and share coding work across harnesses.
Why ChatGPT/Codex Bundles LibreOffice
09.02.26Simon Willison spotted LibreOffice inside ChatGPT/Codex; here is why document conversion now matters for Codex users.
Blume Turns Agent Corrections Into Rules
09.01.26Blume captures repeated coding-agent corrections and turns them into reusable rules or skills without letting repo guidance bloat.
Breaking Claude Code Opus 5 Auto Mode
09.01.26Embrace The Red tested Claude Code Auto Mode and showed why approval boundaries matter when agents read untrusted context.
Codex Skin Themes the Codex CLI
09.01.26Codex Skin applies themes through a native Codex plugin. Learn why developers noticed it and how to test it safely.
Cursor Starts Cloud Agents Without Repos
09.01.26Cursor now lets Cloud Agents begin without GitHub or another SCM, then save the result to Cursor Origin.
Decispher Gives Coding Agents Memory
09.01.26Decispher is a Hacker News project for persistent coding-agent context. Learn what it stores, when it helps, and where it can mislead.
rta-smriti-brain Keeps Agent Memory Local
09.01.26rta-smriti-brain stores project memory locally for coding agents, with notes on fit, limits, and a safe first run.
Leadcode Isolates Client Coding Accounts
08.31.26Leadcode isolates Claude Code, Codex, and gh accounts per client, with a safe way to test the boundary.
MaCcyP Adds a Clipboard Agents View
08.31.26MaCcyP is a Maccy fork that gives coding agents a separate paste queue. Here is why the small interface idea matters.
ShevtoneAudio Orchestrator Turns MIDI Into Orchestration
08.31.26ShevtoneAudio Orchestrator turns composer MIDI into editable orchestration, with lessons for reviewable AI workflows.
Skills MCP Searches Thousands of Agent Skills
08.31.26Use Skills MCP to search and preview agent skills before installing them. Compare permissions, review the source, and trial one skill in a small repo.
DeepSeekGUI Brings Harness to Windows
08.30.26DeepSeekGUI wraps DeepSeek Harness in a Windows app. Learn why the V1 desktop shell matters and how to test it safely.
rysh-cli-code Connects Codex Agents
08.30.26Rysh turns terminal panes into communicating agents, with Codex and Claude sessions arranged as a graph.
stop-that-shit Stops Agent Hashes
08.30.26stop-that-shit guards coding agents from unrequested hashes, extra files, and scope creep in Codex-style workflows.
VibeGuard Security Linting for AI Code
08.30.26VibeGuard checks AI-generated code for common security bugs; here is the Cursor review boundary worth copying.
AgentCloud Gives Cloud Agents iOS Simulators
08.29.26AgentCloud connects MCP-compatible coding agents to disposable iOS simulators so app fixes can be tested end to end.
Coordination Layer for Coding Agents
08.29.26Twing's coordination layer spots duplicated agent work and design conflicts early, with a safe Cursor workflow to try it.
dmx MCP Server Adds Gated Agent Loops
08.29.26dmx runs configurable gated loops inside agentic IDEs, helping developers bound coding-agent work before it drifts.
grith Supervises Coding Agents on Linux
08.29.26grith intercepts Linux syscalls from AI coding agents and queues risky actions before they reach the kernel.
Rundown Summarizes Hacker News with Codex
08.29.26Rundown summarizes Hacker News posts and comments with local Claude Code or Codex CLI, while keeping links back to evidence.
concord-mcp Lets Coding Agents Talk
08.28.26concord-mcp lets Claude Code, Codex, and Cursor coordinate work through MCP instead of manual relays.
Experiential’s Open Model Gateway for Agents
08.28.26Experiential is an open model gateway for agents, with routing, budgets, and one Codex CLI workflow to test it safely.
Harden.run Beats GPT5.5-xhigh on Agent Guards
08.28.26Harden.run tested SLM and IRM agent guards against hard coding-agent security benchmarks. Here is what mattered.
Open Session’s Open-Source Cloud Agent Orchestrator
08.28.26Open Session is an open-source cloud agent orchestrator, and this article shows when Codex users should try it safely.
Simon Willison on Breaking Claude Code Auto Mode
08.28.26Simon Willison covers a Claude Code Opus 5 auto mode bypass and why sandboxing still matters for coding agents.
Z Brings Minimal Agentic Coding to Terminals
08.28.26Assess a minimal terminal agent through its documented tools, permission boundaries and a small task with independently checkable results.
Code Review Habits for AI Code
08.27.26Learn a Cursor-friendly workflow for reviewing AI-generated code with repo rules, MCP boundaries, and a paste-ready checklist.
TaskShell Bridges AI Coding Agents
08.27.26TaskShell is a Show HN project that lets AI coding agents share task state, with a safe Cursor workflow to test it.
Termux-Dev Brings Agents to Android
08.27.26Termux-Dev brings an Android-first coding agent to Termux and desktop; here is what works, what to test, and what to avoid.
The $40 Open-Source Codex Micro
08.27.26Iluvatar Labs built a $40 open-source Codex Micro clone, showing where hardware shortcuts help Codex workflows.
CoolPlugz Turns Jira Tickets Into PRs
08.26.26CoolPlugz is a claude-plugins-app MCP server that links Jira, GitHub, Slack, and Notion into a PR loop.
Cursor Now Runs on iPad
08.26.26Cursor for iPad brings agents, full PR review, Inbox, and bigger-screen markup into the mobile coding workflow.
Google AI Adds Search Study Tools
08.26.26Google AI’s Search study release shows how bounded learning loops can inform safer Codex and MCP workflows.
Google DeepMind Explains Full-Stack AI
08.26.26Google DeepMind’s five-layer framing shows why AI work now spans models, tools, product behavior, and user trust.
OpenAI Restores Codex’s 5-Hour Limit
08.26.26OpenAI brought back 5-hour Codex and Work limits for Plus users, raising a real debate about agent reliability.
Codex Opens Its Agent Harness
08.25.26OpenAI’s Codex platform release explains the open agent harness and how to test it safely in one repo.
Create a Cursor Plugin You Can Install
08.25.26Rajkumar Samra’s post shows how Cursor plugins package context, rules, skills, agents, hooks, and MCP.
Prism Reviewer Action Splits Code Review
08.25.26Prism Reviewer AI uses LangGraph and LiteLLM to split pull request review across multiple AI reviewer agents.
Sloppie Is a Linux Agentic Coding Environment
08.25.26Sloppie is a Linux development environment that turns coding-agent work into review comments, diffs, and terminals.
Will AI Reliance Collapse Coding Expertise?
08.25.26Lars Faye’s AI reliance essay sparked a developer debate about skill loss, coding agents, and review habits.
Continue Is Archived: What Replaces It
08.24.26Continue is archived, but not useless. Learn what changed and how to test a safer Cursor replacement path.
Hands Lets MCP Click Real Chrome
08.24.26Hands shows a Windows MCP path where Codex can observe, click, and type in a real Chrome profile safely.
Meetless Agent Tracks Coding Agents
08.24.26Meetless shows how an active source of truth can make parallel coding agents easier to inspect before merge.
Simon Willison on Coding Agent Review
08.24.26Simon Willison argues that coding agent review is really about proving changes, not reading every generated line.
Zuse Runs 20 Linear Issues in Worktrees
08.24.26Zuse ran one lead agent across 20 Linear issues in worktrees. Here is why it landed and how to review it safely.
A Week Choosing Codex Over Claude
08.23.26A developer’s week with Codex over Claude turned into a sharper question: which agent makes smaller, faster patches?
Cursor Origin Hosts Code Repos
08.23.26Cursor Origin now hosts repos and pull requests in early beta, with GitHub sync and a one-repo checklist to test safely.
fx Is a Tiny Native Coding Agent
08.23.26fx is a tiny native coding agent from Vercel Labs. Learn why its small shape matters and how to test it safely.
New MCP Roadmap for Agent Integrations
08.23.26The new MCP roadmap explains where agent integrations are headed and how Cursor users can wire them safely.
TechSkills Gives Coding Agents Skill Files
08.23.26TechSkills packages reusable engineering skills for coding agents, and this explains when the idea is useful.
claude-code 2.1.238 Adds Header Helpers
08.22.26claude-code 2.1.238 adds plugin header helpers, readline prompt behavior, runner shutdown flags, and memory fixes.
Codex CLI 0.149.0 Adds Agent Dashboard
08.22.26OpenAI Codex CLI 0.149.0 adds an agent dashboard, queue commands, better doctor checks, and safer session restores.
desktop-vibe-fly Sniffs Out Vibe Code
08.22.26desktop-vibe-fly turns AI-coding marker files into desktop scent; learn why it landed and how to try the idea safely.
Heimdall Adds Trust Verdicts to Agent Memory
08.22.26Heimdall is an open-source knowledge layer that verifies AI coding agent memory hits before an agent acts on them.
Proliferate Is a Self-Hostable AI IDE
08.22.26Proliferate is an open-source AI IDE for parallel coding agents, with a safe Cursor-first way to try it.
Show HN: Frugal Tokens Shows Agent Costs
08.22.26Frugal Tokens explores coding-agent session costs, cache misses, and usage patterns so developers can inspect spend before changing workflows.
Codex on AWS Bedrock 10x Charges
08.21.26A Codex CLI issue on Amazon Bedrock showed expensive cache writes. Learn what happened and the safe checks to run first.
Cursor Cloud Agents Get Longer Leashes
08.21.26Cursor’s August 2026 changelog gives cloud agents subscriptions, /goal, custom modes, and safer subagent VMs to test in one repo.
Huzzah Makes AI Coding Less Chatty
08.21.26Huzzah turns AI coding prompts into structured intent, showing where terse specs help and where agents still need review.
machine0 Gives Coding Agents Persistent Cloud VMs
08.21.26Evaluate machine0 persistent VMs for coding agents with a suspend-and-resume trial, an explicit access boundary, and a check of compute and storage charges.
Speko Launches Voice AI Router
08.21.26Speko routes voice AI stacks across STT, LLM, and TTS choices, with a useful lesson for coding-agent evals.
claude-code 2.1.229 Repairs Remote Sessions
08.16.26claude-code 2.1.229 tightens Remote Control sessions, self-hosted hooks, plugin commands, streaming, and crash handling.
Cursor Cloud Agents Start Faster With Builds
08.16.26Cursor Builds prepare Cloud Agent environments ahead of time. Separate disk setup from running services, check source freshness and test a small task.
Does AI Coding Feel Like Leadership?
08.16.26Allen Bargi’s note sparked a real developer debate about whether agentic coding is coding, management, or something in between.
neal Runs Claude and Codex Together
08.16.26neal shows how Claude and Codex can split planner, coder, and reviewer roles in a local repo workflow.
Remarc: contextual feedback for coding agents
08.16.26Contextual feedback helps an agent locate a visual defect; actual permissions, retention checks and rendered acceptance still govern the workflow.
Waku Is a Native Coding-Agent App
08.16.26Waku is a Rust and GPUI desktop app for coding agents. The useful question is whether native control beats chat.
Agentic AI workshop cost and curriculum, explained plainly
08.15.26An agentic AI workshop cost and curriculum overview: what drives the price, what a real syllabus contains, and which line items are padding.
Agentic AI workshop for beginners with no coding experience
08.15.26An agentic AI workshop for beginners with no coding experience should teach judgement, not syntax. What to expect and what to be careful of.
Agentic AI workshop for developers using LangChain
08.15.26An agentic AI workshop for developers using LangChain should cover evaluation and failure handling, not another chain tutorial. What we teach and skip.
Agentic AI workshop for enterprise architects
08.15.26An agentic AI workshop for enterprise architects covers guardrails, repo standards, and rollout evidence, not prompt tips. Here is the syllabus.
Agentic AI workshop for marketing teams in Austin
08.15.26What an agentic AI workshop for marketing teams in Austin should cover, why the format differs from engineering training, and how to judge a vendor.
Agentic coding is a review problem
08.15.26Agentic coding moves the bottleneck from writing code to reading it. What changes for a team, what breaks first, and how to adapt without losing quality.
Agentic coding workshop for senior developers
08.15.26An agentic coding workshop for senior developers has to earn the room. What it should cover, and why the usual beginner curriculum fails here.
AI code review that engineers stop ignoring
08.15.26How to make AI code review useful: what to let it check, what to stop it commenting on, and the metric that tells you it is working.
AI coding workshops for non-technical managers
08.15.26What AI coding workshops for non-technical managers should cover, what they usually get wrong, and how to judge one before you pay for it.
Best AI coding training for career growth
08.15.26The best AI coding training for career growth is not the one with a certificate. Here is what promotion committees actually respond to.
Best AI coding training for remote teams under $3000
08.15.26Best AI coding training for remote teams under $3000: what a remote format changes, what to demand from a vendor, and how to avoid a dead webinar.
Best practices for running OpenAI Codex with a team
08.15.26Best practices for running OpenAI Codex with an engineering team: scoping, review policy, shared context, and the failures that show up in month two.
Claude Code workshop vs GitHub Copilot training
08.15.26Claude Code workshop vs GitHub Copilot training: the two teach different jobs. Here is which one your team needs and why the answer changes.
Compare AI coding workshops from OpenAI, Google, and Microsoft
08.15.26Compare AI coding workshops from OpenAI, Google, and Microsoft against independent training. What vendor programmes do well and where they stop.
Finding an agentic AI workshop near me for startups
08.15.26Looking for an agentic AI workshop near me for startups? Location matters less than you think. What a small team should demand instead.
How to become an AI coding expert in six months
08.15.26How to become an AI coding expert without collecting certificates: the skills that compound, the ones that expire, and how to tell them apart.
How to learn AI coding with Claude for free
08.15.26A practitioner route to learn AI coding with Claude for free: what to build, what to read, and the habits that separate real skill from prompt luck.
How to run a hHands-on agentic AI workshop with real projects
08.15.26A hHands-on agentic AI workshop with real projects only works if the repo is yours. What to prepare, and how to tell demo theatre from training.
The best AI coding training for Python developers
08.15.26What is the best AI coding training for Python developers? It depends on whether you write services, notebooks, or libraries. Here is how to choose.
What changes in AI software development, and what does not
08.15.26AI software development moves the bottleneck from typing code to reviewing it. Here is what shifts on a real team, and what stays exactly the same.
Which AI coding workshop is best for a budget of $2000
08.15.26Which AI coding workshop is best for a budget of $2000? What that money realistically buys, what it does not, and how to spend it without waste.
A Claude CLI workflow for tasks longer than an hour
08.15.26A Claude CLI workflow built around context, checkpoints, and review, for work that takes longer than a single sitting.
A CLI workflow that survives a real working day
08.15.26A CLI workflow for agentic coding that holds up under interruptions, review, and long tasks, with the parts that break.
A Codex CLI workflow that holds up on a Tuesday
08.15.26The Codex CLI workflow we teach engineering teams: plan in read-only, execute in a worktree, verify with one command, and keep sessions short.
A Codex workflow that survives contact with a real repo
08.15.26The Codex workflow we teach: shape the task, plan, small diff, verify, commit. Plus the three points where teams reliably lose time.
Agent teams in Codex and when the extra agents pay off
08.15.26Running agent teams in Codex helps on tasks that split cleanly and hurts on everything else. Here is how to tell the difference.
agentic-ship Runs Lovable-Style Builds on Your Agent
08.15.26agentic-ship packages app-builder conventions as an open-source harness your existing coding agent can run.
AI code governance without a policy nobody reads
08.15.26AI code governance works when it lives in the repo and the pipeline, not in a wiki page. What to enforce, what to measure, and what to leave alone.
AI coding assistant governance without a 40-page policy
08.15.26AI coding assistant governance that engineers actually follow: the four controls worth having and the paperwork you can drop.
Artifex Gives Agents a Media Graph
08.15.26Artifex is a headless CLI runtime for agent-built media graphs, with practical checks for trying it safely in Codex workflows.
Brainless GitHub panels and the trust they borrow
08.15.26Brainless is a shadcn registry for agent-style UI. Wiring a Brainless GitHub panel to real repo access needs care.
Brainless UI, and why agents work better against it
08.15.26A brainless UI keeps logic out of components. It makes agent-written frontend code reviewable, testable, and much easier to fix.
Choosing between Codex models on real work
08.15.26Codex models differ in speed, cost, and how long they hold a plan. How to pick one per task instead of picking once.
Claude Code hooks that earn their place in a repo
08.15.26Claude Code hooks turn a convention the model might follow into a rule it cannot skip. The four we see teams keep, and the ones they regret.
Claude Code skills, and when they beat a prompt
08.15.26Claude Code skills are folders of instructions loaded on demand. What they are good for, where they rot, and how to write one that survives a team.
Claude Code subagents are a context tool
08.15.26Claude Code subagents are not a team of experts. They keep noise out of your main context, and that framing tells you when to use them.
Claude Code vs GitHub Copilot for engineering teams
08.15.26Claude Code vs GitHub Copilot, compared on how each one actually behaves in a real repo, and which teams should run both.
cliclaw Runs Coding CLIs From Telegram
08.15.26cliclaw is a Telegram-controlled macOS daemon for local coding CLIs, with safety checks that make it worth trying.
code review tools with team governance and policies
08.15.26What to check before buying code review tools with team-based governance and policies, and which controls actually change behaviour.
Codex agent teams and the review bottleneck
08.15.26Codex agent teams sound like parallel throughput. In practice the limit is human review. How to structure the work so the parallelism actually pays.
Codex agents, and when they stop helping
08.15.26How Codex agents behave on real repositories, what AGENTS.md should contain, and the point at which adding more agents makes the work slower.
Codex AI training that changes what a team merges
08.15.26What Codex AI training should cover for a working engineering team, how to structure it, and the measures that show whether it worked.
Codex cls, and why clearing context is a real skill
08.15.26People search codex cls looking for a clear command. The useful answer is about when to clear context in a Codex session, and why it matters.
Codex delegate patterns that survive a real codebase
08.15.26How to Codex delegate a task properly: scoping the handoff, writing the acceptance check first, and picking the sandbox mode that matches the risk.
Codex GitHub reviews without the noise
08.15.26Codex GitHub integration puts an agent inside your pull requests. What to hand it, what to keep on your laptop, and how to stop the comment flood.
Codex MCP servers, and when they are worth the tokens
08.15.26A practical guide to Codex MCP setup: what to connect, what to leave out, and the failure modes that cost teams a whole context window.
Codex opt out of training, and what to verify
08.15.26Separate Codex training data controls from file access, credential exposure and retention, then verify the account settings and actual access boundary.
Codex subagents and skills, honestly compared
08.15.26What people mean by Codex subagents and skills, what the tool actually gives you today, and how to get the same effect with prompts and scoped runs.
Codex subagents: what they buy you
08.15.26Codex subagents split work into fresh contexts. Here is when that helps, when it costs you, and how to structure a delegation that returns something usable.
Codex training that survives contact with your repo
08.15.26A useful Codex training exercise ends with a bounded change and evidence that another engineer can review without replaying the session.
Codex vs Claude Code, What Teams Running Both Actually Say
08.15.26Codex vs Claude Code without benchmark hype. Where each tool wins day to day, and why many teams end up running both.
Cursor agents and the work they are actually good at
08.15.26When to hand a task to Cursor agents, how to scope it so review stays cheap, and the failure modes that waste an afternoon.
Cursor AI, what it is good at and where it breaks
08.15.26One is inline: predictive completion and multi-line edits that appear as you type, plus a quick inline edit on a selection.
Cursor MCP servers that are worth the context
08.15.26Cursor MCP connects the editor to your other systems. Which servers pay for themselves, which eat your context window, and how to tell them apart.
Cursor MCP support and the Model Context Protocol in practice
08.15.26Cursor MCP support lets the agent call your own tools through the Model Context Protocol. How to configure it, and the failure modes teams hit first.
Cursor skills your team should build first
08.15.26The Cursor skills that separate fast teams from frustrated ones, and how to turn a working prompt into a rule the whole repo uses.
Designing a Codex agent team for one feature
08.15.26A worked example of splitting a single feature across a Codex agent team, including the handoff format and the parts you should keep for yourself.
Four Codex workflows that hold up under load
08.15.26Most Codex workflows fall apart on real code. These four survive, and each one has a clear failure mode you should know.
Getting a Claude Code review worth reading
08.15.26A Claude Code review is only as good as the scope you give it. How to set that scope, where it beats a human reviewer, and where it wastes your time.
Governance for AI-generated code that engineers accept
08.15.26A practical policy for governance of AI-generated code: what to write down, what to automate, and which controls quietly get ignored.
How .cursor/rules works, past the Cursor rules documentation
08.15.26The Cursor rules documentation explains the .cursor/rules file format. This explains which of the four attachment modes to pick and why it matters.
How agents in Codex actually decide what to do
08.15.26A plain explanation of how agents in Codex read a repo, pick tools, and recover from failure, plus the three habits that make them useful.
How Codex works, and where it stops working
08.15.26A practitioner view of how Codex work gets done: the read-plan-edit-verify loop, what the sandbox allows, and the tasks where it reliably falls over.
How MCP servers actually behave in a real repo
08.15.26MCP servers give a coding agent tools beyond your files. What they cost you in context, and how to keep the list short.
How to brief a Cursor agent properly
08.15.26A Cursor agent edits across files and runs commands on its own. The briefing habits that decide whether that saves an hour or costs you two.
How to merge AI code safely on a busy team
08.15.26A checklist to merge AI code safely: what to read line by line, what to automate, and the bug classes that slip through review.
How to pick between Codex courses for a working team
08.15.26Most Codex courses teach the interface. Here is how to tell which ones change how a team actually ships, and what to ask before you buy.
Looking for Codex CLI? Start with a bounded task
08.15.26If you meant Codex CLI, start with a small repository task and inspect the actual edits, permissions and verification output.
Making agent-readable media assets in a real codebase
08.15.26Coding agents cannot see your PNGs. Agent-readable media assets fix that with naming, manifests, and text sidecars your build already understands.
Maximizing Claude Code Sessions
08.15.26Use Claude Code context and usage reports to review one bounded task, preserve useful handoffs, and distinguish token estimates from subscription billing.
Mole Puts a Budget on Terminal Research
08.15.26Mole is a terminal research agent that enforces spend, checks quotes, and keeps local data inside a clearer boundary.
Offsite training for engineering teams, done properly
08.15.26How to run offsite training for an engineering team so the habits survive the trip home, and when to stay onsite instead.
Onsite programming sessions that survive contact with agents
08.15.26What an onsite programming day with AI coding agents should look like, who it works for, and where the format falls down.
OpenKnowledge vs Obsidian for a team Claude can read
08.15.26Comparing OpenKnowledge vs Obsidian for engineering notes, judged on one thing: can a coding agent read the knowledge without a plugin.
Reading the Anthropic Claude Code hooks documentation properly
08.15.26A practitioner's guide to the Anthropic Claude Code hooks documentation: which events matter, how exit codes work, and the mistakes teams make first.
Rogier Muller, AI coding agent trainer in Amsterdam
08.15.26Who Rogier Muller is, what he teaches engineering teams about AI coding agents, and how the training sessions are actually run.
Rolling out Codex to teams without losing the review gate
08.15.26How Codex teams share configuration, keep review honest, and avoid the two rollout patterns that quietly waste a quarter.
Running a Claude Code security review that finds real bugs
08.15.26A Claude Code security review is good at data-flow bugs and bad at threat modelling. How to scope it so the findings are worth reading.
Running a Codex team of agents without a mess
08.15.26When a Codex team of agents beats a single session, how to keep parallel work from colliding, and the point where it stops paying off.
Running an agent team in Codex without chaos
08.15.26How to run an agent team on Codex: splitting work so parallel agents do not collide, giving each one a verifiable exit, and knowing when one agent is better.
Running Codex across a team without chaos
08.15.26A Codex team needs shared config, shared review habits, and one owner. Here is the setup that survives contact with a real backlog.
Running Codex team agents without stepping on each other
08.15.26Codex team agents work when the shared rules live in the repo and the work is split by boundary. Here is the setup we teach and the failures we see.
Searched curs or and meant Cursor? Start here
08.15.26If you typed curs or into a search box you probably meant Cursor. Here is what it is and how teams use it.
self-bench Turns Private PRs Into Evals
08.15.26self-bench turns completed private PRs into coding-agent evals, with a safer way to measure agents on real repo work.
Setting a codex SLI your team will actually watch
08.15.26A codex SLI turns vague agent adoption talk into a number. Which indicators are worth tracking, which are vanity, and how to instrument them.
Setting up Codex CLI MCP without regret
08.15.26Codex CLI MCP support means the agent can call tools provided by an external process: a database client, a docs index, an issue tracker, a browser.
Setting up one Codex team agent everyone shares
08.15.26A shared Codex team agent gives you one identity, one config, and an audit trail. Here is how to set it up and what it should never be allowed to do.
Standardising Codex agents across an engineering team
08.15.26Rolling out Codex agents to a team: shared repo context, an agreed review policy, sane sandbox defaults, and the metrics that tell you it is working.
The best MCP servers for Claude Code are the few you keep
08.15.26The best MCP servers for Claude Code are the three or four that reach data the agent cannot get from your files. Here is how to choose.
The Claude Code CLI workflow after the novelty wears off
08.15.26A Claude Code CLI workflow built around plan mode, short sessions, slash commands you wrote yourself, and hooks that enforce rules nobody remembers.
The Cursor rules that are worth writing
08.15.26Most Cursor rules files are too long to work. Here is what to keep, what to delete, and how to tell whether a rule changed the output.
What a Codex bootcamp should cover, and what to skip
08.15.26What a codex bootcamp needs to teach engineers in two days, which exercises work, and the parts that waste everyone's time.
What a Codex course should teach your engineers
08.15.26How to judge a Codex course before you buy one, what a good syllabus contains, and when a team is better off skipping training entirely.
What codex team mode actually means for a shared repo
08.15.26Codex team mode is mostly convention, not a setting. What shared config, review rules, and task ownership look like when a whole team runs Codex.
What good MCP technical training actually covers
08.15.26An MCP exercise should test tool selection, authorization, timeout handling and credential ownership before an integration reaches real records.
What the claude mcp command actually configures
08.15.26The claude mcp command adds, lists, and removes MCP servers. The part that matters is scope, and picking the wrong one breaks it for your team.
When Claude Code MCP servers help, and when they hurt
08.15.26Claude Code MCP servers give the agent real data instead of guesses. They also cost context on every request. How to decide which ones to keep.
When Codex sub agents help and when they hurt
08.15.26Codex sub agents split a task across separate contexts. Useful for parallel search, risky for anything that needs shared state.
Working with Opus 5 in Claude Code without burning budget
08.15.26When to reach for Opus 5 in Claude Code, when a smaller model is the better call, and the habits that decide whether the bigger model actually pays off.
Your CLAUDE.md is probably too long
08.15.26CLAUDE.md supplies persistent repository instructions. Why long ones stop working, what belongs in it, and a rewrite that takes twenty minutes.
/show-me Makes Coding Agents Draw
08.14.26HumanLayer's /show-me turns coding-agent explanations into compact visuals so developers can review shape, flow, and risk faster.
claude-code 2.1.232 Enables Agent Forks
08.14.26claude-code v2.1.232 changes subagent defaults, cross-session messaging, GitLab token redaction, and what to test first.
Codex in ChatGPT Linux Preview
08.14.26OpenAI’s Linux preview brings Codex into the ChatGPT desktop app, with a safe first repo workflow to try.
Hearth Turns Family Notes Into Apps
08.14.26Hearth shows how a shared household workspace can let an agent use family context and build small apps safely.
Kery Shows PR Features Working
08.14.26Kery tests web-app pull requests in a browser and leaves visual evidence so agent-written UI changes are easier to trust.
Read it easy Is a Read-Only Code Editor
08.14.26Read it easy is a read-only desktop code editor built for source reading. Here is why its Go to Definition idea matters.
brave-devtools-mcp Connects Brave to Agents
08.13.26brave-devtools-mcp connects Brave DevTools to coding agents, with a safe Cursor workflow and permission boundary.
Claude Code 2.1.229 Stabilizes Remote Control
08.13.26Claude Code 2.1.229 steadies Remote Control, MCP OAuth, streaming, hooks, and plugin command sources.
Discovered Materials Uses Agents for Chip Heat
08.13.26Discovered Materials (YC P26) uses AI agents to search for semiconductor materials, and Cursor users get a review lesson.
FEDERaiDE Routes Agents in Your Terminal
08.13.26FEDERaiDE is a terminal multi-agent harness with P2P routes, memories, and an IDE. Here is when to try it safely.
Simon Willison Ships alchemy-utils Alpha
08.13.26Simon Willison’s alchemy-utils 0.1a0 turns an AI-built database spike into a small alpha worth studying.
tmux-agent-switcher Watches Coding Agents
08.13.26tmux-agent-switcher tracks AI coding agents in tmux so you can spot blocked Codex or Claude panes quickly.
Best Programming Language for Coding Agents?
08.12.26Dan Luu’s token-efficiency post asks whether language choice matters when coding agents read and write code.
Claude Code User-Agent Email Leak Report
08.12.26A reported Claude Code curl User-Agent email leak shows why agent-run network commands need explicit header review.
Hoplite YC S26 Brings Coding Agents to the Cloud
08.12.26Hoplite (YC S26) moves coding agent setup into cloud sandboxes. Here is what it does well and what to check before you trust it.
keen-code: A Go Coding Agent
08.12.26keen-code is a Go terminal coding agent that shows what a small, opinionated agent harness can teach Codex users.
Parley Lets Coding Agents Talk
08.12.26Parley connects coding agents over MCP so they can ask questions, hand off work, and claim files without using a human relay.
Airship Visually Edits Your Running App
08.11.26Airship is a visual editor for running apps that lets Codex, Claude Code, or OpenCode change real source.
Ante Ships Offline Coding in One Binary
08.11.26Evaluate Ante in a disposable repository: verify the local model, runtime access, licensing and test evidence before accepting an agent-generated patch.
Claude Code Makes Auto Mode Default
08.11.26Claude Code now defaults to Auto mode. Here is what changed, why it landed, and the safest first repo check.
Oqoqo Measures Agents on Real Tasks
08.11.26Oqoqo shows how realistic agent evals can test product surfaces, MCP, CLIs, and Cursor review workflows.
Try Benzi Tests Code Maps Against Claude Code
08.09.26Try Benzi maps codebases for agents, checks writes with static analysis, and gives Claude Code users a measurement lesson.
Vibsync Shares Memory Across Cursor, Claude, Codex
08.09.26Vibsync gives Cursor, Claude Code, and Codex one MCP memory; this shows where it helps and where to set boundaries.
Managing AI Coding Costs at Scale, Is It Worth It?
08.08.26Databricks published a plan for managing AI coding costs at scale. Here is how to test the spend, review, and metering claims.
Mirafold Gives Terminal Agents a Browser UI
08.08.26Mirafold wraps terminal coding agents in a browser UI, showing when the extra surface helps and when to stay in the shell.
Mirafold Wraps Coding Agents in Browser UI
08.08.26Mirafold puts Codex, Claude Code, and Gemini CLI output in a browser UI without changing the agent underneath.
aident-skill Connects Codex and Claude Code to Apps
08.07.26Aident Loadout gives Claude Code and Codex app actions; learn what the GitHub project does and when to try it safely.
mcp-use v2 Gets a Stateless Rebuild
08.07.26mcp-use v2 was rebuilt for stateless MCP, with faster launches, smaller installs, and sharper integration boundaries.
mcp-use v2 Goes Stateless
08.07.26mcp-use v2 was rebuilt for stateless MCP. Learn what changed, why it matters, and where to set review boundaries.
mcp-use v2 Rebuilds for Stateless MCP
08.07.26mcp-use v2 rewrites a TypeScript MCP framework for stateless servers, with faster launches and safer agent boundaries.
adlc-team-skills Teaches Agents Team Rules
08.06.26adlc-team-skills turns team coding conventions into reviewable agent skills for Claude Code, Codex, and Cursor workflows.
claude-code 2.1.223 Tightens Marketplace Permissions
08.06.26v2.1.223 adds marketplace wildcards, model fallback warnings, teleport hints, and security fixes for Claude Code.
Frontrun Tests Python Races
08.06.26Frontrun explores Python thread, async, and process interleavings so race conditions become reproducible.
Armature Shows MCP Session Analytics
08.04.26Armature reconstructs MCP agent sessions so developers can see use cases, failures, and review evidence.
book-skills Turns Work Books Into Claude Skills
08.04.26book-skills packages management and programming books as Claude Code skills, with a safe way to test when that helps.
Cursor Adds Google Workspace Plugins
08.04.26Cursor’s Google Workspace Plugins connect agents to Gmail, Drive, Calendar, Docs, and Sheets from the editor.
Rudder Measures Your Input on AI Code
08.04.26Rudder turns prompt history into tests so Codex users can see whether generated code reflects their intent.
claude-code-meter Tracks Claude Code Usage Pace
08.03.26claude-code-meter tracks Claude Code and Codex usage pace, so you can see a burn rate problem before you hit a limit reset.
MicroCodex: Reviewing a Small Coding Agent
08.03.26Evaluate MicroCodex on one small repository task. Record its release, runtime dependencies, changed files and test evidence before accepting a patch.
Sprocket Lets an Agent Buy Parts
08.03.26Sprocket is a Show HN agent for hardware, software, and web purchases. Here is what mattered and how to test it safely.
Claude MIDI Twister Shows Agent State
08.01.26A MIDI controller can display coding-agent state, but its usefulness depends on the events captured and the privacy of the local bridge.
Ski Runs Voice Coding Locally
08.01.26Test Ski with transcript approval enabled, exact identifier checks, and a small patch. Separate local speech processing from the connected coding agent.
skill-language-server Refactors Agent Skills
08.01.26skill-language-server adds LSP tooling for agent skills, so Cursor users can review renames and references safely.
Yamlet Makes Agent Specs Smaller
08.01.26Yamlet is an open-source toolkit for writing tiny YAML specs that Claude Code can challenge, check, and turn into test files.
agent-manager Puts AI Agents in tmux
07.31.26agent-manager keeps coding agents in tmux sessions and shows where review, prompts, and stalled panes need attention.
codex-security Tests the Repo Boundary
07.29.26OpenAI’s codex-security scans code for vulnerabilities, but the real lesson is where to draw the repo boundary before agents fix code.
Dn Turns Issues Into Agent Plans
07.29.26dn turns GitHub issues and markdown specs into durable plans that agents can execute, review, and resume.
Telnyx Kimi K3, Is It Fast Enough to Use?
07.29.26Telnyx now serves Moonshot AI Kimi K3 through its inference API. Here is how to test latency before you trust it for coding work.
verified-3d-mesh-intersection Verifies 3D CSG
07.29.26verified-3d-mesh-intersection proves a 3D CSG mesh kernel in Lean and shows a sharper review pattern for AI code.
verified-3d-mesh-intersection Verifies 3D Mesh CSG
07.29.26A Lean 4 mesh-intersection project shows why reviewers can trust a small spec, a checker, and sharp boundary tests.
claude-code 2.1.218 Backgrounds /code-review
07.28.26claude-code 2.1.218 moves /code-review into a background subagent and tightens MCP, terminal, and accessibility edges.
Decispher Adds Grok CLI Support
07.28.26Decispher now connects Grok CLI sessions to architectural decisions and reviewable agent traces.
FeyNoBg Opens a Background Removal Library
07.28.26FeyNoBg is an open background-removal model and Python library, with a useful lesson for reviewing AI-made code.
FeyNoBg Removes Backgrounds in Public
07.28.26Feyn released FeyNoBg and NoBg; here is what developers should test before wiring it into agent-written code.
FeyNoBg Ships Background Removal Library
07.28.26Feyn released FeyNoBg and NoBg, showing how to test an AI image tool before wiring it into real repos.
Boffin Adds Per-Edit Agent Guardrails
07.27.26Boffin routes file-specific constraints to coding agents before each edit. Compare it with repo rules and use a verification checklist for risky changes.
Claude Code 2.1.218 Backgrounds Reviews
07.27.26Claude Code 2.1.218 moves reviews into a background subagent and fixes sharp edges in terminals, paths, and accessibility.
coding-agent-skill-library Serves Skills via MCP
07.27.26coding-agent-skill-library exposes reusable agent skills through MCP, with a safe Cursor workflow and rule-file boundary.
Termic Runs CLI Coding Agents
07.27.26A desktop shell for terminal agents should make session boundaries, changed files and verification evidence easier to inspect.
Argus Manages VS Code Worktree Agents
07.26.26Argus is a VS Code extension that manages parallel Codex and Claude Code worktree sessions in one window.
claude-code 2.1.219 Surfaces MCP Failures
07.26.26claude-code 2.1.219 makes MCP failures, directory changes, and sandbox network rules easier to test before upgrading.
Cursor Router Powers Auto Mode
07.26.26Cursor Router powers Auto mode in Cursor, routing each request by task complexity and cost with admin controls.
AgentCost Tracks Local Agent Spend
07.25.26What local agent-cost logs can establish, where provider billing differs, and how to compare the two before changing a team’s workflow.
Claude Code 2.1.219 Adds Opus 5
07.25.26Claude Code 2.1.219 adds Opus 5, stricter sandbox networking, better MCP errors, and a new DirectoryAdded hook.
Show HN: WhipDesk Controls Dev Machines From Phones
07.25.26WhipDesk is an open-source phone remote desktop for AI coding agents, with a safe Codex handoff checklist.
Echo Claims Fable-Level Results for Less
07.24.26Echo claims Fable-level output from routed open models; here is what to trust, test, and review in Cursor.
Echo Tests Cheaper Open-Weight Routing
07.24.26Echo routes open-weight models per task, raising useful questions about evals, cost, and review habits for AI code.
OneCLI Keeps Secrets Out of Agents
07.24.26OneCLI shows a cleaner boundary for AI coding agents: fake keys in the session, real secrets behind a gateway.
Bento Puts Slides in One HTML File
07.23.26Bento puts editing, viewing, slide data, animation, and collaboration into one HTML file you can inspect.
Code Review Habits for Coding Agents
07.23.26Learn a practical team workflow for reviewing AI-generated code from Codex, Cursor Agent, MCP boundaries, and PR checks.
Review AI-Generated Code in Cursor
07.23.26Learn a Cursor-first workflow for reviewing AI-generated code with rules, PR checks, and a paste-ready checklist.
Review Habits for AI-Generated Code
07.23.26Teams learn a simple review workflow for AI-generated code, with a checklist, comparison table, and Claude Code habits.
serve-avd Streams Android Emulators to Browsers
07.23.26serve-avd puts an Android Emulator in a browser so Cursor agents can inspect and test mobile UI with a human nearby.
Yorishiro Gives Coding Agents a Body
07.23.26Yorishiro is a macOS terminal that gives coding agents a body, and shows why visible agent state matters.
claude-code 2.1.216 Loosens Sandboxes
07.22.26claude-code 2.1.216 adds a filesystem sandbox escape hatch and fixes stalls, resumes, web answers, and worktree isolation.
Claudexor Routes Coding-Agent Quotas
07.22.26Assess Claudexor with a disposable repository: check quota visibility, credential scope, patch review and evidence before routing production work.
CodeAlmanac Saves What Agent Chats Forget
07.22.26CodeAlmanac turns coding-agent chats into a local wiki, so agents can reuse decisions, flows, and gotchas.
CodeAlmanac Turns Agent Chats Into Wikis
07.22.26CodeAlmanac turns agent conversations into a local repo wiki. This explains the project, the doubts, and a safe first test.
CodeAlmanac Writes Wikis From Agent Chats
07.22.26CodeAlmanac turns Claude Code and Codex chats into a local wiki. Learn where it fits, what to watch, and a safe MCP path.
joydex Turns a Flight Throttle Into Codex Controls
07.22.26Joydex maps flight-sim hardware to OpenAI Codex controls and shows when a physical agent interface is worth trying.
Ask HN: Shipping Apps Without Seeing Code
07.21.26A July 2026 Ask HN asked whether people ship production apps without seeing code, and why Cursor review still matters.
Claude Code 2.1.215 Makes Reviews Explicit
07.21.26Claude Code 2.1.215 stops auto-running review skills. Here is when to call /verify and /code-review yourself instead.
claw-coder Runs an Autonomous Local Agent
07.21.26Gabriel Blessed’s claw-coder post shows why local agents need privacy controls and machine-safety boundaries.
claw-coder Runs Coding Agents Locally
07.21.26Gabriel Blessed’s claw-coder post shows why local agent autonomy is really a runtime safety problem.
claw-coder Takes Autonomous Coding Local
07.21.26Gabriel Blessed’s claw-coder pitch makes local agent safety concrete, with a safe-first way to try it.
Cursor and Codex Hit by Sandbox Escapes
07.21.26Sandbox escapes in Cursor, Codex, Gemini CLI, and Antigravity show why agent review needs a tighter repo boundary.
OpenAI Cuts Codex Context Window
07.21.26Context metadata changes are a reason to check task boundaries and verification practices, not to infer that a larger window proves better work.
PocketVeto Is a Bluetooth Agent Remote
07.21.26PocketVeto gates risky AI coding actions over Bluetooth and shows what Cursor users can borrow from the idea.
Agent-talk Gives Coding Agents a Backchannel
07.18.26agent-talk adds a message backchannel between agent sessions; here is what changed, what is rough, and how to test it safely.
claude-code 2.1.212 Splits Forks and Subtasks
07.18.26Claude Code v2.1.212 changes /fork, /subtask, agent limits, and MCP backgrounding. Here is what to test.
Hermes Agent Runs Kimi K3 in a Sandbox
07.18.26Hermes Agent’s Show HN demo runs Kimi K3 in Sanbox, with a safe Codex CLI review loop before patches land.
Claude Code 2.1.212 Tames Background Sessions
07.17.26Claude Code 2.1.212 changes /fork, MCP backgrounding, and safety limits so review sessions stay usable.
Codex Micro Makes Agent Control Physical
07.17.26Codex Micro is a compact OpenAI and Work Louder device for steering coding agents with explicit, physical controls.
Brainless: Claude Code-Style Shadcn Components
07.16.26Use Brainless shadcn components for agent-style interfaces, with visible GitHub MCP permissions, action previews, and clear read-only versus write access.
deja-vu Shares Agent Memory Over SSH
07.16.26deja-vu turns local coding-agent logs into searchable, shareable memory for Claude Code, Codex, and opencode.
Codex Encrypts Sub-Agent Prompts
07.15.26Codex encrypted sub-agent messages, and one GitHub issue shows why readable agent trails still matter.
Juggler Turns Coding Agents Into a Workbench
07.15.26Inspect Juggler sessions against a real diff and test output. Check remote-access settings and keep a short handoff that another reviewer can verify.
Microsoft’s Claude Code and Copilot CLI Study
07.15.26Microsoft’s CLI study found more merged PRs, but the useful lesson is how to measure agent work without trusting PR count alone.
claude-meseeks Adds Meeseeks Waiting Alerts
07.14.26Test claude-meseeks audio alerts when Claude Code needs input. Compare it with a general notifier and check terminal permissions in a small repository.
clawk Puts Coding Agents in VMs
07.14.26Clawk runs coding agents in disposable Linux VMs. Examine mounted-repo risks, network allow-lists, and the boundaries a VM adds around local commands.
Evaluate Prompt Quality in Claude Code and Codex
07.14.26Learn how a prompt-quality measurement post turns Claude Code and Codex prompts into repeatable engineering checks.
Sanbox Gives AI Agents MicroVM Sandboxes
07.13.26Sanbox runs AI agents in isolated, resumable sandboxes with filesystem state, run events, and Codex review use cases.
Terence Tao Builds Apps With Coding Agents
07.13.26Use Tao’s mathematical app experiments to design a bounded coding task: record assumptions, check known cases, and review the rendered interaction.
Why Claude Code Sends 33k Tokens First
07.13.26Systima measured Claude Code and OpenCode token overhead; here is the debate and a small repo experiment to run yourself.
claude-code 2.1.205 Fixes Workflow Edges
07.12.26claude-code v2.1.205 fixes transcript safety, schema handling, background agents, MCP import, and worktree cleanup.
searxng-ai-kit Turns SearXNG into an MCP Server
07.12.26Evaluate searxng-ai-kit on a public dependency question. Check query privacy, primary sources, exposed tools and local tests before accepting a patch.
9lives Heals Playwright Tests Safely
07.11.269lives repairs broken Playwright selectors, shows a diff, and gives Codex users a safer test-healing pattern.
Abralo Puts Claude Agents in One Window
07.11.26Abralo is a free window for several Claude Code agents, useful when terminal splits hide what needs review.
local-motion Runs Cursor Agents Locally
07.11.26local-motion helps Cursor users run a local coding model on macOS and avoid the usual memory and setup traps.
claude-code 2.1.206 Smooths Repo Friction
07.10.26claude-code 2.1.206 fixes small workflow surprises in /cd, /doctor, git pushes, worktrees, MCP, and login.
Databricks Tests Agents on Millions of Lines
07.10.26Databricks benchmarked coding agents on a huge repo. Learn why task cost, harness design, and repo shape matter.
Kastra Enforces Policies for Cursor Agents
07.10.26Evaluate Kastra authorization for Cursor tool calls with a disposable test service, denial evidence, and checks for alternate execution paths.
Ask HN: What Agent Sandboxes Are Missing
07.09.26Ask HN’s agent sandbox thread shows why Codex work needs clear MCP edges, network controls, and review receipts.
Claude Code 2.1.205 Hardens Agent State
07.09.26Claude Code 2.1.205 fixes session safety, schema output, background agents, MCP imports, and Windows edge cases.
Moo Versions Machines for Agent Branches
07.09.26Moo gives each agent branch its own Linux machine state, so file changes and runtime services stop colliding.
Moo Versions the Whole Machine
07.09.26Moo isolates each branch or agent attempt in a Linux microVM, making parallel AI coding less collision-prone.
ultralytics-mcp Connects YOLO to Cursor
07.09.26ultralytics-mcp lets agents control YOLO datasets and training through MCP, with a safe Cursor setup boundary.
claude-code 2.1.202 Tunes Dynamic Workflows
07.08.26Claude Code 2.1.202 adds dynamic workflow sizing and telemetry fields, plus fixes that make remote sessions less brittle.
Local MCP Connects Mac Apps to Agents
07.08.26Local MCP connects AI assistants to Mac-only context, and the safe first test is a narrow, read-only MCP boundary.
Rowboat vs Claude Desktop for AI Work
07.08.26Rowboat turns the desktop AI assistant into local work surfaces. Here is where it beats Claude Desktop, and where it does not.
Rowboat vs Claude Desktop: Local Work Surfaces
07.08.26Rowboat brings a local-first workbench to Claude-style desktop AI, and shows where Codex CLI still fits best.
Claude Code 2.1.202 Adds Workflow Controls
07.07.26Claude Code 2.1.202 adds workflow size guidance, OTel attributes, and fixes worth testing in one real repo.
Clean Repos Make Coding Agents Cheaper
07.07.26A May 2026 arXiv study found clean code did not raise pass rates, but it made coding agents cheaper to run and easier to review.
Code Cleanliness Changes Coding Agent Footprints
07.07.26Review a code-cleanliness study through agent cost and navigation, then compare one repository change using token use, touched files and test results.
Whyline: Company Memory for Cursor
07.07.26Whyline connects company decisions to Cursor through MCP so agents can answer “why” with receipts, not guesses.
Ask HN: Banned From Claude Code
07.06.26A reported Claude Code account ban raises questions about vendor dependence, portable repo workflows, fallback tools, and narrowly scoped MCP permissions.
TikZ Editor Makes LaTeX Figures Less Painful
07.06.26Try TikZ Editor on one existing LaTeX figure. Check source preservation, macros, the rendered document and the final diff before keeping a visual edit.
Ask HN’s LLM Coding Flow Debate
07.05.26Why a Hacker News debate about LLM coding flow matters for Codex CLI, MCP, specs, and verification loops.
Dan Luu on Agentic Coding: Keep the Loop Bounded
07.05.26Use a small debugging exercise inspired by Dan Luu’s agentic coding notes to define scope, retries, verification and the evidence a reviewer needs.
open-knowledge vs Obsidian and Notion
07.05.26open-knowledge is an open-source, AI-first alternative to Obsidian and Notion. This piece compares them and adds a Claude Code review workflow.
Scopewalker MCP Measures Code Complexity
07.05.26scopewalker-mcp gives Cursor agents local, read-only complexity metrics before they edit your repo.
Manufact YC S25 Turns MCP Servers into Hosted Apps
07.03.26Manufact (YC S25) turns MCP servers into hosted apps. Here is the safe Cursor boundary for connecting agents to real systems.
OpenAI Codex Sensitive Files Issue
07.03.26Codex Workshop research on issue #2847, agent ignore files, and safer Codex CLI workflows around secrets.
Router Picks Models Inside Cursor
07.03.26A story-first look at workweave/router, smart model routing, and the Cursor skill rubric worth keeping.
Claude Code Rollout Drill
07.02.26A team rollout plan for turning Claude Code changelog updates into safer conventions, permissions, and review habits.
Claude Workflow Training for Teams
07.02.26A Claude Code team rollout plan for shared agent workflows, training loops, review receipts, and safer autonomy.
Codex Agent Rules for Real Repos
07.02.26Set up Codex agents with scoped AGENTS.md rules, safe MCP access, verification commands, and reviewable handoffs.
Codex Team Rollouts for Real Repos
07.02.26A Codex-ready team rollout plan for shared agent workflows, MCP boundaries, and safer review habits.
Cursor Guardrails for Real Teams
07.02.26A practical guide to Cursor team workflows with rules, skills, MCP, subagents, and a rollout checklist.
Cursor Shared Workflows for Teams
07.02.26A practical Cursor-first rollout plan for agentic coding training, MCP boundaries, repo rules, and safer reviews.
Bound Claude Code Agent Teams
07.01.26A team convention for running Claude Code subagents with clear scopes, MCP limits, and review receipts.
Codex CLI MCP Setup
07.01.26Connect Codex CLI to MCP with scoped AGENTS.md rules, safe permissions, and a verification loop your team can review.
Cursor Rules vs AGENTS.md, a Team Playbook that Scales
07.01.26How to split Cursor .mdc rules from AGENTS.md, so engineering teams get durable rules, skill handoffs, and reviews that scale.
Safer Codex Rollouts for Teams
07.01.26A practical Codex rollout for team AI coding governance with MCP boundaries, AGENTS.md rules, and review guardrails.
Train Safer AI Coding Habits
07.01.26Use a bounded rollout, review receipts, and tool limits to train coding agents safely across a team.
Adopt Claude Code as a Team
06.30.26A practical Claude Code rollout plan for shared CLAUDE.md rules, MCP access, hooks, skills, and review habits.
Codex Remote in Real Repos
06.30.26Use Codex Remote with GitHub, AGENTS.md, MCP boundaries, and reviewable diffs in Codex CLI workflows.
Safe AI Coding Team Rollouts
06.30.26A practical rollout plan for training engineering teams to use Codex and coding agents with guardrails.
Teach AI Coding With Guardrails
06.30.26A practical team rollout plan for Cursor users: rules, MCP boundaries, workshops, and review guardrails for agentic coding.
Train Teams on Coding Agents Safely
06.30.26A practical rollout plan for Claude Code teams: rules, MCP boundaries, skills, hooks, and review gates for safer agentic coding.
Use Cursor Agents With Guardrails
06.30.26A practical Cursor team workflow for agents, rules, skills, AGENTS.md, MCP, and reviewable background-agent work.
A Practical AI Review Workflow
06.29.26A Codex-first convention for aligning AI code assistants, MCP boundaries, and review guardrails across a team.
A Team Rubric for Cursor Skills
06.29.26A practical Cursor convention for accepting skills, subagents, and rules before they become team defaults.
Codex Remote for GitHub Workflows
06.29.26A practical Codex Remote GA guide for teams using GitHub, AGENTS.md, MCP, and Codex CLI verification loops.
Review Rules for AI Coding Agents
06.29.26A practical Claude Code convention for aligning teams on AI-assisted code review, MCP boundaries, and review guardrails.
Run a Claude Code Team Workshop
06.29.26Run a Claude Code team workshop in a real repository: set CLAUDE.md conventions, scope MCP access, add hooks, and finish with a reviewed pull request.
A Safer Agent Review Convention
06.28.26A practical Claude Code convention for safer agent-assisted reviews, MCP boundaries, and engineering team adoption.
Codex Review Guardrails That Stick
06.28.26A practical Codex convention for safer agent-assisted reviews, MCP boundaries, and team-owned AGENTS.md checks.
Learn Claude Code as a Team
06.28.26A practical Claude Code team rollout plan with CLAUDE.md, hooks, MCP, skills, and review conventions.
Roll Out Cursor Team Training
06.28.26Plan a team Cursor rollout around repository instructions, focused agent tasks, scoped tools and a workshop exercise that tests the review process.
Run Codex as an Engineering Team
06.28.26Team workflow for Codex CLI, AGENTS.md, verification loops, and reviewable diffs after Codex Remote GA.
Give Coding Agents Team Rules
06.27.26A practical team convention for orchestrating Codex and other agents with clear roles, MCP limits, and review rules.
Govern Coding Agents as a Team in Cursor
06.27.26A practical governance framework for coding agents in Cursor, with agent roles, repo rules, MCP boundaries, and a human review gate.
Orchestrate Coding Agents Safely
06.27.26A practical team convention for orchestrating Claude Code agents with scoped rules, MCP limits, and review gates.
Train Cursor Agents as a Team
06.27.26A practical Cursor team rollout plan for rules, skills, subagents, and reviewable AI coding workflows.
Wire Codex to MCP Safely
06.27.26A practical Codex CLI workflow for MCP servers, AGENTS.md rules, Remote runs, and reviewable verification.
Choose Code Review Agents by Receipts
06.26.26Review AI-assisted pull requests with scoped Cursor tasks, explicit MCP permissions and a receipt that records findings, tests and the human decision.
Claude Code Hook Conventions
06.26.26Keep team conventions close to the repository and verify that hook behavior, connector permissions and review evidence match the written policy.
Cursor Agents Need Team Conventions
06.26.26A practical Cursor team convention for agents, rules, skills, AGENTS.md, and safer reviewable workflows.
Pick Code Review Agents Safely
06.26.26A Claude Code workflow for safer AI code review, with review receipts, MCP boundaries, and team guardrails.
Run Review Agents With Receipts
06.26.26A practical Codex workflow for llm code review, with AGENTS.md guardrails, MCP boundaries, and a copyable review receipt.
Train Your Team on Codex Remote
06.26.26A practical Codex training guide for teams adopting Codex Remote, AGENTS.md, MCP, and reviewable CLI workflows.
AGENTS.md for Codex Teams
06.25.26A practical AGENTS.md convention for Codex CLI teams that want safer MCP access, verification loops, and reviewable diffs.
Compare AI Coding Agents Safely
06.25.26A practical governance matrix for comparing Cursor, Claude Code, and Codex in enterprise ai code generation workflows.
Compare Coding Agents for Teams
06.25.26A practical governance matrix for comparing Codex, Claude Code, and Cursor across repo rules, MCP boundaries, and review loops.
Compare Coding Agents With Guardrails
06.25.26Compare Claude Code with a governance matrix covering repo rules, tool permissions, test evidence, and review expectations for each run.
MCP Setup for Claude Teams
06.25.26Make a Claude Code integration maintainable by assigning ownership, checking actual connector access and testing the intended hook events.
Use Cursor Rules as Team Rails
06.25.26A practical Cursor rules workflow for teams using subagents, skills, AGENTS.md, and reviewable coding habits.
Accept Claude Skills Like Code
06.24.26A team guide for accepting Claude Code skills with CLAUDE.md, hooks, MCP notes, and a review rubric.
Bounded Benchmarks for Coding Agents
06.24.26How engineering teams can use signed, isolated benchmarks to govern Claude Code and other coding agents.
Review Code With Codex CLI
06.24.26A practical Codex CLI review workflow with AGENTS.md rules, MCP checks, verification loops, and a PR receipt.
Team Boundaries for Coding Agents
06.24.26A practical workflow for setting AI coding boundaries, signed benchmark checks, and review guardrails in Cursor.
Train Your Team on Cursor 3.9
06.24.26Use Cursor 3.9’s Customize page to roll out rules, skills, subagents, MCPs, and team habits.
Verify Coding Agents in Isolation
06.24.26Signed isolation bundles help teams test coding agents with clear tool boundaries, review guardrails, and repeatable evidence.
A Cursor MCP integration checklist
06.23.26Connect one Cursor MCP server with scoped credentials, verify a known read and denied access, and leave a short integration note the team can repeat.
Check Codex MCP access before a task
06.23.26Verify Codex MCP configuration, authentication and session access with a harmless read, then record the permission boundary before changing code.
Govern Autonomous Codex Workflows
06.23.26A practical Codex workflow convention for governing autonomous coding agents with states, MCP boundaries, and review checks.
Govern Autonomous Coding Agents
06.23.26A team convention for agentic coding governance: boundaries, MCP permissions, and review checks Cursor users can adopt.
Govern Codex Agents With Workflows
06.23.26A practical Codex team convention for autonomous coding agents, review guardrails, MCP boundaries, and safe verification loops.
Guardrails for Autonomous Coding Agents
06.23.26A practical Cursor team convention for autonomous coding agents, with rules, MCP limits, and review guardrails.
How Autonomous Coding Agents Ship Safely
06.23.26A practical Codex team convention for autonomous coding agents, MCP boundaries, and code review guardrails.
How Coding Agents Stay Inside Bounds
06.23.26A practical team convention for governing Claude Code, MCP access, hooks, and autonomous agent reviews.
Let Coding Agents Work Safely
06.23.26A practical Claude Code team convention for agentic coding workflows, MCP boundaries, hooks, and safer review.
Make Coding Agents Follow a Workflow
06.23.26A practical convention for governing coding agents with Claude Code rules, MCP boundaries, and review guardrails.
Rules for Autonomous Coding Agents
06.23.26A practical Cursor-first convention for agentic coding governance, MCP boundaries, and reviewable autonomous changes.
Subagent conventions for Claude Code teams
06.23.26Define one Claude Code subagent role, restrict its tools, test its handoff on a small task, and keep evidence that a teammate can review.
A Team Operating Model for Claude Code
06.22.26A practical Claude Code team operating model covering CLAUDE.md, hooks, MCP permissions, and review habits.
Agentic Coding for Mixed-Skill Teams
06.22.26A practical team convention for matching coding agents to skill levels, Cursor rules, MCP access, and review guardrails.
Codex 26.616 Workflow Check
06.22.26How to respond to Codex app 26.616 with AGENTS.md rules, CLI verification, MCP boundaries, and review habits.
Codex Conventions for Mixed-Skill Teams
06.22.26A practical Codex convention for AI coding training, skill rubrics, MCP boundaries, and review guardrails.
Codex Workflow After App 26.616
06.22.26A practical Codex workflow response to app 26.616: AGENTS.md rules, CLI verification, MCP boundaries, and reviewable diffs.
Cursor 3.8 Code Review with Subagents and Skills
06.22.26Cursor 3.8 code review workflow using Automations, subagents, skills, and MCP, plus a review receipt template teams can reuse.
Govern Coding Agents by Skill
06.22.26Shared Cursor skills need boundaries a reviewer can inspect, with different acceptance criteria for routine and sensitive changes.
Make Cursor Reviews Leave a Trail
06.22.26A review receipt ties the Cursor diff to its intended scope, actual commands, skipped checks and remaining uncertainty.
Skill Gates for AI Coding Teams
06.22.26A Claude Code convention for accepting agent skills, setting MCP boundaries, and reviewing mixed-skill team work.
Skill Rubrics for Coding Agents
06.22.26A team convention for reviewing agent skills, Codex workflows, MCP boundaries, and mixed-experience AI coding.
Team Conventions for Claude Code
06.22.26A practical guide to CLAUDE.md, hooks, MCP, and review habits for teams adopting Claude Code.
AI Code Review Needs a Receipt
06.21.26A practical Codex-first review workflow for governing coding agents across tools, MCP servers, and team rules.
Code Review Agents Need Receipts
06.21.26A practical review workflow for Cursor teams using agents, review receipts, MCP boundaries, and shared rules.
Codex AGENTS.md Team Rules
06.21.26A team convention for Codex AGENTS.md files, with scoped rules, MCP boundaries, and a reviewable verification loop.
Cursor Automation Team Conventions
06.21.26A practical Cursor team convention for using /automate, rules, skills, AGENTS.md, and review habits safely.
Pick the Review Workflow, Not the Bot
06.21.26A Claude Code-first guide to AI code review guardrails, MCP boundaries, and a pasteable review receipt.
Practical Rules for Claude Code Teams
06.21.26A practical guide to CLAUDE.md, hooks, MCP, and review habits for using Claude Code across a real engineering team.
A Safer Codex Review Loop
06.20.26Use Codex CLI, AGENTS.md, verification, and review receipts to make AI-assisted code review safer for teams.
A Team Claude Code Review Workflow
06.20.26A practical Claude Code review workflow with CLAUDE.md, MCP boundaries, hooks, and a paste-ready PR receipt.
AI Coding ROI With Guardrails
06.20.26A practical governance workflow for measuring AI coding ROI with Cursor rules, MCP boundaries, and review guardrails.
Cursor Rules for Team Automations
06.20.26A practical Cursor team convention for rules, skills, AGENTS.md, and safe always-on automations.
Governed AI Coding at Team Scale
06.20.26A Claude Code workflow for measuring team AI coding ROI with skills, MCP boundaries, and review guardrails.
ROI Guardrails for Coding Agents
06.20.26A practical governance workflow for Codex teams using AGENTS.md, MCP boundaries, skills, and review checks.
Add MCP to Codex Safely
06.19.26A practical Codex CLI MCP guide for teams: configure one server, document AGENTS.md boundaries, and verify diffs.
Cursor MCP for Team Workflows
06.19.26A practical Cursor MCP guide for teams using rules, skills, automations, and reviewable agent workflows.
Team AI Coding Training Plan
06.19.26Review a Codex workflow by checking the changed paths, actual permission controls and evidence attached to the handoff.
Train Coding Agents Safely
06.19.26A Cursor-first training guide for rolling out coding agents with rules, MCP boundaries, and review guardrails.
Claude Code or Copilot for Teams
06.18.26A practical Claude Code vs GitHub Copilot guide for teams setting CLAUDE.md, hooks, MCP, and review conventions.
Cursor Review Workflows With Subagents
06.18.26Build a Cursor 3.7 review workflow with cloud agents, repo rules, focused subagents, and skills. Record the checks and review evidence in every pull request.
Install Codex CLI for Teams
06.18.26Install Codex CLI with AGENTS.md, MCP boundaries, and a verification loop your team can reuse for reviewable diffs.
A practical context budget for Codex CLI
06.17.26Team convention for Codex CLI context limits, AGENTS.md rules, MCP boundaries, and verification loops.
Claude Code conventions for an engineering team
06.17.26Team conventions for Claude Code: CLAUDE.md, hooks, MCP, skills, and review habits engineers can actually use.
code review tools Need Receipts
06.17.26A practical read on the workflow, tradeoffs, and next steps. Read the workflow, review rules, and team training patterns for AI coding tooling.
Cursor Agents Team Checklist
06.17.26A Cursor Workshop guide for using Cursor agents, rules, skills, and Bugbot in reviewable team workflows.
AI Code Review Workflow for Teams
06.16.26A practical team convention for reviewing AI-assisted code without slowing delivery or losing ownership.
Codex Code Review for CLI Teams
06.16.26A Codex CLI review workflow for AGENTS.md rules, MCP boundaries, verification loops, and review receipts.
Cursor Skills and Team Rules for Large Codebases
06.16.26How Cursor skills, .cursor/skills files, and team rules work together so large codebases stay reviewable, not just faster to edit.
How to enable agent teams in Claude Code
06.16.26A practical Claude Code agent-team checklist for CLAUDE.md, subagents, hooks, MCP, and review control.
Cursor MCP Support and How to Add a Server Safely
06.15.26A practical look at Cursor MCP support, how to add one server, scope it with a rule, and review agent work without losing control.
Hands-On Claude Code Workshops for UK Development Teams
06.15.26A practical Claude Code team convention guide for shared CLAUDE.md rules, review habits, and adoption under delivery pressure.
Cursor Skills for Team Workflows
06.14.26Specific skill descriptions and scoped tasks help a team judge whether a Cursor handoff contains enough evidence to accept.
Shared agent workflows for review risk reduction
06.14.26A practical read on the workflow, tradeoffs, and next steps. Read the workflow, review rules, and team training patterns for AI coding tooling.
Cursor Agent Review Loop
06.13.26Use cursor agent with rules, subagents, and review checks to ship safer code faster in Cursor teams.
Cursor Rules for Team Workflows
06.12.26Cursor rules help teams keep scoped .cursor/rules files, subagents, and reviews aligned in Cursor.
AI code review habits
06.11.26A practical read on the workflow, tradeoffs, and next steps. Read the workflow, review rules, and team training patterns for AI coding tooling.
AI coding training and compliance regulations: a working model
06.03.26Build an AI coding review process around repository scope, secret handling, recorded checks and escalation decisions, with an onboarding checklist.
AI coding training requirements for DevOps teams on Codex CLI
06.02.26The five operational training modules DevOps teams need before Codex CLI touches delivery paths: guardrails, secrets, CI evidence, rollback, ownership.
What is MCP in Cursor IDE? The answer for Codex CLI teams
06.02.26The practical answer to what is MCP in Cursor IDE, plus the repo contract Codex CLI teams need before connectors touch shared code.
Codex Windows review loop
05.31.26Codex CLI review workflow for openai codex cli review code changes, AGENTS.md, MCP, and verification loops.
Cursor 3.6 auto-review and rules
05.31.26Cursor auto-review is one part of the approval path; inspect how allowlists, sandbox boundaries and human decisions interact.
Safer Agents for Legacy Code
05.31.26ai coding training for legacy codebases with review guardrails, MCP boundaries, and phased rollout steps.
Codex Windows update: MCP and mobile use
05.30.26A Codex CLI workflow guide for how to add mcp to codex, AGENTS.md, and reviewable verification loops.
Agent Code Review Without Drift
05.29.26A practical review contract connects task scope, regression evidence and the engineer responsible for accepting an agent-authored change.
Codex CLI 0.135.0 workflow
05.29.26A repeatable CLI workflow records the installation, applicable instructions, connector access and commands that verify the resulting change.
Cursor in Jira: tighter scope, cleaner handoffs
05.29.26How Cursor in Jira scopes work, where Cursor MCP fits, and the team checklist that keeps agent work reviewable.
Codex CLI boundaries
05.28.26A Codex CLI guide for AGENTS.md, MCP, verification loops, and reviewable diffs. Read the workflow, review rules, and team training patterns for MCP.
Cursor Jira review workflow
05.28.26A Cursor code review workflow for Jira handoffs, rules, subagents, and a review receipt teams can use this week.
Onsite or Virtual AI Coding Training
05.28.26Practical guidance for distributed engineering teams choosing onsite ai coding training for distributed teams or virtual delivery.
Codex CLI review receipts
05.27.26A Codex code review workflow for AGENTS.md, MCP, and verification loops that turns CLI diffs into reviewable receipts.
Cursor in Jira, with clear handoffs
05.27.26A Jira handoff should expose scope, subagent responsibilities, connector access and the checks a reviewer still needs to inspect.
Training agents without chaos
05.27.26Practical ai coding training cost for engineering teams: guardrails, review checks, and a budget worksheet.
Agentic coding guardrails
05.25.26Practical ai coding training for large teams: review guardrails, MCP boundaries, and team habits that improve delivery.
Claude Code Hooks, PreToolUse, and MCP Permissions
05.25.26A plain guide to Claude Code hooks, PreToolUse, and MCP permissions, turned into team conventions a reviewer can check.
Codex CLI workflows for reviewable diffs
05.25.26A Codex CLI workflow guide for openai codex cli github, AGENTS.md, MCP, and reviewable verification loops.
Cursor Jira workflows without chaos
05.25.26Connect a Jira brief to the Cursor task, applicable rules and verification evidence so reviewers can compare the result with the request.
Claude Code MCP team conventions
05.24.26Practical Claude Code guidance for claude mcp, CLAUDE.md, hooks, and reviewable team conventions.
Codex CLI Goal Mode, Appshots, and the Real Workflow
05.24.26What Codex CLI Goal Mode and Appshots change day to day, and the AGENTS.md and verification habits that make them worth using.
Claude Code team skills and conventions
05.23.26Claude Code 2.1.150 tightens team conventions around claude code skills, CLAUDE.md, hooks, and shared repo context.
Codex CLI review loops
05.22.26A Codex CLI workflow guide for codex cli mcp, AGENTS.md, MCP boundaries, and verification loops teams can use to ship reviewable production work.
Cursor Automations in the Agents Window
05.22.26Cursor 3.5 moves Automations into the Agents Window, with multi-repo and no-repo workflows for cursor ai teams.
Claude Code team conventions
05.21.26Claude Code teams need one shared way to set context, review changes, and keep agent work inside repo rules.
Codex CLI, GitHub, and MCP
05.21.26A Codex CLI workflow guide for codex github, AGENTS.md, MCP, and reviewable diffs in production repos.
Cursor Automations Across Multiple Repos, Without Drift
05.21.26How to scope Cursor Automations across multiple repos and MCP connectors so agent runs stay reviewable instead of drifting.
Claude Code hooks for teams
05.20.26A practical guide to Claude Code hooks, CLAUDE.md, MCP, and team review habits. Read the workflow, review rules, and team training patterns for MCP.
Codex CLI 0.132.0, AGENTS.md and MCP Official Docs
05.20.26What the Codex CLI 0.132.0 release changes for AGENTS.md rules and MCP connectors, and how to keep diffs reviewable.
Cursor Jira handoffs without drift
05.20.26Turn Jira tickets into scoped Cursor tasks with clear acceptance criteria, shared instructions and a completion check before the ticket closes.
Govern AI Coding Before It Drifts
05.20.26Practical ai coding governance for engineering teams: review guardrails, MCP boundaries, and shared rules across tools.
Claude Code review receipts
05.19.26A practical Claude Code review workflow for team conventions, CLAUDE.md, hooks, MCP, and a copyable review receipt.
Cursor Composer 2.5 team guide
05.19.26Organize Cursor team instructions into focused rules, single-purpose agent tasks and reusable skills, then review their scope with a shared checklist.
Agentic Coding Breaks At The Handoff
05.17.26Most teams do not lose control when an agent writes bad code. They lose it when nobody can explain the change ten minutes later. The handoff is the interface.
Claude Code 2.1.139 team conventions
05.17.26Claude Code 2.1.139 team conventions: a CLAUDE TOC, red-folder approvals, data-class tags on MCP connectors, and a weekly retro note.
Codex governance: four contracts that hold in review
05.17.26A codex governance note for engineering teams: the slash catalog, verification latch, browser bridge note, and model pin that keep Codex CLI work reviewable.
Best practices for agentic coding in real environments
05.16.26An operating guide to best practices for agentic coding in real environments: rule-file precedence, scope ledgers, replay receipts, connector cards.
CLAUDE_CODE_STOP_HOOK_BLOCK_CAP in Claude Code 2.1.143
05.16.26What CLAUDE_CODE_STOP_HOOK_BLOCK_CAP means in Claude Code 2.1.143, and how to investigate a repeatedly blocking Stop hook.
Codex mobile CLI: runs that survive review
05.16.26A codex mobile cli pattern for running Codex CLI from anywhere: verification latches, model pins, and connector rosters that keep remote runs reviewable.
Claude Code 2.1.142 team conventions
05.15.26Claude Code 2.1.142 team conventions for parallel agent streams: a skill index, a hook budget, a CLAUDE TOC, and red-folder approvals.
Codex mobile CLI docs your team can read anywhere
05.15.26The codex mobile cli question is a docs question: how a team keeps AGENTS.md rules, run notes, and verification transcripts readable away from the desk.
Codex workflows for mobile handoffs
05.15.26Codex workflows for mobile handoffs: the repo contract of model pins, connector rosters, done checklists, and slash catalogs that lets agent work change hands.
Cursor 3.4 cloud agents
05.15.26A workflow note on Cursor 3.4 cloud agents: connector stewards, glob diets, prompt expiry tags, and receipts that survive remote runs.
Cursor Bugbot effort levels as review policy
05.15.26Cursor Bugbot effort levels mapped to review risk: lanes, an escalation list, and a triage table so deadline pressure stops deciding review depth.
How to Set Up Agentic Coding Workflows and Guardrails
05.15.26How to set up agentic coding workflows and guardrails for Codex, Claude Code, and Cursor, with receipts a reviewer can actually check.
Codex workflows: governance that lives in the repo
05.14.26How to govern codex workflows from the repo: a connector roster, a ten-line done checklist, a slash catalog, and a verification latch reviewers can replay.
Cursor Cloud Agent Setup and the Environment Contract
05.14.26A practical guide to Cursor cloud agent setup, covering reproducible environments, a secrets boundary, and review evidence before agents edit.
Codex auto review: setup, strengths, and limits
05.13.26Codex-auto-review trials showed Codex catching syntax drift and missing permission drift. The fix is transcript evidence and repo contracts, not more autonomy.
Cursor subagents and skills for teams
05.13.26Cursor subagents and skills for team repos: glob diets, prompt expiry tags, skill preambles, and mutex paths that keep parallel work explainable.
Fast mode is not the default: when fast models earn it
05.13.26The fast model is a tradeoff you make on purpose: scope ledgers, replay sandwiches, and connector cards that keep fast agent runs reviewable.
AI coding agents workflow guardrails for browser control
05.11.26Workflow guardrails for AI coding agents with browser control: child receipts, decision stubs, scope ledgers, and a supremacy clause reviewers can audit.
Claude Code 2.1.136: team convention fixes
05.11.26Claude Code 2.1.136 team convention fixes for engineering teams: a weekly retro note, a skill index, a hook budget, and a CLAUDE TOC reviewers can audit.
Is Codex CLI Giving MCP Too Much Reach?
05.11.26Codex CLI does not widen MCP reach on its own. The fix for reviewer anxiety is an AGENTS.md line naming what each connector may do.
AI agent guardrails that hold
05.10.26A field guide to AI agent guardrails for recursive agent chains: connector ownership, child receipts, and review evidence that survives the merge queue.
Claude Code 2.1.138: team conventions check
05.10.26A Claude Code 2.1.138 team conventions check: PR receipts, a weekly retro note, a skill index, a hook budget, and a CLAUDE TOC reviewers can audit.
Codex workflows for Chrome and the CLI
05.10.26Codex workflows that cross into Chrome: the browser bridge note, model pin, connector roster, and done checklist that keep two surfaces telling one story.
Agentic workflows from PR to merge
05.09.26A PR review workflow for agentic coding teams: connector ownership, scoped tasks, replay transcripts, and human approval lanes from PR to merge.
Claude Code 2.1.137 fixes Windows activation
05.09.26Claude Code 2.1.137 fixes Windows activation; the team conventions that matter are a weekly retro note, a skill index, a hook budget, and a CLAUDE TOC.
Cursor subagents and skills for teams
05.09.26An operational memo on Cursor subagents and skills for teams: mutex paths, skill preambles, prompt expiry tags, and review receipts in the repo.
OpenAI Codex CLI AGENTS.md Official Docs, Explained
05.09.26What the OpenAI Codex CLI AGENTS.md official docs cover, and the review habits that keep an update from breaking your workflow.
Composer 2 for Cursor teams
05.08.26A measurement guide for Composer 2: known-case evals, acceptance gates, and the review notes that separate faster drafts from better ones.
Claude Code 2.1.132 team conventions
05.07.26Claude Code 2.1.132 team conventions for engineering teams: a skill index, a hook budget, a CLAUDE TOC, and red-folder approvals reviewers can audit.
Cursor 3.3 context for subagents and skills
05.07.26Cursor 3.3 context for subagents and skills: review receipts, connector stewards, glob diets, and expiry tags that cut merge-queue fatigue.
Claude Code 2.1.129 team conventions
05.06.26Claude Code 2.1.129 team conventions: skill precedence, a hook budget, a CLAUDE TOC, and red-folder approvals that end review-thread archaeology.
Codex CLI Workspace Tools for Reviewable MCP Connectors
05.06.26Codex CLI workspace tools that make MCP connectors reviewable, including a model pin, a connector roster, and a done checklist.
Cursor 3 subagents and skills
05.06.26Review a Cursor subagent handoff with an independent caller search, actual connector permissions and a check of the combined diff.
Claude Code 2.1.128 team conventions
05.05.26Claude Code 2.1.128 team conventions: red-folder approvals, data-class tags, a weekly retro note, and a skill index that ends precedence trivia nights.
Cursor team controls for admins
05.05.26Cursor team controls for admins as an operating guide: connector stewards, mutex paths, review receipts, and rules that stay trustworthy.
AI agent guardrails: why every harness needs them
05.04.26Why agent harnesses need guardrails: AI agent guardrails that turn complete-sounding summaries into receipts reviewers can actually verify.
Claude Code conventions for shared repos
05.04.26Claude Code conventions for shared repos: a weekly retro note, a skill index, a hook budget, and a CLAUDE TOC that keep parallel streams reviewable.
Cursor Canvases for team workflows
05.04.26A guide for using Cursor Canvases as decision artifacts that bind owners, paths, and rules, instead of diagrams that drift away from review.
Claude Code 2.1.121: MCP guardrails
05.03.26Claude Code 2.1.121 MCP guardrails: data-class tags, a weekly retro note, a skill index, and a hook budget reviewers can check before merge.
Cursor Security Review beta
05.03.26A playbook for Cursor Security Review: red-folder routing, finding triage, and the PR evidence a named human owner still signs.
Claude Code 2.1.122 team conventions
05.02.26Claude Code 2.1.122 team conventions for clogged merge trains: a weekly retro note, a skill index, a hook budget, and a CLAUDE TOC reviewers can audit.
Codex 5.5: pin the model before you swap it
05.02.26Codex 5.5 questions are model governance questions: pin the default model and escalation rule in AGENTS.md, and keep browser checks bridged to CLI receipts.
Cursor team marketplace: rules with named owners
05.02.26A Cursor team marketplace works when shared rules, skills, and MCP connectors carry named owners: review receipts, stewards, glob diets, expiry tags.
Claude Code 2.1.126: MCP, hooks, skills
05.01.26Use Claude Code 2.1.126 with clear MCP ownership, a hook budget, a skill index, and a short weekly review that keeps team conventions current.
Codex CLI /goal Command and AGENTS.md Workflow
05.01.26The Codex CLI /goal command explained, and how pairing it with AGENTS.md keeps task intent and repo boundaries clear at handoff.
Cursor Subagents, the Official Docs and the Missing Contract
05.01.26What the Cursor subagents official docs cover in .cursor/agents, and the ownership contract your team still has to write.
Cursor 3.2: subagents, worktrees, multi-root
04.29.26A migration memo for Cursor 3.2 teams deciding how to use subagents, worktrees, and multi-root workspaces without losing review ownership.
Agent boundaries for teams running coding agents
04.28.26How to set agent boundaries for teams: connector ownership, written scopes, and review receipts that keep agent diffs explainable after the session ends.
How to clean up agent-written code
04.26.26A working memo on how to clean up agent-written code: restore visible scope, ownership, and verification receipts to agent diffs before review.
Morning signal review: an AI code review checklist
04.25.26Run a morning signal review with an AI code review checklist: precedence files, replay proof, connector owners, and receipts checked before merge.
Headless SaaS for agents: APIs before dashboards
04.24.26An operational memo on headless SaaS for agents: scope ledgers, decision stubs, and replay receipts for vendor APIs without dashboards.
Agent drift: how to read coding agent output
04.23.26How to spot agent drift in coding output: the reading habits, receipts, and scope checks that catch it before the merge.
Local first: why local checks before CI win for agents
04.22.26Local checks before CI, argued for agent work: replay sandwiches, connector cards, and child receipts created where the work happens.
MCP for team workflows: scopes, owners, receipts
04.21.26MCP for team workflows means treating every connector as a production dependency: declared scope, named owner, written rollback, reviewable receipts.
Specs for coding agents that hold up in review
04.19.26A field guide to specs for coding agents: five-line scope ledgers, decision stubs, and connector cards that a tired reviewer can check.
Stop adding a regression test for every bug
04.18.26Why a regression test for every bug fails as strategy: guard failure classes, keep replay receipts, and stop growing suites that prove nothing.
An agent-friendly codebase beats a clever prompt
04.16.26An agent-friendly codebase keeps scopes, receipts, and verification commands in files, so agent diffs stay reviewable and delegation stays safe.
An AI coding workflow that holds up under audit
04.15.26An AI coding workflow built on receipts: child receipt blocks, decision stubs, scope ledgers, and precedence files that survive audit.
Lead sourcing with agents that survives review
04.14.26A field guide to lead sourcing with agents: connector cards, replay receipts, and review evidence for outreach pipelines built on Cursor and MCP.
Narrow skills for coding agents beat broad ones
04.13.26Skills for coding agents stay reviewable when each one has a single job, a declared scope, and a pinned verification command. Broad skills hide drift.
Evaluating AI coding tools that hold up
04.11.26Evaluating AI coding tools by the audit trail they leave: connector cards, child receipts, decision stubs, and scope ledgers a stranger could review.
MCP for UI components: boundaries before the demo
04.10.26MCP for UI components without silent rework: connector cards, CLAUDE.md precedence, replay records, and child receipts for design-system agent work.
Fast evals for better coding agent decisions
04.09.26Fast evals for coding agent workflows: replace review archaeology with receipt checks, decision stubs, and scope ledgers reviewers can run in minutes.
Keep parallel coding agents from colliding
04.08.26Parallel coding agents clog merge trains when scopes overlap. Scope ledgers, CLAUDE.md precedence, and replay receipts keep streams out of each other.
Specs and tests: the stable stack for AI coding
04.07.26Specs and tests as the stable stack for agent work: four named fixes that turn fuzzy scopes into reviewable, parallel-safe delivery.
Stop using CSS selectors in E2E tests
04.06.26CSS selectors in E2E tests churn every time an agent regenerates markup. Durable selectors, decision stubs, and scope ledgers keep the suite reviewable.
A better bug-finding prompt for coding agents
04.05.26A better bug-finding prompt is one reviewers can replay: receipts, precedence, and connector boundaries for agent-led debugging.
Agentic team workflows that survive review
04.04.26A field guide to agentic team workflows: the scope, ownership, and verification receipts that keep parallel agent output reviewable after the session ends.
AI coding tools that keep working after rollout
04.03.26Why AI coding tools degrade after rollout, and the file-backed contracts, precedence, replay receipts, connector cards, that keep them working.
Cursor Composer layers in agentic coding
04.02.26A field guide to Cursor Composer layers in agentic coding: decision stubs, scope ledgers, and precedence files that keep work reviewable.
Headless agent runs in CI need receipts
04.01.26Headless agent runs in CI work when AGENTS.md mandates replay receipts, connectors carry owner cards, and child agents report exactly what they touched.
Long-running agent loops you can still review
03.31.26How to keep long-running agent loops reviewable: replay contracts, connector boundaries, and receipts that outlive the session.
Coding agent loops for messy code
03.30.26Coding agent loops survive messy code when scope ledgers, CLAUDE.md precedence, replay receipts, and connector cards keep every iteration reviewable.
AI coding tools that last past the demo
03.29.26AI coding tools last when their output survives review: CLAUDE.md precedence, replay sandwiches, connector cards, and child receipts, applied in practice.
AI coding team workflow beats prompt tuning
03.28.26An AI coding team workflow built on child receipts, decision stubs, and scope ledgers outperforms prompt tuning, and it survives the people who wrote it.
Markdown files for coding agents: the real interface
03.27.26Markdown files for coding agents, CLAUDE.md, AGENTS.md, rules, and connector cards, are the contract layer that keeps agent work reviewable and owned.
Browser checks for coding agents that reviewers can replay
03.26.26Browser checks for coding agents: turn UI claims into replayable receipts with intent lines, command transcripts, connector cards, and decision stubs.
Keep the AI coding stack current and agents bounded
03.25.26Updating the AI coding stack without losing agent bounds: connector cards, child receipts, decision stubs, and scope ledgers that survive upgrade weeks.
Reviewing AI generated code defensively
03.24.26Reviewing AI generated code defensively: decision stubs, scope ledgers, and replay receipts that let reviewers defend a merge without replaying the chat.
Coding Agent Browser Automation Needs an Owner
03.23.26Coding agent browser automation moves faster but widens blast radius fast. Give every connector a named owner before you trust it.
AI coding wrappers that hold up under review
03.22.26A governance guide to AI coding wrappers: the repo contracts Cursor, Claude Code, and Codex need so agent work stays reviewable.
Subagent prompts: why every fork needs its own brief
03.20.26Why subagent prompts need their own scope, paths, and verification: four named fixes that keep forked agent work explainable in review.
AI coding tool rollout without surprise rework
03.18.26An AI coding tool rollout survives scale when discipline travels as artifacts: replay sandwiches, connector cards, and decision stubs every team can audit.
Reliable AI coding workflows under pressure
03.16.26Reliable AI coding workflows run on receipts, not autonomy: scope ledgers, replay sandwiches, and connector cards that keep agent streams reviewable.
AI coding workflow patterns that survive review
03.15.26Four AI coding workflow patterns that keep agent work reviewable: CLAUDE.md precedence, replay sandwiches, connector cards, and child receipts.
Async subagents that speed up AI coding workflows
03.14.26Async subagents speed up AI coding workflows when every fork returns receipts: paths touched, commands run, and tests that prove regression guards.
How returning Markdown from docs shapes agentic coding
03.13.26Returning markdown from docs gives Cursor, Claude Code, and Codex one reviewable contract: scope, constraints, verification, and owner on every run.
Cursor .mdc vs CLAUDE.md vs AGENTS.md
03.11.26Compare Cursor .mdc rules, Claude Code CLAUDE.md, and Codex AGENTS.md: what each file controls, where it belongs, and how teams review it.
Playwright MCP: faster agent loops with receipts
03.09.26A Playwright MCP workflow note for faster agent loops, focused on connector ownership, scoped access, review logs, and test evidence.
What to do when AI coding tools regress
03.07.26When AI coding tools regress, the teams that recover fastest are the ones whose receipts survive the update: connector cards, child receipts, decision stubs.
Running multi-agent teams without losing the review trail
03.01.26Running multi-agent teams across Cursor, Claude Code, and Codex: scope ledgers, precedence files, and replay records that keep every diff reviewable.
Cursor 2.4 subagents and skills for engineering teams
01.28.26A Cursor 2.4 operating model for subagents and skills: scope ledgers, rule precedence, artifact-first review, and a one-branch training drill.
Put research into practice
Work with Rogier and Vasilis on the tools and codebase your team uses. Onsite, offsite, or remote.