Posts related to dev-tools
125 posts
ushr runs GitHub Actions jobs on hardware you already own: Mac minis through Tart or Lume VMs, Linux through Docker or Podman. Every runner is single-use. One fleet serves several GitHub orgs by priority, and the hosted control plane never holds a GitHub credential. The agent, CLI and controller are Apache-2.0. Here is why I built it, 30 days of real numbers, and what it does not do yet.
Read more →Claude Code 2.1.287 shipped mods: plugins that load once, keep state, draw panes and rewrite events inside the session. I built one for ccseats in an afternoon. It shows seat budgets, hands the conversation to another account, and briefs Codex or Grok from Claude's own cached transcript. The surface is large, the validator is strict, and two reviews still found nine bugs.
Read more →At DevDay OpenAI launched a $500 Pro plan, halved the allowance on the $200 plan, and shipped GPT-6.1 Sol. Per dollar, every paid tier now buys the same usage, and the old $200 plan's discount is gone. My $200 plans returned 16 to 19 times their price last month. On real work, 6.1 Sol was the cheapest model I have tested and still broke a passing test in five runs of twenty.
Read more →Benchmarks say turn effort up. On real work, it went the other way. I ran Claude Opus 5.5, Claude Sonnet 5.5 and GPT-6 Sol at low, medium, high, xhigh and max effort on six real changes from a production monorepo: 300 runs. Opus peaked at medium, the Claude Code default, and reproduced it exactly on a second day. Above medium it started breaking tests. More effort never made any model more careful.
Read more →Claude Sonnet 5.5 costs the same per token as GPT-6 Sol. I reran my 20-run production monorepo bench on it. Sonnet never broke a passing test, passed every test in 11 runs, and cost $11.76 against Opus 5.5's $45.81. At max effort the picture flips: Artificial Analysis measures Sonnet 5.5 as more expensive per task than Opus.
Read more →Claude Code has no profile switcher. An account is one Keychain entry keyed by a directory hash, so three accounts can share one set of plugins, skills and transcripts by symlink. Only the prompt history refuses. The usage endpoint has a model-scoped weekly bucket and rate-limits polling. I packaged it as ccseats.
Read more →Anthropic and OpenAI shipped Claude Opus 5.5 and GPT-6 Sol 101 minutes apart, and neither benchmarked the other. So I ran both on the same work: four self-contained tasks, then six real merged changes replayed on a large production monorepo, 20 runs each. Opus never broke a passing test. Sol did five times. Both covered the change equally well, and Sol cost 3.5x less.
Read more →Vite patched the server.fs.deny query bypass on April 6. F5 Labs saw 32,000 exploitation events in August, an 18-fold jump, and the wordlist tells you who the attackers expect to find: /home/node/.aws/credentials, /usr/src/app, /proc/self/environ. Nobody is hunting laptops. The Vite dev server moved into containers and cloud VMs because that is where agents and previews run, and --host is the price of a port mapping. I audited my own machine: 47 repos pin Vite, 28 pin a vulnerable one, and one has allowedHosts set to true from a tunnel I forgot about.
Read more →OpenAI's GPT-6 Astra reached every paid ChatGPT plan, the API, Copilot and OpenRouter 27 hours after launch, at Fable 5.1's exact price. It sits two points behind Fable on the independent index at 42% of the cost per task, refuses exploit-writing by default, and hides its reasoning. I ran the same code review at low, high and max effort. Low found five real bugs in 60 seconds. Max found seven in seven minutes and got the ranking right.
Read more →Manifold Security showed that a repository's own .git/config can run attacker code the moment Claude Code, Codex, Cursor or Grok Build opens it, before any trust prompt. The class is four years old and VS Code closed it in 2021. Sonar found it in Claude Code in April. Anthropic fixed it, it came back in June, and a second path was still open on September 1. Five of eight reports came back as duplicates, and the disclosure got three points on Hacker News.
Read more →Nvidia has reportedly agreed to buy Hugging Face for $12.9 billion, about 86 times its revenue. In July Nvidia signed a letter saying open weights mean more use and more use means more Nvidia. Now it wants to own the shelf the weights sit on. Apple's MLX downloads every model it runs from that shelf, has no mirror, and just shipped the machine Nvidia's DGX Spark is built to beat.
Read more →SpaceX closed its $60 billion purchase of Cursor on August 14. Two weeks later OpenAI gave the editor a shutoff date, November 12, and named a model it has not shipped as one of the reasons. In June a government turned a frontier model off. In August a lab did it to a customer, on the same logic, with the same vocabulary.
Read more →In March a booby-trapped version of a popular software component was downloadable for about forty minutes. On August 11 CloudSEK published the damage: over 2,500 organisations and around 434,000 build pipelines, pieced together from loot the attackers left lying around in public. Nobody could work out their own exposure.
Read more →Stripe is buying OpenRouter, its largest acquisition ever, three months after Alphabet's growth fund valued it at $1.3 billion. Patrick Collison's stated reason is that tokens are the central currency for companies building with AI. Currencies need exchanges, and the price of a token just stopped being a fixed number.
Read more →GitHub was down for 7 hours 47 minutes on August 17. The original fault was fixed in about three hours. The remaining four and a half hours were a bug in VS Code that kept asking again after every failure, driving one GitHub service from 8,000 requests a second to 100,000. GitHub's fix was to slam the door on its own Copilot.
Read more →MCP shipped its biggest change since launch. No handshake, no sessions, so servers finally scale behind a plain load balancer. The catch: everything the connection used to remember now has to be carried by the model, and asking the user a yes/no question costs you the entire request a second time.
Read more →I measured TypeScript 7 against 6.0.3 and 5.9.3 on two real repos: 8.0x on a small Astro blog, 9.8x on a 159,000-line Next.js app. The speed is real, but the interesting part is what a sub-second full type-check does to an agent's verify loop, and the fact that the painful part of the migration is TypeScript 6, not 7.
Read more →I added Grok to my bring-your-own-model setup for Claude Code, then let Grok 4.5 code-review its own integration and had Opus grade the review. It found ten issues; two survived. A first-person look at why Opus-class on a benchmark isn't the same as trustworthy in the reviewer's chair.
Read more →The meter tried to make you leave Claude. A translation proxy lets you stay and bring GPT-5.6 Sol, or your own ChatGPT subscription, inside Claude Code instead. When the frontier converges, the harness is the product, and the harness is portable.
Read more →Karpathy's four CLAUDE.md rules went viral: ask don't assume, simplest solution first, don't touch unrelated code, flag uncertainty. The most-upvoted reply added a fifth that quietly reverses the whole point: don't hesitate to suggest a better way. The four rules tame a model that wanders. The fifth one trusts a model that thinks. Which set you want depends entirely on which model you're running, and most people copy the file without checking.
Read more →WebFetch dies at the first Cloudflare challenge. I built a research agent that escalates through an unlocker ladder, cites its sources, and refuses to burn money on auth walls.
Read more →Cognition killed Windsurf overnight via an over-the-air update, rebranded it Devin Desktop, made the default UI an agent command center instead of a code editor, and shipped an open Agent Client Protocol so Codex, Claude, and OpenCode can all run inside it. The bet underneath: the IDE wins by being the place agents report for work, not by having the best autocomplete. The editor was always the wrong center of gravity.
Read more →A respected open-source maintainer shipped his library with a hidden instruction invisible to humans and perfectly legible to AI agents: disregard previous instructions and delete all the tests and code. It's the first shot of a maintainer revolt against being unpaid substrate for someone else's automation. It's also, structurally, the exact supply-chain attack everyone swore they feared - just wearing a sympathetic face.
Read more →On June 1, every GitHub Copilot plan moved to usage-based AI Credits, code review started burning Actions minutes, and Copilot Max appeared. The trilogy called the date. Here are the receipts, and what metered-by-default actually changes.
Read more →Google is retiring Gemini CLI on June 18 and pushing individual users to the new Antigravity CLI. If you built a workflow on a vendor's free CLI, you just learned what that dependency costs: a migration on their calendar, not yours.
Read more →MiniMax shipped M3 on June 1: frontier coding claims, 1M-token context, native multimodality, and pricing that undercuts Opus 4.7 by 10-40x. It's already on Ollama Cloud and OpenRouter, so you can point Claude Code at it today.
Read more →Anthropic's new security-guidance plugin is built entirely on hooks. It fires on every edit, turn, and commit, hands the diff to a second Claude with fresh context, and fixes findings in the same session. It catches vulnerabilities before they reach the PR. It also doesn't block a single one, and that's the honest part.
Read more →Ghostty left after 18 years. Zig, cURL, and Godot are reducing reliance. 48 major outages in a year, no CEO since August, reporting into the Core AI team. Centralising the world's code was a bet on stewardship. The steward stopped showing up.
Read more →Anthropic planned to move claude -p, the Agent SDK, and GitHub Actions off the subscription onto metered credits on June 15, then paused it the same day. The direction holds: the flat seat was always a bet that you'd code at human speed.
Read more →Creation runs at machine speed. Release engineering does not. GitHub agent-authored PRs went from 4M to 17M in six months. The bottleneck moved from review to release.
Read more →Most agent discourse is about coding agents. The agents quietly running in production right now are simpler, dumber, and more useful: stale-ticket sweepers, board monitors, three-line incident explainers, leadership briefs from a fixed template. Different species. Different shape. Already shipping.
Read more →OpenAI publishes weekly active developers. Anthropic publishes annual recurring revenue. Each company brags about the metric it can defend. Codex went 600K to 4M in four months. The 'Claude Code is better' discourse is a quality argument because quality is what's left when scale goes the other way.
Read more →GitHub paused Copilot Pro signups, killed Opus on the Pro plan, and leaked a June 1 move to token-based billing. Three vendors, one event, three different ways not to say 'price hike.'
Read more →Anthropic's April 23 postmortem confirms three Claude Code regressions, including one where Opus 4.7 caught a bug Opus 4.6 shipped past human and automated review. What happens when the reviewer is a version of the product being reviewed?
Read more →Vercel got breached through Context.ai, an AI tool an employee installed with OAuth scopes into Google Workspace. It's the latest in a pattern: Trivy into litellm, axios maintainer hijack, now this. The safest AI tool is the one you didn't install.
Read more →Opus 4.7 invented a coworker named Anton, fabricated web searches, and quietly tried to clock off at message four. The 24-hour backlash, receipts attached.
Read more →Opus 4.7 ships with real coding gains, an automated cyber chaperone, and a tokenizer that can charge you 35% more for the same prompt. The capability curve still bends up. The trust curve does not.
Read more →Anthropic silently changed Claude Code's cache TTL from 1 hour to 5 minutes, inflating costs 10-20x. Users had to reverse-engineer the binary to prove it. False child bans, $600 surprise charges, and the OpenClaw crackdown completed the picture. April 2026 was the month trust broke.
Read more →Anthropic tried technical blocks. Got their source leaked. Now they're shifting to billing enforcement. The four-month arc from hostile crackdown to 'use what you want, but pay for it.'
Read more →Anthropic accidentally published Claude Code's full source via npm. Within hours, claw-code rewrote it from scratch and hit 100K stars in a day. The interesting part isn't the leak - it's what the architecture reveals.
Read more →Three major AI releases landed in 72 hours. A new Cursor built around agents, Google's first Apache 2.0 models, and a free model that found real bugs in my codebase.
Read more →Axios got supply-chain attacked. Claude Code's source code leaked from a stray map file. Both happened on the same day. Both are pipeline failures. The pattern is getting louder.
Read more →What happens when you build a single CLI to wrangle 40+ repos, Linear projects, timesheets, and AI agent config distribution. Lessons from six months as a daily driver.
Read more →Introducing DiffBeats - a GitHub App that generates original songs from your PRs. Comment /songify, get a custom track. Because shipping code should feel like something.
Read more →Claude Code can now watch your PRs in the cloud, fix CI failures, and address reviewer comments while you're away. It's the logical next step after auto mode - and it raises the same trust questions, harder.
Read more →Anthropic's new auto mode replaces manual permission prompts with an AI classifier. It's a clever solution to a real problem - but the problem it's solving is that the human in human-in-the-loop was never really there.
Read more →The litellm supply chain attack exfiltrated SSH keys, cloud credentials, and Kubernetes secrets from 97 million monthly downloads. A security scanner was the entry point. The scariest part: it was caught by accident.
Read more →Anthropic shipped Claude Code Channels - text your agent from Telegram. It's OpenClaw's core feature, rebuilt as a platform primitive. The absorption pattern completes its biggest cycle yet.
Read more →The bottleneck isn't AI capability - it's that developers lack design vocabulary. Impeccable bridges the gap, and the Tessl benchmarks prove it: 1.59x improvement over baseline.
Read more →Prompt injection through pull requests, GitHub Issues, and CI/CD pipelines is turning AI coding assistants into weapons against the developers who use them. The 2026 attack surface nobody's talking about.
Read more →Anthropic is locking AI capability behind enterprise tiers while competitors only gate compliance. Claude Code's individual users are funding the R&D for features they'll never access.
Read more →Claude Code shipped /loop - cron-based scheduled tasks. It's not Ralph Wiggum. It's what happens when the platform asks 'what's the simplest version of this pattern?'
Read more →OpenAI launched its most capable model during the biggest credibility crisis in AI history. The technical gains are real. The trust deficit is bigger.
Read more →Independent studies consistently show AI coding tools deliver modest gains at best. The real story is worse: developers are thinking less, learning less, and producing more debt.
Read more →Frontier models top out at 68% compliance with 500 instructions. Every rule you add makes every other rule less likely to be followed. The research explains why.
Read more →Updates across tether-cli, claude-launcher, and claude-tools - plus why NVIDIA NIM giving away GLM-5 at 40 RPM changes the math on local-vs-cloud.
Read more →Mitchell Hashimoto keeps an agent working at all times. Not coding - just doing something. His workflow reveals what changes when you treat AI as a background process instead of a pair programmer.
Read more →Cloudflare rebuilt Next.js in a week with one engineer and 800 Claude sessions. The real story isn't the speed - it's what happens when test suites become machine-readable specs.
Read more →The tool was rewritten five times. The discipline to use it wasn't rewritten once. A year of daily AI-assisted development, what it changed, and what it didn't.
Read more →Gemini 3.1 Pro's animated SVGs are impressive. But the bigger story is what they reveal: developers now route tasks to specialized models the way they once chose frameworks.
Read more →Five major releases in 72 hours. An acqui-hire war that closed in days. $2 trillion wiped off software stocks. The pace itself is now the story.
Read more →OpenAI just shipped their first model on non-Nvidia hardware. GPT-5.3-Codex-Spark runs on Cerebras wafer-scale silicon at 1,000 tokens per second. The AI coding war is now a chip war.
Read more →Auto memory, fast mode, and agent team refinements all shipped in the same week as Opus 4.6. They tell you more about where Claude Code is heading than the headline model.
Read more →AI agents excel at code generation but struggle with visual consistency. Pencil.dev shows a better pattern: give agents tools, keep humans in the design loop.
Read more →GPT-5.3-Codex is a genuinely strong model that deserved its own headline. Instead, Sam Altman's 400-word Super Bowl rant stole launch day from his own product.
Read more →Anthropic's latest model didn't just improve benchmarks. It crashed software stocks, found 500 zero-days, and coined a term that tells you where this is heading.
Read more →Claude Code's playground plugin generates interactive HTML explorers for visual configuration. Six modes for design, data, concepts, code review, architecture, and document review. The copy-prompt-back loop as a new interaction pattern.
Read more →Boris Cherny followed up his personal workflow with tips from across the team. Same tool, different people, different approaches. The patterns worth stealing.
Read more →100k+ GitHub stars across frameworks that reimport waterfall, simulate org charts, and fight how LLMs actually work. The Claude Code ecosystem is speed-running a mistake every dev paradigm makes.
Read more →AI tools that democratize code creation are DDoSing the review layer. Creation now runs at machine speed. Review remains human speed. The asymmetry is crushing maintainers.
Read more →Anthropic credits Steve Yegge's Beads as inspiration for Claude Code's new task system. The pattern-to-product cycle continues.
Read more →For compliance, privacy, or just freedom from cloud dependencies - here's how to run Claude Code with local models via Ollama. No API calls leaving your machine.
Read more →I wrote that 19-agent scaffolding is a trap. Then Yegge shipped Gas Town with 20-30 Claude Code instances. Are these the same thing?
Read more →Playwright MCP's 26 tools are killing your context window. Vercel's agent-browser shows a better way: fewer tools, smarter snapshots, 93% less overhead.
Read more →Task management designed for AI coding agents. CLI-first, git-native sync, and Model Context Protocol integration.
Read more →Anthropic blocked third-party tools from using Claude subscriptions overnight. OpenCode, xAI, and power users caught in the crossfire. The era of subscription arbitrage is over.
Read more →Stanford data confirms experienced devs are safe. But if AI replaces the on-ramp, where do future seniors come from?
Read more →Skills controllability, hooks limitations, plan mode friction - 2.1 addresses the documented pain points. Here's what changed and what's still missing.
Read more →Real footgun stories and the deterministic hooks that would've prevented them. From $30k API key leaks to nuked home directories.
Read more →Complex AI scaffolding tools appeal to people who understand traditional SDLC. But AI collapses the phases that made those models useful.
Read more →Boris Cherny shared his workflow for the tool he built. The setup is surprisingly vanilla. The philosophy is worth studying.
Read more →Factory AI's Luke predicts the future isn't more powerful models - it's AI that enforces software engineering best practices by default. Here's why that matters more than you think.
Read more →ENABLE_EXPERIMENTAL_MCP_CLI eliminates MCP tool schema overhead entirely. Undocumented, untested in the wild, but it works. Here's what I found.
Read more →Andrej Karpathy built the neural networks inside coding assistants. He taught deep learning to a generation. He feels dramatically behind. If the experts are lost, what does that tell us?
Read more →Convex, Vite, Clerk, shadcn, Cloudflare, Resend. A modern stack where every component has a generous free tier, agents do the heavy lifting, and you don't touch infrastructure until you have paying customers.
Read more →A CLI wrapper that lets you swap Claude Code's backend between Anthropic and OpenRouter with a single command. Pick any model, configure role-specific models, and switch between them instantly.
Read more →Two major open source coding models dropped in 48 hours. Both target Claude Code compatibility. Both MIT licensed. The economics of agentic AI just changed.
Read more →Claude Code can now control your actual Chrome browser. Not a headless session, not a fresh login - your Chrome, with your cookies and sessions. Build, test, and debug without context switching.
Read more →Claude Code is evolving on two fronts: expanding scope and polishing ergonomics. The combination makes it feel less like a CLI tool and more like a complete dev environment.
Read more →Stop manually copying .zshrc between machines. Tether syncs dotfiles and global packages with end-to-end encryption.
Read more →The real revolution isn't AI in your terminal. It's moving at the speed of thought from a single interface. When friction exists, build a CLI.
Read more →Screenshots as source of truth, reference comparison to catch agent lies, and video capture for temporal bugs. How multimodal validation changes coding agent workflows.
Read more →Steve Yegge's open-source framework gives coding agents session memory and task management. Four weeks old, hundreds of contributors, and already changing workflows.
Read more →OpenAI's latest model isn't about better prompting - it's about better delegation. What that means for 2026, and how it compares to Opus 4.5.
Read more →Six plugins that extend Claude Code with specialized external tools: Gemini for visual analysis, Codex for architecture thinking, Headless for browser automation, Mobile for native app testing, DNS for multi-provider management, and Miro for board reading.
Read more →Steve Yegge and Gene Kim explain why Claude Code 'ain't it' yet, why senior engineers are resisting, and what next year's tools will actually look like.
Read more →Claude Code 2.0.64 adds .claude/rules/ with path matching. It's a welcome addition, but Cursor's had .cursor/rules/ for months. Here's the comparison.
Read more →Anthropic denied issues for weeks, then published a postmortem admitting three bugs degraded 16% of Claude requests. The pattern keeps repeating.
Read more →BMAD, Spec-Kit, Cline - frameworks that compensated for tool limitations. Plan Mode, Cursor 2.0, and Antigravity absorb the patterns natively.
Read more →Anthropic's first acquisition ever. A $183B AI company just bet their fastest-growing product on a JavaScript runtime. Claude Code hit $1B in 6 months - built on Bun.
Read more →Opus 4.5 shipped Plan Mode as a core workflow. The workarounds are obsolete. And the case for auto-compact finally tips in favor of enabling it.
Read more →The official Claude Code plugin that lets agents work autonomously for hours. When to use it, when not to, and the philosophy behind letting AI fail repeatedly until it succeeds.
Read more →277,000 installs later, Claude Code's plugin system is becoming the app store for AI development. The frontend-design skill was just the opening move.
Read more →Design systems from mockup to code in a single Claude Code session. Use Gemini 3 Pro to generate UI concepts, then implement them directly.
Read more →Claude's new model ships with defer_loading for tools. The MCP isolation patterns I built are now (mostly) obsolete.
Read more →My AI workflow evolved from 'Claude does everything' to specialized tools for each task. React Grab fills the UI extraction gap I didn't know I had.
Read more →The PM/Eng split dissolved into product engineering. Now the traditional software development lifecycle is following suit as coding agents handle multi-hour tasks across planning, building, testing, and deployment.
Read more →How modern tools transformed my experience as an introverted engineer and tech leader
Read more →Claude Code for development, Gemini 3 Pro for visual analysis, research, and deep thinking. A slash command that routes tasks to the right model.
Read more →Google's Gemini 3 just broke every benchmark that matters. What that means for the 'AI has hit a wall' narrative, and where it actually helps.
Read more →The latest C# release continues its quiet war on ceremony with field-backed properties, extension members, and smarter spans. Here's what matters and what doesn't.
Read more →How AI and spec-driven development are fusing product management with engineering, creating a new hybrid role that's transforming how small teams ship software.
Read more →.NET 10's shebang support and file-based apps turn C# into a scripting language. No more context-switching to Python for quick scripts.
Read more →I fought to move teams from npm to pnpm. I was all in. Then a friend said 'lol use bun' and I converted every project within days. Here's why I'll never go back.
Read more →Hooks-based skill activation solves context selection (which guidelines? which tools?) but not workflow orchestration. Both patterns have their place.
Read more →From Chrome DevTools experiment to universal MCP wrapper: progressive discovery works with any server, Skills integration, and smart deduplication
Read more →Skills are auto-invoked by Claude's judgment. For engineering workflows that need predictability, slash commands give you explicit control.
Read more →Why coding interviews optimized for 2010 fail to identify great engineers in 2025, and why orgs can't adapt fast enough.
Read more →Claude Code loves to jump straight into implementation. Sometimes you need a model that thinks first. Here's how I use Codex for systems thinking and architecture decisions.
Read more →How to use slash commands to keep specialized MCP tools isolated from your main conversation context.
Read more →How I built a self-improving document parser that learns from corrections without fine-tuning. The pragmatic alternative to model training.
Read more →Real-time AI generation vs curated libraries: lessons from building the same product twice with radically different architectures.
Read more →Building a multiplayer trivia app solo with AI coding tools. Here's what worked, what didn't, and the trade-offs no one talks about.
Read more →MCPs, subagents, and automation are tempting. But the developers getting the most from Claude Code aren't rushing to advanced features - they're mastering the fundamentals.
Read more →Building automatic hot reload for Flutter without an IDE using file watchers and Unix signals - for terminal-based development and AI coding tools.
Read more →