Articles about development, AI and open source.
At DevDay, OpenAI added MCP Events support to ChatGPT. A plugin can now wake the agent on its own when a new ticket, order, or complaint comes in. What it does, how it works under the hood, who is behind the standard, and what to watch out for.
What mods in Claude Code are, how they differ from hooks, and how developers, designers, PMs, QA, and ops could use them. Plus what to check before installing someone else's mod.
A new version of Claude Profile Manager: per-profile MCP servers and settings, separate session history, profile management from the command line, cpm exec for scripts and CI, and an API key check before cloud sync.
On June 12, 2026, the US government forced Anthropic to switch off Fable 5 and Mythos 5 — three days after launch, with a single order. This isn't a fight about price or an outage, it's a new class of risk: the kill switch on your infrastructure is held by a foreign executive. And that's exactly why European AI sovereignty is risk management, not ideology.
Anthropic dropped Claude Fable 5 at $10/$50 and half of Twitter closed the budget debate with "2x more expensive, I'll wait." But a new tokenizer, always-on thinking, and the refusal mechanics make two the wrong number — up and down. What you need to price is the finished outcome.
Claude Code starts every session from zero. What if the agent reflected each evening instead of summarizing? A thought experiment on memory and pruning.
Transcribe meetings and interviews into text right on your Mac – offline, free, and in Czech. A step-by-step guide from installing Homebrew to your first finished transcript, without uploading sensitive data to anyone else's server.
A CMU/UMD paper shows that what limits SSM-attention hybrids on deep reasoning isn't memory capacity but the number of passes over the context. A 'sleep' mechanism runs N recurrent passes at window eviction and consolidates the context into fast weights, moving computational depth offline.
navigator.modelContext still returns undefined, but behind that word hides a new way for the web to expose tools straight to an agent — no server. Turn on the flag, register a tool, and test it.
Anthropic shipped the /goal slash command. A session-scoped wrapper over a prompt-based Stop hook: you write a condition, a small evaluator reads the transcript after each turn, and the session ends only when it says yes. When to reach for /goal, when for /loop, when for a Stop hook — and why you write the condition for Haiku, not for the main model.
Anthropic announced GA of Claude Platform on AWS. Full Claude API through IAM, invoice in the AWS console, identical pricing. Skills, Managed Agents, Advisor pattern and MCP connector – all the same day, no weeks of waiting like with Bedrock. What it actually unlocks for a dev and the team.
Meta Ads AI Connectors in open beta. Official MCP server and CLI, 29 tools with read+write access to your ad account. Claude and ChatGPT can pause campaigns, adjust budgets, and write ad copy. What it does, what it doesn't, and why it's the biggest Meta ecosystem story this year.
One day in Karlín, three people, Expo + Supabase + NativeWind. What you can ship in 7 hours for a non-profit that employs people with disabilities — and what I'd do differently next time.
From March 4 to April 20, 2026, three concurrent degradation modes ran in Claude Code. Anthropic frames them as bugs — but two of the three were deliberate decisions.
My bet: by spring 2027, isolated context per sub-task gets a folk name and jumps from engineering blogs into pitch decks. Why the Lütke–Karpathy–Anthropic pattern keeps repeating.
The certification question has one correct answer out of four — a PreToolUse hook on Read and Grep. In production, your `.env` still leaks through Bash. A working hook that closes it.
Agents deliver sloppy code because they get sloppy briefs. A practical recipe for turning a Jira ticket into binary acceptance criteria, three-tier boundaries and self-verification.
Whisprflow replaced my spec writing. I dictate a messy brain dump into Claude Code, the agent turns it into a structured spec and asks whether it understood me. Ten minutes instead of an hour — and a better output.
Pinterest layout used to mean Masonry.js or CSS columns with broken reading order. display: grid-lanes changes that — native masonry with a two-line @supports fallback.
PreToolUse, PostToolUse, Notification. What Claude Code hooks are, how to configure them in settings.json, and the three patterns I use in every production repo.
Not a spec sheet. A direct comparison of Cursor and Claude Code based on daily production use — editor integration, agentic workflow, hooks, skills, cost and when to pick each.
A talk on AI agents in software development. CLAUDE.md, hooks, skills, Skillsmith, git worktrees and real-world case studies — everything I showed at Yolk Studio.
The Claude Agent SDK gives you the same primitives Claude Code uses — tools, hooks, subagents, permissions. A practical walkthrough of shipping one to production.
Personal subscription, company team, Vertex AI setup — and Claude Code stores everything in one directory. I built a CLI tool that gives each account its own isolated environment.
Your team uses Claude, Cursor, and Copilot — and you're maintaining the same coding standards in five places. I built an open-source tool that fixes this with one command.
AI agents waste tokens parsing your navbars and cookie banners. I built an open-source library called next-markdown-mirror that fixes this — and it costs $0.