Posts

Ever had a Claude Code session debug a gnarly JWT rotation bug, then three weeks later watch your new Codex agent wrestle the same issue from scratch? Yeah, me too. So I run three coding agents across eight projects. That’s months of conversation logs — design decisions, debugging rabbit holes, configuration gotchas. All sitting on my disk in format-specific JSONL files I couldn’t search. Honestly, every new agent session felt like I was hiring a dev who’d never seen the codebase. The short version: Deja-vu is a single Go binary that reads every session log Claude Code, Codex, opencode, aider, Gemini CLI, Cursor, Antigravity, and Grok Build ever wrote — and gives you instant search plus MCP recall over the whole thing. But no database, no daemon, no service to run. Yet it works retroactively on months of logs from before you installed it. So I’ve been running it for a week. Now here’s what I found. ...

Ever asked Claude Code to build a landing page and got the same purple-to-pink gradient hero, the same centred-everything layout, the same icon-tile feature cards with a fake Trustpilot rating underneath? Yeah, me too. Every LLM has a default design vocabulary. It’s aggressively boring. But Hallmark — a design skill by Nutlope (the same person behind screenshot-to-code, now under Together AI) — refuses to ship that. It sits inside Claude Code, Cursor, and Codex with 57 slop-test gates, a pre-emit self-critique, and twenty design themes that genuinely look different from each other. 11,568 GitHub stars and 3,372 stars in a single day says the community agrees this is a real problem. ...

Ever looked at your Claude Code subscription dashboard on a Wednesday and realized you’d burned through half your monthly Fable 5 quota already? Yeah, me too. Two weeks ago I was staring at 60% of my 2M-token quota gone by midweek, wondering if I was just bad at coding or if something was fundamentally broken about how I was using the tool. Turns out, it’s neither. The problem is that every single Claude Code session — every search, every edit, every test run, every git blame — draws from the same Fable 5 quota pool. And most of those tasks don’t need a frontier model to do them. ...

Ever spent an afternoon debugging a Playwright script because some website changed a single CSS class name? Yeah, me too. But browser automation has been a pain for years — you chain selectors, wait for elements, and pray nothing moves. browser-use flips that entirely. So here’s the idea instead of a fragile selector chain: give an LLM a screenshot of the page and a goal, and let it figure out the clicks and keystrokes. The result? 105,000 GitHub stars and the fastest-growing open-source browser agent framework I’ve seen this year. ...

Ever had your pager go off at 3 AM for what turned out to be a single noisy pod restarting in a loop? Yeah, me too. Then you spend the next hour digging through Datadog dashboards, filtering logs, and manually stitching together what happened — while your phone keeps buzzing with alerts that are all the same incident. I’ve been running AI agent stacks on my own infra for a while — containerized agents with sandboxd, orchestrating them with PilotDeck. So when I saw Superlog hit 1,029 stars in 44 days and land in Y Combinator’s P26 batch, I paid attention. An open-source observability tool that uses AI agents to automatically aggregate noise into incidents and investigate root causes? And I had to try it. ...

Ever had a Claude Code session where you spawn three subagents, paste the same 9,000-token plan into each prompt, and cross your fingers? Look, me too. So last week I ran a build script across three parallel agents. But each one re-read the full plan on every turn. Even my token bill for that single session? Just shy of $3. And worse — when I found a bug in the plan after spawning them, I had to paste the correction into three separate conversations. And one of them never got the update and ran the wrong analysis for 20 minutes before I noticed. ...

So you’ve got Claude Code in one terminal. Codex in another. OpenCode on your laptop, Cursor CLI on your desktop. Each has its own config, its own model preferences, its own CLI quirks. I was spending more time switching context between agent setups than actually writing code — and let me guess, so were you. Still, that’s the exact pain Claude Code Router (CCR) fixes. With 35,792 GitHub stars and a Kimi (Moonshot AI) sponsorship, it’s not a side project anymore. It’s the agent control plane the ecosystem needed. ...

But ever looked at a page Claude Code just built and thought — yep, that’s AI? Purple gradient hero, 3-column feature grid, Inter font everywhere, all centred in a 100vh box. It works, but it screams “template.” That’s where Hallmark comes in. It’s a design skill for Claude Code, Cursor, and Codex that enforces 58 slop-test gates before any code leaves the agent’s mouth. 6,613 stars on GitHub, trending #2 today with 1,015 stars in the last 24 hours. Made by Nutlope (the Together AI team). ...

Ever watched your AI agent turn “add a date picker” into 404 lines of code, a flatpickr npm dependency, and a wrapper component with its own CSS? Yeah. Me too. So when I saw a GitHub repo called Ponytail hit 83,449 stars in its first month—with a tagline that reads “He says nothing. He writes one line. It works."—I had to install it immediately. TL;DR: Ponytail is a plugin for Claude Code, Codex, and 20+ agent CLIs that installs a 7-rung “ladder of restraint” before every code generation step. But the result? 54% less code, 22% fewer tokens, 20% lower cost, 27% faster sessions—and 100% safety on adversarial inputs. Yet a bare “write one-liners” prompt only manages 95% safety. ...

Disclosure: Some links below are affiliate links. If you sign up through them, I may earn a commission at no extra cost to you. Vultr — starts at $6/mo DigitalOcean — $200 credit for new users Here’s a question I haven’t seen anyone ask about AI coding agents: what happens when the model is wrong — not just bad code, but malicious? Every agentic tool today — Cursor, Claude Code, Codex, OpenCode — works the same way: the model proposes an action, and the tool trusts it. Maybe it asks for confirmation before running shell commands. But the model picks which commands to run. The model chooses which files to read. The model decides which npm packages to install. The model holds the keys. ...