A year ago, building a custom tool for one library upgrade would've been a waste of time. With agents, the math flips. Things not worth building before are suddenly worth it: run once, then delete. I walk through one I built for a React Query migration. blog.tomaszgil.me/throwaway-by...

Astro 7 is here! A new Rust compiler, a new Rust Markdown/MDX processor, Vite 8 and more. Get ready for 60%+ faster builds. astro.build/blog/astro-7/

The best software I’ve written has always been software I was a user of. Using it every day made the rough edges impossible to ignore.

The number of sessions I managed to have with Fable before the shutdown: 1. 🤐

Huge thanks to meet.js for having me at the Summit — the heart of Poland's JS community since 2011 — for my first-ever large-stage talk. Terrifying, and one of the best things I've done as an engineer. Recording's up! 👇 www.youtube.com/watch?v=mfCM...

Final post in the Offline Support for Web Apps series. 📴 Image bytes don't belong in your app state. This one is about splitting the data layer — data in one place, image bytes in another — and letting a service worker handle the transport. More in the article. blog.tomaszgil.me/offline-supp...

Hear me out: self-healing codebases. 👀 Periodically feed the agent all rules and architecture docs and let it apply "safe" fixes.

I've been moving off of Cursor, to Claude Code. Man, it's slow. 🥱

New post in Offline Support for Web Apps series. 📴 Prefetching can make offline UX feel seamless — or completely tank performance. This one breaks down how to do it carefully: when, what, and how much. More in the article. blog.tomaszgil.me/offline-supp...

I'm switching away from Cursor, diving into Claude Code. 👀 What is your preferred way of interacting with it? I'm specifically interested in something for managing your conversations within and across projects.

Days since I used `useImperativeHandle` in React: 0.

Today: trying out Ralph. 🤖

TypeScript 6.0 is now available! This release brings better type-checking for methods, new standard library features, new module features for Node.js, and more! But most important, this release brings us one step closer to the upcoming native-speed 7.0! devblogs.microsoft.com/typescript/a...

📚 Continuing my series about design-systems, today I wrote about why I believe data-testid is a bad practice and and how role-based selectors actually help ensure your app is accessible.

I’ve stopped writing code by hand. I hesitated to post this because it already feels obsolete — my workflow is evolving as I write this. Still, I break down how I currently work with AI agents end-to-end. More in the article. blog.tomaszgil.me/the-agent-fi...

AI promised me I’d finally have time to post unhinged memes on social media. Now I’m just doing more work and endlessly tweaking my workflow. Send help.

Roughly two weeks ago I drafted a post about how I use AI in my workflow, reflecting on the last 6 months or so. It already feels obsolete. I wonder if I should even post it. 🙈

New post in the Offline Support for Web Apps series. 📴 This one moves from queue implementation to real application usage — wiring the queue into React, capturing offline mutations, and syncing them safely later. More in the article. blog.tomaszgil.me/offline-supp...

Being good at context switching becomes a lot more important when you have multiple agents running simultaneously on different things. In my experience, most devs prefer focused work and being “in the zone”. I feel this is gonna be a big adjustment and why many are still sceptical. It’s hard.

i really don't care about using AI to ship more stuff it's really hard to come up with stuff worth shipping i want to ship the same amount of stuff with higher quality both in product and code

I was fixing debouncing problems in a global search component. AI couldn't figure it out, multiple tries, attempting weird hacks. Jumped in, debugged rendering cycles, found an unmemoized component at the top of the tree, causing full page re-renders. One-line fix. Take that, LLMs.

Just finished speaking at meet.js Summit 2026 about Shipping Offline Support on the Web. 🎤 It was my first time presenting at a conference of this scale — and what an experience. Huge thanks to the organizers. Appreciate the engaged audience, sharp questions and conversations afterward. 🙏

There's no legitimate reason to "write code by hand" anymore.

I didn't realise how or when it happened exactly, but after almost 10 years in the industry, I have a small network of people I can rely on, discuss things and ask for help. It's incredibly valuable. I'm really grateful for that.

Slides ready. Excited for next week. 👀

If you haven’t had a chance to grab your ticket for the 15th meet.js Summit yet (psst… I’ll be speaking there 👀), here’s a -10% promo code for you! 🎟️ cart.easy.tools/checkout/bra... See you on March 4th! 🚀

Ok, I think I'm ready. 👀 cursor.com/docs/configu...

I morged a featue today, no tim wasted.

Today, I prepared an animated, interactive visualization with Codex 5.3, for my upcoming talk about offline support on the web. I’ve never written custom canvas/svg animations. Until today. So much fun.

It's been only yesterday, when a friend mentioned that with only skimming through AI-authored code, he already feels it's harder to come back to it to maintain it, fix bugs, etc.

everyone's talking about their teams like they were at the peak of efficiency and bottlenecked by ability to produce code here's what things actually look like - your org rarely has good ideas. ideas being expensive to implement was actually helping 1/4

I'm beginning to suspect that a key skill in working effectively with coding agents is developing an intuition for when you don't need to closely review every line of code they produce. This feels deeply uncomfortable!

The 2025 State of React survey results are now available! 2025.stateofreact.com/en-US Thanks to @aurorascharff.no for providing a great conclusion.

New post in the Offline Support for Web Apps series. 📴 This one dives into retry strategies for offline mutations: why a single attempt isn’t enough, how exponential backoff works, and how to keep syncing pauseable and safe. More in the article. blog.tomaszgil.me/offline-supp...

Honored to join the speaker lineup at meet.js Summit 2026. I'll be speaking about shipping offline support on the web. See you there! 🚀

I prepared a rough script for my talk next month. I focused only on the most important bits... I shapes up to be around 30+ minutes. Now I need to cut it in half for it to fit in the conference slot. Send help. 🤐

Status update: wife's sick, kid's sick, I'm still holding on.

My first ever CFP submission just went through. I guess I have a month to prepare a talk now. 🤐

New post in the Offline Support for Web Apps series. 📴 This one moves into the UI layer: exposing queue state, subscribing to updates, and wiring a single long-lived foreground queue into React. More in the article. blog.tomaszgil.me/offline-supp...

"For you" feed has become quite usable for me recently. 👀

Trying to avoid getting sick when half of your kid's daycare has flu.

Moltbook, reddit-like forum, where agents talk with other agents (mostly using Opus 4.5 with Claude Code via OpenClaw). This is something else. Enter at your own peril. www.moltbook.com/post/34809c7...

I never really understood the phenomenon of tools like Claude Code or Codex CLI and the fact that they run in a terminal. Terminal is the wrong place for UIs like that. It's almost like trying to have a full editor experience in a terminal... Oh, wait.

Once, my friend asked me if by teaching others I'm not creating competition for myself. I get where he was coming from, but that's not how I see it. We aren't in a race, we are in a life-long hackathon. If someone levels up, everyone wins. And I win, too.

Approx. 90% of the code I ship is written by agents, yet I still don't really know what MCPs are. AMA.

Small tweaks in how you use LLMs can save a lot of tokens. Here are 3 steps that worked well for me in practice. 👇

New post in Offline Support for Web Apps series. 📴 This one dives into the hard parts of a foreground queue: atomic read–modify–write, enqueue-time deduplication & bounds, and a single-flight sync loop you can safely pause. More in the article. blog.tomaszgil.me/offline-supp...

Hot take: if your company does not allow engineers to use Cursor or Claude Code, they're actively hurting your career. (no, Copilot doesn't count)

Now that AI writes most of my code, I think I should get back to posting more memes.

“This is a small codebase, there can’t be much dead code”… Added knip.dev. Deleted ~800 lines immediately. Yeah. Absolute no-brainer. Incredible value. 👏 @webpro.nl