Ivan Jeremic

@theo @haydendevs Codex feels gr8 to wk w/ rn.

If Codex maintains a high quality bar and refuses to ship slop, they will win.

@TylerFCloutier @AnthonyPoschen Thanks! I guess if someone were to build a "Vitess for SQLite" it would solve those limitations too.

@ivanjrmc @AnthonyPoschen SQLite is great and if you tune it properly and do appropriate batch commits you can get quite high throughput (100k+), but you have to be careful about lock fairness. Also you don't get replication and subscriptions and type code generation and all the rest of it.

I've decided to get into the balls game. Let's visualize the difference between SpacetimeDB throughput and throughput of Node + Postgres.

@TylerFCloutier @AnthonyPoschen I would really be curious to hear the answer to that regarding SQLite?

@AnthonyPoschen Why don't people actually do that you think? 🤔

We wildly underestimate TypeScript’s compiler. It was the perfect chance to evolve JS without breaking the web: just var + const (compiler fixes scoping), fn instead of function, cleaner syntax, real fixes. We had the tool. We barely used it.

@matteocollina Yes! Because then all three big runtimes might one day become one.

Bun rewrote itself from Zig to Rust. AI did most of the work. 98% of the test suite passed on the first run. The question isn't hypothetical anymore. Should we rewrite Node.js in Rust?

@JLarky @webdevcody WebDevAgent

@webdevcody WebDevNotCodyAnymore

no

The WordPress plugin/PHP ecosystem invented easily running scripts and getting hacked.😅

@matteocollina @thdxr Why not rebuild npm around Cargo? Its security mess is probably from being too massive & loose. Cargo’s strict model could fix that, or just integrate AI scanners to vet every package for exploits.

@thdxr What’s your take? I’m pretty aware of the constraints, and there isn’t a lot of wiggle room.

everyone's ideas for fixing the npm security issue shows how basically no one is capable of thinking at the scale of this problem

God, I hope we will see a merger of Bun and Deno projects like we did with Node.js and io.js.

@ChrisLaubAI @garrytan A Rust dev not long ago also "killed" Electron...

A Rust dev just killed Headless Chrome. It's called Obscura. The open-source headless browser purpose-built for AI agents and scrapers at scale. Chrome vs Obscura: - Memory: 200MB+ → 30MB - Binary: 300MB+ → 70MB - Page load: 500ms → 85ms - Startup: 2s → Instant - Anti-detect: None → Built-in Single binary. No Node, no Chrome, no dependencies. Stealth mode is brutal: → Per-session fingerprint randomization (GPU, canvas, audio, battery) → 3,520 tracker domains blocked by default → navigator.webdriver masked to match real Chrome → Native function masking so detectors can't sniff it out Drop-in replacement for Puppeteer and Playwright over CDP. Zero code changes. If you run agents or serious scraping at scale, this repo prints money. 100% Opensource.

Love the @tan_stack CLI, but it would be great to support installing into the current directory without requiring a project name.

@theo Vanilla CSS is now an option too, since AI writes it anyway.

Can’t believe I spent years learning all the Tailwind class names just to have AI write them for me

@ASUS_ROG I keep getting bombarded with administration requests from your VirtualPet software, which I never installed. I did not check the box for Virtual Pet when installing Armoury Crate. This is broken and unacceptable—please fix it!

Having a separate domain for VitePlus makes the ecosystem around Vite more confusing. There was no need for it; it could have just been a route on the @vite_js website.

One frustrating thing about working in Codex on a project: chats don’t automatically share project context. If I’ve already made architecture decisions in other chats, I have to restate them again. Any chat under the same project should understand the repo, prior decisions, and what was discussed before. Project-level memory should be the default.

@rauchg I would even say that people who don't know how to code have zero chance of publishing an app to production.

Not knowing how to code giving you an advantage is absolute nonsense. The more you understand, the better your prompts, the better the feedback you give, the better product you ship. What will change is that the intricacies of syntax, compilers, module systems, the finer details of type systems, won’t matter as much to everyone. But you should absolutely understand how the pieces fit together. From syscall to pixels. Learn how data flows, because you’ll be able to secure your systems. Learn about performance, because you’ll be able to push your agent further. Learn about APIs, because they determine how to integrate systems. Learn about how systems fail, because you’ll be able to make reliable programs.

@devongovett There is a browser plugin I use exactly for this scenario.

All web forms should have auto save. I just spent 2 hours writing a very detailed comment on a GitHub PR, went to press Cmd + A to select all and copy, but accidentally hit Cmd + Q instead and lost all of my work. Infuriating. 😡

@cursor_ai, removing/hiding the Agent/Editor toggle is not just a mistake—it's also a bad business move.

@levelsio PHP is the problem; it would never get lost in JS.

I literally told it was in gen_webhook.php, it just had to add this one line Claude Code from a month ago would have one-shotted this in 5 miliseconds Claude Code today gets lost and confused

Claude Code with Opus 4.6 was so dumb today I finally had to write my own code again A sad state of affairs 🥹

@RyanCarniato @DelaneyGillilan I'd pay for a stream w/ u two 😅

As much as I hate to say it. You need to compare against better competitors. Going up against Next.js is like shooting fish in a barrel, even in frontend circles. I think also the example needs to be simple enough that no one can contest it, yet play to the competitors wheelhouse. These things only have weight if the competitor recognizes the scenario which usually means you have to play to their strengths. In the early days of Solid I put it up against DOM diffing scenarios because that was the Solid's worst case and React's arguably best case. The real problem is that if people can't carry the performance context in their head these things fall apart. Like Qwik incredibly good at hydration never came up with the right demo/benchmark to show off their performance. The problem is that if no one cares about the performance difference between Next and literally anything else in the frontend world, you are going to have a hard time getting the right attention. Generally use their own benchmarks against them. Which I imagine is hard though because it isn't apples to apples. Hypermedia's concern has never been performance. It is that people don't believe it is capable of scaling with interaction complexity in an ergonomic way. I know you believe you can prove that wrong. Honestly the biggest obstacle here is similar to what like Marko faced when first trying to push MPAs. The paradigm is different enough people don't see it. I think you just got to keep showing it. I am definitely interested in having you on my stream in the future.. Usually it follows where my focus is. But I imagine I will be swinging back that way coming up.

Glad to see other people are staying to do independent tests on it. We won every single category and that's even in PHP! How is this not the norm for web dev? zweiundeins.gmbh/en/methodology…