Rasmus Schultz

@mindplay.dk

Passionate web-developer since 1998 using various languages, databases and tools.

I opened X for some reason lol the current experience is something like scroll, block, scroll, scroll, block, scroll, block, omg block, block, close this isn't social media anymore - there is nothing social about it it's a monument to human stupidity on the internet 🪦🥀

raw SQL in your frontend 🙀 electric.ax real talk: for some reason we feel better about an ORM with a REST or GraphQL frontend doing essentially the same thing with row level security 🤷‍♂️ is this rejection of the idea of raw SQL in the frontend mostly rational or mostly instinctive culture shock? 🤔

Electric | Agents on sync

Electric provides the data primitives and infra to build collaborative, multi-agent systems. Including Postgres Sync, Durable Streams, TanStack DB and PGlite.

electric.ax

bsky is flooding with MiniMax H3 release posts, and not a single post gets a single like... while X is having a total meltdown 😂 what I want to know is: hailuoai.video/generate/ai-... did this model just perfectly recall a 1982 hit song? bc that would point to a rather serious problem, right? 😶

AI Face: Born in Any Country Simulation – AI Generated Video | Hailuo AI

Generate & Play Hailuo AI video:Upload a photo, enter a country, and AI generates your simulated look as if born there.

hailuoai.video

just an hour since release, and Kimi K3 is already up on Fireworks openrouter.ai/moonshotai/k... 4 times lower latency, twice as fast looks like they still have some stability problems, but this model will be everywhere in a day or two 🙂 Claude Opus on any plan is probably cheaper now though? 😏

Kimi K3 - API Pricing & Benchmarks

Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. $3 per million input tokens, $15 per million output tokens. 1,048,576 token context window. Higher uptime with 3 pr...

openrouter.ai

tfw you spent a day making a meaningful contribution to an open source project, no one replies for two weeks, and then the author deletes his entire GitHub account 🥲

it's amazing how companies manage to make development work anything between awesome fun, and unbearable mindnumbing bullshit I've been at both ends (and I'm thankfully at the good end now) I honestly don't think the ones who are really successful are also the ones who make the work unbearable? 🤔

have you ever needed to contribute to a repo that is clearly 100% vibe coded slop? after about two hours, you give up and let your agent produce more slop to shoehorn in the change, right? it's not like the owner is going to notice or care "machine-only codebases" open source is getting weird 😳

having a bot auto-close issues without comment is the strongest signal you can send that you're not serious about software development, at all. makes your backlog look good at first glance tho. 🥲

Released dprint 0.55—a pluggable and configurable code formatting platform that unifies all your formatters behind a single command. This substantial release lets you specify formatting plugins from npm.

A dprint.json configuration file with npm specifiers in the plugins array.

you come across an interesting paper or article you're not going to read do you (A) save it to your bookmarks or reading list that you never open or (B) leave an open tab in the background for 3 months before closing all the tabs 🤔

I looked at 8 alternative NPM packages today 6 were slop - of those, 2 didn't work and 2 were vibed in a day and abandoned 2 were clearly AI but usable, but I already found critical bugs in both imagine NPM two years from now, you might as well not install anything and just generate your own slop

please calm yourselves today the 50% figure is a CEO interview claim there's no verified benchmark, the chip isn't deployed, no technical details or system level data or anything tangible has been released it's less an announcement and more just good old fashioned market manipulation

does this feel weirdly passive-aggressive to anyone else? "your issues are not important unless we say they are" or at least like a weirdly broken github workflow? making it *look* like fewer issues 🤨 just tag issues as "new" by default and close them if you don't think they're worth while 🤷‍♂️

Bild

did they lobotomize Claude Sonnet 4.6 this week? I'm getting the most hilariously wrong answers these past few days this model has been fine for almost everything I need, and now, just days before they launch their new flagship model, it seems to have turned completely effin dumb coincidence? 🤨

"make a scrolling div where the top/bottom edges fade into the background" even Claude Opus with extended thinking couldn't get this, no matter how many attempts and how much guidance guess this isn't in the training data 🤷‍♂️ we're not going unemployed just yet, folks jsfiddle.net/pLd1ao0z/1/

scrolling div with fade in/out at edges - JSFiddle - Code Playground

JSFiddle - Test your JavaScript, CSS, HTML or CoffeeScript online with JSFiddle.

jsfiddle.net

this is nonsense github.com/vercel-labs/... you don't need to build an entirely new language that is "optimized for agents" you could have added JSON output to the frontend of any existing compiler and you can't "optimize a language for LLMs" - they weren't trained on your new language duh? 🤷‍♂️

GitHub - vercel-labs/zero: The programming language for agents

The programming language for agents. Contribute to vercel-labs/zero development by creating an account on GitHub.

github.com