elbo

I just published gatsby-remark-embedded-codesanbox, a @gatsbyjs plug-in to embed @codesandboxapp in your markdown given a local folder of files: github.com/elboman/gatsby… 🎉🎉

@sudoingX I actually have issues with OpenCode complaining about the jinja template (running through LM Studio). Even updated to latest version and getting the same problem

let me save you 3 hours of head scratching. if you're running local models like Qwen3.5-35B-A3B through Claude Code via llama.cpp's Anthropic endpoint, the chain will break every 3 to 5 minutes. tool call fails. flow stops. you reprompt. it recovers. 2 minutes later it stops again. the model is fine. the harness chokes on local inference latency. switch to OpenCode. same localhost endpoint. same model. same GPU. the chain doesn't break. the tradeoff: OpenCode sometimes loops. the model forgets what it already read and repeats the same tool call. but a loop you can interrupt. a broken chain kills your momentum and you start over. watch both side by side. proprietary agent vs open source agent. same 3B model. different failure modes. pick your poison.

@GregFeingold I just ordered my, looking forward to it! Someone already started a custom firmware: github.com/crosspoint-rea… Might be interesting to join forces there!

I got that mini e-reader that everyone was talking about! I love how small and light it is, fits perfectly with my iPhone Air. The software is super bare bones and DIY. I’m going to use Claude Code to make a new OS for it. To be continued!

Missing that `Copy as markdown` button? I was, so I shipped a fix this weekend: Just type ✨ i n t o . m d / ✨ in front of any URL. Get clean, LLM-friendly markdown for any website or docs page.

@trueadm 🔥🔥🔥

I've had so many folks ask now, that I had to go for it. You asked me to build that ideal Inferno X framework that I described ages ago, so I went ahead and did just that. I built something that actually works and, well, it works well. Oh, and doesn't use signals! I mean it's still less than a week of hacking it together, but it gives me Inferno-esque vibes of when an underdog could come in and change things up. DX wise, IMO it's in its own category. I'm hoping to get my npm issues sorted out soon so I can publish the super early alpha (do not use) version so you can at least try it out. Watch this space. :)

No apple maps, I literally never want to copy the empty input

@CompuIves Wow congrats! 🥳

I have some personal news, times two! I recently proposed… and now we’re engaged! AND we’re going to move to San Francisco in 2025! It’s a huge change, but I’m so thrilled!! Really I can’t wait!! 🎉🎉🎉

@AntonK52 Is the noise cancelling any good? Tempted to go from pro 1 to normal 4. Also, how’s the transparency mode?

A few thoughts on the AirPods 4 - still fall out easily - trying out noise cancelation without a rubber tip is more impressive - there’s no volume control on the buds like on the pro ones

@AntonK52 yeah I figured this is what coolify does too 😅

@elboman github action on push main ssh host cd dir git pull

stupid question but I need to ask at this point: for people deploying laravel on vps, do you just ssh and git pull? am I missing something?

oh ok @coolifyio is the answer 😅

@AntonK52 a decent combo of intense hacking/prototyping 0→1 and then remote work for long term/clear planned work is the 🐐 imho

Im all up for remote work but man it is easier to iterate on stuff when people are in the same room

@_boraturan @JNYBGR Haven’t test it yet, caniuse shows partial support for video, so we might be lucky. Will let you know!

@elboman @JNYBGR does it work on Safari mobile?

There are dozens of online MP4 to WebM converters And all of them hook up to server and convert on the backend But i am now 100% sure it can be done in the browser using WebCodecs: faster, private, offline Nobody’s been crazy enough to build it.. Someone has to do it, right?

@PhilippSpiess Congrats 🙌

some personal news 👀 spiess.dev/blog/im-joinin…

@farid_ahamat @GergelyOrosz you can measure value if the test is directly coupled with revenue stream and covering transient/e2e flows that are hard to catch otherwise. i.e. revenue loss prevention tests are valuable and somewhat direct to measure

@GergelyOrosz How do u measure cost and value for tests? I assume cost can be measured by dev time, but how do you measure a test's value?

Every. Single. Time. When we built end-to-end tests (or close to these) They were so taxing to maintain. And every time we sat down to ask “what value are we getting, versus the cost?” it turned out the value wasn’t that much. Worth reflecting how valuable your E2E ones are.

compression and speed is good and avc is supported basically everywhere. I’m just surprised there’s no support for hvec or av1 in any major browser

ok I finally got the video encoding working (and therefore compression) with h264 powered by webcodecs, all hardware accelerated and native

@AntonK52 I’m sure the Venn diagram of the two groups has a sizeable union

configuring vim is less geeky than this

@taylorotwell exciting! but unfortunate you had your finger in the corner

Coming soon... 👀

btw one cool thing I found gpt useful for it actually to generate d.ts of JSDoc types for a given raw JS code. I wonder if a bot + llm could be put together to contribute type defs for raw JS libraries with minimal manual work or cleanup

how do people handle working with big libraries that are untyped? I find I stumble upon many and it's tough to manage