Micah Engle-Eshleman

@hamza_q_ Wow! Would be neat to see this integrated into WhisperX. AFAIK it’s still using Pyannote

What other applications of diarization are you now excited for being enabled thanks to much faster speed? Interesting applications in video editing come to mind, for example. Cheers everyone; enjoy fast diarization!

Senko is now 3x faster on Mac thanks to CoreML. It now processes 1 hour of audio in ~8 seconds on an M3 MacBook, making it by far the the fastest open-src speaker diarization pipeline on Mac, all while maintaining good accuracy.

🎉 Preact 11 beta is here! After years of thoughtful development, we're delivering major improvements with minimal breaking changes. Better hydration, forward refs by default, and performance upgrades - all while maintaining the stability you love about Preact X.

@dlnieves @thmsmlr Cloudflare tunnels are great for exposing local resources as public subdomains. No static IP required

@thmsmlr What is your advice on ISP and IP changes?

Don’t let them scare you out of self-hosting You want at GPU at home?! RIP your electricity bill. Residential Electricity cost: $0.2/kWh Max 4090 power draw: 450W Cloud 4090 GPU: >$0.7/h -.-

PWAs can capture links, but support is fragmented across OSes and platforms. I tested all platforms and wrote about my findings in my new article "The state of link capturing in PWAs" Read if for free on Medium 👇

Chuck Palahniuk, oof

Warnings are beautiful. Vike has 183 of them ✨ For all kinds of things, ranging from small paper cuts to important performance issues.

Really enjoyed this article on how Notion is using SQlite in the browser and the challenge of supporting multiple tabs. This stuff is tricky, cool to see it happening at scale! Notion article: notion.com/blog/how-we-sp… Relevant Github Discussion: github.com/rhashimoto/wa-…

@evisdrenova @thorstenball Funny you say that, I've thought of trying to implement the same thing with Chrome's built-in LLM to help explain API errors to users that can be a bit cryptic (e.g. Pydantic model validation failures). Haven't tried it yet but I bet it would work

I played around with the idea of a having an AI layer in the error propagation path that tells the user what happened, and how to fix it. Essentially take error + app context and feed it into an LLM and rerun to the user. That “how to fix it” is the biggest missing piece right now. But it’s really hard to get the UX and context right.

I truly believe that this is one of the hardest problems in software: How do you bubble up errors in a system so that the user can see them, make sense of them, and knows what to do? It's so hard and I've bumped into it in nearly every project I worked.

@aboodman @zero__ms Fair point. I've had to do weird things like storing / comparing Last-Modified header values on cached responses before using them. Looking forward to using this in my app sometime.

I honestly think API caching just doesn't make a ton of sense for lots of applications. As soon as you start caching you immediately have consistency problems – the cache responses are stale. And it's a perfect zero-sum game. The more you turn the caching knob to improve perf, the worse consistency gets. Instead of cache, _replicate_. Sync engines replicate data continuously to the client. It's as if the TTL is infinite AND the data is always fresh. And there isn't increased load on your db. In fact there is _reduced_ load because the sync server acts as a read replica.

Zero 0.13: Horizontal scalability, zero-downtime restarts, and SST support Complete details at zero.rocicorp.dev/docs/release-n…

@aboodman @zero__ms Makes sense, thanks! Sounds like you loose out on being able to reuse public cached-at-CDN-layer API responses between multiple users, but existing users will be fetching less data overall.

In a synced app, the app itself is typically a static shell. This caches very well. The data is entirely separate and is synced to the client. The client stores tens of MB of data locally, so you can think of that as a data cache. In Zero, there's also a separate server called zero-cache which acts as a network cache of the data too.

@tannerlinsley @irshathv2 Sweet! Really looking forward to this (only pre-rendering). Blocked by Vinxi bugs currently but sounds like that dependency will be going away as well?

@irshathv2 SPA mode for start is coming. You'll be able to turn off everything and opt-in to only SSR, only server functions or only prerendering (among other things)

It's 2025 and the React docs still don't show Vite + vite-react as a 1st class valid target despite an ocean of developers still using CRA for SPAs likely never caring in the slightest what blessed fullstack meta framework is the "idiomatic" way to use react.

Excited to finally start self-hosting a GPU server for Adblock Podcast. Finished the new “rack” today!

Why do desktop apps think this is OK? If I launched an app, it means I need to do something with that app. It is _the_ worst time for you to update. It's been updating for over a minute. Forget it, I'll just use the website. This is another reason why the web wins.

This story about SpaceX engineers transporting a rocket from Texas to Florida is insanely hardcore

Shout out to hugeicons.com, great source of free SVG icons. I used to use heroicons.com from the Tailwind team, but keep finding myself swapping them out for the hugeicons equivalents

I have great news! I was wrong on the Internet! In my Jan 2023 cargo airship post, I wrongly assumed airships would be limited to trucking-level per-ton-km revenues. It turns out airships can actually BEAT 747 freight service times. This means revenues and margins can be MUCH higher. @JimCoutre figured this out and has started @ShipByAirship to usher in the new transoceanic cargo airship era. I'm thrilled to be a part of this journey not only as a supportive poaster but also as a pre-seed investor. Airship Industries will deliver cargo faster than today's air freight service with 75 percent fewer emissions and with far fewer opportunities for theft and delay. Please join me in congratulating Jim on the launch and in celebrating our glorious airship-filled future. elidourado.com/p/airship-indu…

@HookForm If you'd doing conditional logic with RHF, try it out here! Github README: #conditional-logic-for-react-hook-forms" target="_blank" rel="nofollow noopener">github.com/micahjon/rhf-c… NPM: npmjs.com/package/rhf-co…

Just dropped a new version of React Hook Forms conditional logic hooks with better type-narrowing & autocompletion in useCondition() rhf-conditional-logic v0.1.2 @HookForm

I’ve been using Claude AI daily for a couple weeks now and I continue to be impressed. Such a great conversation partner about solar, wine making, Typescript, and all sorts of other things I need guidance on