programmarchy

Nice work. The attention to detail is insane. Love it. One time purchase is a hell of a deal... I bought.

Couple feedback points. Some of the sound effects are cool. The sound for the Tabs component was a little grating to me, personally.

I initially checked out via Google, but got weirded out by the strange app name (dcfmseuukjnydbmzxwhi.supabase.co). Probably a Supabase thing, but I backtracked to email authentication.

Welcome to the OCD lounge. I also put my cables into bags, which helps me sleep better at night.

Yep, worked on me...

Really needs a 20MB video playing in the background of the hero.

Might have been Ken Silverman. https://advsys.net/ken

Good luck with that. I used to be an OCD freak about code before LLMs, but AI coding has largely freed me of that limitation. I've become very comfortable giving AI a long leash, but only after being meticulous about curating the context.

These days I spend most of the day in discussions and planning, producing documentation, agonizing over architectural decisions, edge cases, and naming conventions. Once that's all settled I'll hand off implementation work to run overnight. In the morning, I'll review and fix, but I'm usually pleasantly surprised with the results.

One pitfall is long leash without a curated context, which is more like "slot machine" coding. Usually not effective, and may have addictive effects since it does occasionally work.

To spice things up lately, I've been encouraging the model to produce its own "capstone" -- a feature it decides to build on its own, however it wishes, with the tools at its disposal. So far it's been conservative, creating useful tools for development rather than customer facing features, but I'm curious to dial up the temperature to see what it might come up with.

I haven’t used Effect but the problem I see with using it is that it seems to want to completely swallow the whole app architecture. At that point, why not just use a functional language?

Railroads had enormous subsidies, too. This is how infrastructure is built, even in "capitalist" economies because it operates at the level of national security. Even to this day, passenger rail is not profitable, although freight is very profitable. So it wouldn't be surprising if "passenger" AI remains unprofitable, while "freight" AI becomes very profitable.

Pretty bold statement to say it's useless for most people outside of tech. Almost every "normal" person I know including my in-laws are using it regularly. It's becoming the go-to for asking questions rather than Google, Bing, etc.

And the privacy battle was lost 25 years ago. People don't really care if corporations know about their search history (Google), or their private lives (Facebook). You're beating a dead horse there.

AI might actually RTFM

Also had a severe issue with Bunny, serving videos. They had a cache poisoning issue where they served a few frames from pornographic videos right in the middle of our educational videos. They did not have the multi-tenancy thing fully figured out, and it became a nightmare scenario. After that, we moved to a provider that explicitly did not allow porn.

I really dislike how apps add themselves to the menubar. And I hate if there's no option to remove it from the menubar. Icons with indicators like WiFi, Battery, etc. make sense. But if an app does not need an indicator like that, just add the capabilities to the Dock icon!

Then what’s the point of skills like apple-reminders? Isn’t the implication for a personal assistant styled OpenClaw setup that you allow it access to those tools on your behalf? Otherwise where is the benefit?

FastAPI -> OpenAPI -> openapi-typescript

Academia is more than a transaction. It's a social good. One that's also subsidized by the US taxpayer.

Compile step makes things more complicated.

I spent some time attempting to "derive" a theme given a primary and secondary color, but realized my color theory wasn't strong enough to build something reliable (I tried with both hsl and oklch). Curious if that's really possible.

PDF is arguably a confusing format for LLMs to read.

Exactly what came to mind for me as well. Information is a difference that makes a difference.

An analogy

Statement: Statistically, seatbelts reduce the chances you’ll die in a car accident.

You: But, what if your car crashes into a lake and you get trapped underwater?

This proves the GPs point about saving energy. Think about how inefficient it would be for each individual to create their own transcript via an LLM.

Would it be possible to embed Bitrig in an existing app? Similar to how you might embed React Native into a portion of your app to accelerate development in a specific area, it’d be awesome to preserve the mature areas of our apps and use Bitrig to quickly iterate on new features.

One of the core things Node.js got right was streams. (Anyone remember substack’s presentation “Thinking in streams”?) It’s good to see them continue to push that forward.

This sounds like a breath of fresh air as a disenchanted Spotify user. My only hesitation is that I’ve lost touch with collecting music. I used to rip CDs and download music and curate a library etc, but I’ve lost my collection and collecting habits since adopting streaming. How do people collect music nowadays? Is there a legit way (fairly compensating artists) to do it?

For most APIs that doesn’t deliver any value which can’t be gained from API docs, so it’s hard to justify. However, these days it could be very useful if you want an AI to be able to navigate your API. But MCP has the spotlight now.

With 5M you can get white hat audits. Even big boys like Okta have had serious fuckups [1].

[1] https://trust.okta.com/security-advisories/okta-ad-ldap-dele...

shadcn distribution model for utils is a good idea. i wanted something for react hooks as well and was surprised that didn’t seem to exist either.

Wow, this looks awesome. Been using Temporal, but this fits so perfectly into my stack (Postgres, Pydantic), and the first-class support for DAG workflows is chef's kiss. Going to take a stab at porting over some of my workflows.

Seems like you should be correct. A shadcn button is just react, tailwind, and @radix/react-slot. But if you simply create a new shadcn Next.js template (i.e. pnpm dlx shadcn@latest init) and add a button, the "First Load JS" is ~100kB. Maybe you could blame that on Next.js bloat and we should also compare it to a Vite setup, but it's still surprising.

With OpenAI models, my understanding is that token output is restricted so that each next token must conform to the specified grammar (ie json schema) so you’re guaranteed to get either a function call or an error.

Edit: per simonw’s sibling comment, ollama also has this feature.

Okay, that makes sense. I thought they could just log in to a dummy site, not that it was proxying requests through to a real site. Yikes.