Cal Lavicka

"Strong preference for software that reduces chaos instead of adding more of it" everyone should strive for this

@aarondfrancis @dillon_mulroy It’s really revealing to me who actually cared about code quality before

The only reason I won't stop reviewing the agent's code is that I've reviewed the agent's code.

I've been seeing a lot of folk on the timeline say IDEs are dying. But I don't get it. Are you guys not reviewing the code your agents generate 😅

@dillon_mulroy I feel like i'm taking crazy pills. People are seriously merging this shit into production without looking at it?

s/4.5/4.6/

@_developit People truly be saying coding is solved and yet we can't get basic focus management right

Has anyone else noticed mobile frontends (native and web) are getting increasingly terrible (as we descend into UX vibe coding?)? I keep noticing I've developed ridiculous coping strategies like constantly copying input to clipboard in case it gets lost when sending.

@mitsuhiko nah opus 5 will fix everything trust me just one more model

The biggest issue for me with agents is that they are hard to resist. But then you can build yourself some shit into the codebase that you get to regret in record time. And no, I don't think you can vibe yourself back to sanity with better models.

When you give every engineer a magic button, it separates the ones who truly care about the craft and the ones who were never going to succeed anyways

@alvinsng pnpm i solid-js

There certainly are times that they speed me up, but I am scared of what our codebase might have looked like if I didn't keep such a close eye on the code outputs

@RhysSullivan I feel like MCPs got a bad rap because the protocol itself was poorly designed and the initial examples were terrible

the overindexing on CLIs is kind of insane to me it's building a primitive that's not portable, properly discoverable, has no good approval flow DCR / CIMD to APIs would go so much further but CLI is just the current hype thing

Why does Claude love O(n^3) algorithms

Wouldn't it be easier to review the before/after of a UI component than the underlying code? 🔜

The <Suspense> is over. Solid 2.0 Beta is now released (next tag on npm). 🎉 github.com/solidjs/solid/…

@ericzakariasson Tab will always have my heart, but can't deny the shift

we are in the middle of a major shift of how software gets produced today: 1/3 of PRs are from cloud tomorrow: 1/3 of PRs from local extrapolate

@ritakozlov Gotta start taking the train then

all the best launches are always happening when i'm in an uber to jfk. don't they know i get car sick

@RhysSullivan Keeping it synced via webhooks seems hard, last I checked there were a bunch of github features that don't trigger webhooks

some notes on it: - you can jump back to github at any point through the cmd + k menu - logs support copying the CLI command to give to your agent, along with copying them with header data about run id etc so it has context - it works via installing a github app and syncing all of your data to convex then keeping it live through webhooks - the ui i'm still not super happy with, there's a bit of jank between showing public repos on the homepage and personalized ones - you can use [ and ] to toggle the sidebar on the PR page to get a larger view - repo ingestion takes a little bit of time but once thats done its fast forever - you can run this locally with bun dev:web or bun dev, 0 config is needed putting this one out a little earlier than i normally would, there's some prefetching stuff i need to solve with the ui to make it truly instant but that'll come in the next few days, wanted to get it out early and improve on it

i had opus credits to burn so i recreated github with instant navigation and the ui that i would want you can use it today fastergh.com, most of your workflow should work on it, private repos are supported, and it's open source

@ryanvogel @thdxr nice sweatshirt

@thdxr this is true I’m sitting next to you

i'm rawdogging this flight (letting codex run while i stare at the spinners)

@JosephD @jaredpalmer Oh definitely. I still think there’s work to be done to make them understand the refactors I want to do better, but with the current architecture they’re always going to struggle on their own.

They can't be good at refactors by design, they're working with partial information no matter what. Even humans aren't leveraging all the knowledge yet and instead relying on whats in their head. The AI can't possibly see at once all of the plumbing and understand how to safely keep that together. It will always be guess and check by the current methods.

It feels like current generation of coding agents are overfit to make code run (and render) successfully on every prompt. Too often, they emit “fallback” behavior and avoid breaking changes as a reward hack, sometimes not even telling you about it.