𝗧𝗢𝗠 𝘈𝘺𝘭𝘰𝘵𝘵

When you lift a train over your head, normal people don't think you're strong; they think the train is made of cardboard and you are trying to trick them.

@thsottiaux Let me transition a ChatGPT Pro chat thread to Codex Make it easier to mine my thread history for stuff I’ve said in one Thread and then forgot to record anywhere else. Maybe make it easier to browse/search/organize my past posts. Maybe I should just make a plugin for it

What is something that you feel is surprising that Codex still can't do well and we should have gotten right a while ago?

it's all his fault

I blame seg

@wesbos we elder millenials has seen the rise and fall of more technologies than all past AND FUTURE generations combined our parents saw human programming get invented we have seen its end but just like VHS, some of us still love it for nastalgia

@SubtleGradient first in my bloodline to understand these words

Give me your hottest takes on generative UI, MCP apps, agents returning widgets, and chat as the final interface Working on a talk for JS Nation and I want to spotlight some differing opinions

@wesbos instead of ~1987 era desktop app or a ~1997 era web site or a ~2007 era mobile app or a ~2017 era web app what if 2027 had retro futuristic ~1977 era CLI/TUI apps in chat and 2037 had them in voice + AR holograms

@wesbos chat is nearly the O'st of Gs UX pattern we had bash, zork, and BBSs before desktop windows and menus AI chat is like the terminal, but mainstream. ChatGPT was like bash * zork. Gen UI in chat is like TUI apps in bash. a BBS is a remote TUI app. That's a HUGE opportunity

@dhh YES !! bring back cartidges

Also, software on cartridge is the great constraint. We can't just fix it in post! It's gotta be perfect before it's romified.

I just adore how quickly @PalmerLuckey's modretro Gameboy goes from off to in the game! No updates, no boot lag. Just instant into the greatest game of all time (Tetris).

@jarredsumner bun made me fall in love with Zig but Zig is like JavaScript and Rust is like TypeScript for long term super serious business stuff, I'd probably use Rust / TypeScript for whimsical toys and wacky side quests I'm on Zig and ESM JS (with ts jsdocs) bun is Pro™ software now

The people dunking on Bun’s Rust port are having a hard time coming up with a strong technical argument.

@mutewinter ya, Linear is held up on a pedestal like the end all and be all of what web apps can be. it's pretty great, but it seems Safari is not getting the love personally I'm discovering that task oriented project management is not really the right abstraction for me

Do these people actually use Linear or just enjoy fantasizing about it? It’s a mess in Safari. Multi second delays, failed syncing, broken drag states.

@ChmalSzymon that'd be pretty hip

What if React Native handled threading like Lynx does: keeping UI operations on the main thread and business logic on a background thread? 🤔

what is LazyGit was a macOS app it is now I made a thing that lets me double click whatever.lazygit folders to open its parent in lazygit in LazyGit macOS app

@tldraw and yeah, QuickLook works

what if tldraw was like SVG? resurrected my 15 year old Appify-UI project and built TLCanvas app that manages .tlcanvas bundles (folders full of tldraw canvas stuff) immabouta use tldraw for ALL the things. each of the ones actually. we love @tldraw

@dhh this is the thesis to my family's game studio the retro collectible games of 2047 in 2027

Half the joy of collecting retro gaming is the box art, the cartridges, the artifacts. We've lost something important giving up on those physical manifestations of software. Comeback?

@kitlangton SHAMELESS PLUG: @ effect-native/tui-testing-library@beta may or may not be helpful to you :D

Here's a little GitHub TUI for managing PRs. bunx @kitlangton/ghui github.com/kitlangton/ghui

@dhh YES !! my favorite nerds are all coming together

What a delightful TUI! Next version of Omarchy will ship it as a lazy-installed default. So you can just type ghui and it'll get pulled, just like we do with opencode, codex, etc. github.com/basecamp/omarc…

@VictorTaelin @SteveMoraco or anti-entropy. you know what I mean :P

@VictorTaelin @SteveMoraco "token golfing" has helped me with this a bit especially the Binary Cognitive Mechanics flavor that uses a GAN-like loop to iteratively attack & refine the compressed info such that signal entropy is maximally retained across all cognitive dimensions, provably

seriously, working with AI is MISERABLE for one and only one reason: having to re-explain the same thing "oh yeah this new session obviously doesn't know what proper case trees are, so let me explain it for the 5000th time in my life" I'm tired AGENTS.md doesn't solve this because it is impossible to fit the entire domain knowledge without nuking the context - it would be 1m+ tokens worth RAGs don't solve this, the agent won't search unknown unknowns SKILLs don't solve this unless I keep like a collection of 1750 skills with specific cuts of domain knowledge for each possible subset of my domain that I might need in a given chat, but that's a lot of manual work recursive LLMs or whatever don't solve this for the same reason, you can't dump a domain book and expect the AGENT will magically guess that it is supposed to search for a specific bit knowledge. unknown unknowns fine tuning doesn't solve this (OSS models suck and OpenAI / Anthropic gave up on user fine tuning) I honestly think a good product around fine tuning on your domain would be a major hit and an underdog lab should take this opportunity

@dceddia @VictorTaelin I've been toying around with various kinds of fractal decomposition such that relevant complexity is discoverable in context just-in-time, but I'm tired boss

Maybe progressive disclosure would help with this? I'm not sure what shape it would take, or if agents would do well with it, just spitballing here. I'm imagining like, an LOD tree for knowledge. And maybe this is just SKILLs with more hierarchy? e.g knowledge.md Don't use BigInt - see [bigint/why.md] bigint/why.md We don't use BigInt in this project because short description. When BigInt seems like the right solution see [bigint/but-why.md] But like you say, the hard part is a) creating this thing in the first place b) keeping it up to date c) at some point knowledge.md will still get too big Agents can probably do (a) and (b) but idk how well they'd actually do it and (c) is still eventually an issue. I guess it's kinda just handrolled compression at the end of the day. Maybe train the model on .zip-formatted data and we can just upload our compressed markdown blobs 😂

@VictorTaelin unfortunately your personal cognitive mechanics * "giftedness" is sufficently atypical to ensure permanent frustration with any toolset capable of becoming generally popular you're going to have to solve it for yourself only