thread of things I made to make it easier to share w ppl I meet instead of having to dig thru my tweets 👇.

What do you want to know?.

RT @ChShersh: Life goals is making a useful project to earn enough money so that I can finally work on useless projects.

was going to @ several ppl for their thoughts but i already have too many projects to dream about!.

bearish about execution via text prompts but the concept is really compelling, reminds me of how "single continuous shot" is the default for Line Rider. wonder if we can build a scaffold and have genAI fill in the details (on second thought, the scaffold might be enough).

well if that aint oddly satisfying. i think i can create a system that generates a bunch of animations of this category, brb reviewing 2d symmetry groups.

ok sure this direction could work

one of these days i will make something that warrants this type of UI. probably a DAW.

i watched a presentation on how to architect LARGE software projects and now i have a mighty urge to architect a LARGE software project

you shouldn't let poets (LLMs, demagogues, your own imagination) lie to you.

arc-agi was designed with these so-called Core Knowledge priors:. - Numbers and Counting, Basic Geometry and Topology: "draw some circles".- Objectness, Goal-directedness: "draw the rest of the fucking owl". alternatively: objects aren't real! teleology is a myth!.

what do you call the type of programmer whose performance 10xs when they hit the ballmer peak.

how many programmer-hours in the world have been lost to this type of bug 😑

funny enough, adding indirection helped (and also clarifies how I should modify this graph algorithm). now we're at the limit of comprehensibility. there are some "parallel" sets of relations (with real meaning) that you can kinda identity via literal parallel arrows

I think a barrier to understanding things that deal with a lot of relations (like category theory) is how arrows quickly get out of hand and become incomprehensible

has anyone vibe coded a programming language yet.

RT @poetengineer__: everyone's wired different. but its hard for me to imagine ever running out of ideas. if anything, i sometimes slow dow….

from this view, AI existential risk will seem overblown. we already know how humans and human orgs work, just apply the same principles to keep them in check.

*it's infeasible to replace reality with simulation. simulation will always have a tradeoff between accuracy, generality, and efficiency, and won't ever be able to predict the future at a hi-res as that requires very hi-res present information.

calling it now: AGI and ASI will be human-shaped, and seeing them as unfathomable is holding us back. like humans, AGI will need to be continuously learning and interfacing with reality*. and scaling "intelligence" already has a solution: see human orgs. it then boils down to a.