aaron

design leader for AI products // head of design at AE Studio // building agent systems, behavioral QA, and strange interfaces

[aaron @aa_on_ai]

“State of the Art” by Jim James came out in 2013 and somehow feels considerably more about 2026

not anti-technology exactly... more like - use the machine, don’t let the machine become the life

https://t.co/HNhpYHYpKV

[aaron @aa_on_ai]

source on the hugging face incident https://t.co/GJLkJu4tc5

visual made with the tokens tool from @BrettFromDJ's https://t.co/L3qoSw07tg

[aaron @aa_on_ai]

after openai’s agents hacked hugging face, agentic trust is the talk of the town

the agents I trust know which parts of the agency are still mine

and the systems around them assume that trust can fail

[aaron @aa_on_ai]

@designertom you don’t need a traditional background in alignment/safety to work on solutions to this

we’re hiring at @AEStudioLA to explore neglected approaches to ai alignment: https://t.co/sB96Xa77hQ

[Anshu @anshuc]

Extremely proud of this article I wrote for the legendary @lennysan! This is the culmination of months of research into why most AI designs are slop, and my top techniques for fixing this.

LLMs can be super creative, but as we post-train them more and more, we stifle that creativity. We train them to predict tokens that balance everyone’s preferences. But good design does the opposite: it’s risky, bold, and opinionated. You have to put your LLM in a different state of mind to get this kind of design from it. You have to break out of the average slop, into the fringes of the bell curve where unseen ideas live.

This post is all about how to do that. Hope it helps, and let me know if I should go deeper on any of this for future posts!

https://t.co/VyCzxmQ3qc

[aaron @aa_on_ai]

electrochemistry [easy: success] - give the disco ball physics

https://t.co/VfeG0PINHT

[aaron @aa_on_ai]

1. liquid gooey by @jakubantalik
https://t.co/1n5N80k0DU

2. liquid glass by @samasante
https://t.co/o4Ouu9fB9d

3. dither-kit by @grimcodes
https://t.co/ecn8QN7NIA

4. dialkit by @joshpuckett
https://t.co/0FH88TfamJ

5. morphicons by @guillermolg00
https://t.co/Dxzr0EWdxi

[aaron @aa_on_ai]

I keep a running pile of weird little tools that make building on the web more fun

here are 5 worth trying

[aaron @aa_on_ai]

agents are changing my design work in a way I didn’t expect

I’m designing fewer task flows and spending more time deciding what software is allowed to do for people

the interface doesn’t disappear... it becomes where people set intent, judge the result, and take back control

[aaron @aa_on_ai]

source material

openai’s black hat talk, where the story starts
https://t.co/0uooRISuxr

hugging face’s technical timeline
https://t.co/HTWtGnEvha

zvi’s interpretation
https://t.co/gnpH4ZeIG7

clawbotomy, where I want to test the room next
https://t.co/U1u2CZPRMp

[aaron @aa_on_ai]

we keep interviewing the minds... it may be time to interview the building https://t.co/fE0qECsN6T

[aaron @aa_on_ai]

@drewwilson @valiocon yes please... would love to come hang in san diego 🤞

[aaron @aa_on_ai]

@designertom sorry! just thought you went IN on the computer too for a second there

[aaron @aa_on_ai]

I rebuilt agentic design system into a repo-local skill pack for coding agents

a ui brief becomes a browser-tested repair, evidence, and a separate grader verdict before the agent calls it done

open source and live https://t.co/pVg1eI1CUJ

[aaron @aa_on_ai]

spent yesterday turning a pet project into an authored 3d game instead of a tech demo

the funniest bug was every object having a real hit zone + a secret offset ghost hitbox

very 90s adventure game energy... point-and-click UX is affordance archaeology

still very WIP

[Peter Gostev @petergostev]

Sometimes you need to recognise if you are in an abusive relationship

[aaron @aa_on_ai]

really impressive work... and great for AE

I built https://t.co/C0KHo3ufd6 for it too

the reverse lobotomy part has me thinking... if knowledge can live in removable modules, could bad habits too? could models/agents stay more targeted instead of learning everything?

[Anthropic @AnthropicAI]

We’re pleased to have collaborated with AE Studio on this research.

Read more here: https://t.co/9exnFltK7A

[aaron @aa_on_ai]

@kylezantos so simple to use, great job!

[aaron @aa_on_ai]

@designertom @openclaw hermes was way on top of it ⚡️