James Yang
26 posts

James Yang
@jamyanges
Saving Tokens @Brevitas_sys || Founder @fdotinc || 4x Hackathon winner || Serial Founder
San Francisco Katılım Mart 2022
38 Takip Edilen27 Takipçiler

if you have less than 10 coding agents open at a time, you're ngmi
introducing Gridbash, a way to spawn CLI agents in a clean & manageable grid
> use many claude/codex accounts without logging out
> deploy manager agents that autoprompt for you (like /goal)
> organize your terminals in grids & groups
> see all your terminals at once to save human context
> prompt as many terminals at once, good for sending "continue" lol
I've used this in pretty much all of my hackathons & decided to open-source it, to which it has gained 50+ stars (yay!)
tokenmaxx with it now:
`npm install gridbash` or github.com/jasonsuhari/gr…
English

hi im james
- just graduated high school(skipped grad)
- math + bba at @UWaterloo
- building brevitas @fdotinc rn
- saw a cool robotic fight the other day
- down in sf till august 10th
- wanna meet any people down here hmu
(Random parrot guy I met at the robot fight)

English

@jamyanges bro u didn't make it in the 101. or 102 for that matter
English

Wow..
Ju Lin@urjr1
Just tested Samsung’s tiny 7M TRM model in Clusy. The claim is pretty wild: it can score very high on ARC-AGI-1, even against much larger models. So I ran the actual eval instead of just reading takes about it. Clusy pulled the repo, made the plan, loaded the checkpoint, ran the benchmark, showed solved ARC grids, then tested the obvious question: is it really reasoning, or is it leaning on puzzle IDs / recursion / augmentation voting? Replay here: app.clusy.io/share/bA9seSDW… This is the kind of thing I want research workflows to become: run it, inspect it, challenge it. Try it at clusy.io now.
QST

Just tested Samsung’s tiny 7M TRM model in Clusy.
The claim is pretty wild: it can score very high on ARC-AGI-1, even against much larger models.
So I ran the actual eval instead of just reading takes about it.
Clusy pulled the repo, made the plan, loaded the checkpoint, ran the benchmark, showed solved ARC grids, then tested the obvious question:
is it really reasoning, or is it leaning on puzzle IDs / recursion / augmentation voting?
Replay here:
app.clusy.io/share/bA9seSDW…
This is the kind of thing I want research workflows to become: run it, inspect it, challenge it. Try it at clusy.io now.



English
James Yang retweetledi

Makes us think about whether if its right to keep making the context windows bigger
Anish Paleja@Anipaleja
English
James Yang retweetledi

It sounds obvious once you notice it. Two machines are talking through a language designed for humans.
I don't think that's where AI infrastructure ends up.
That's why we built Brevitas: brevitassystems.com
Anish Paleja@Anipaleja
English

James Yang retweetledi
James Yang retweetledi
James Yang retweetledi





