ceo @flowith to shape the interface for AGI. tech optimist, AI engineer, designer. previously founded X ACADEMY and Realm.
San Francisco, CA
- Tweets 1,100
- Following 877
- Followers 13,510
- Likes 5,304
We've been quiet for 8 months.
Because we've been busy building the infrastructure for a 100% agent-led companies.
Still in the beta phase, but I can't hold back this preview.
Introducing Matrix, where anyone can launch a 0-Person Company that actually earns.
And yes, Matrix beta already achieved SOTA on the frontier harness, matching Fable's performance.
165
129
794
411,160
it is late 26 and i still haven't seen the definitive local personal agent stack.
i want to run my own models on my own data. i want a context layer that scales forever and is built for agents from day one. and i want claude code plugged right into it.
what is the closest thing we have to this today?
13
1
14
1,528
this is insane. mac rental businesses are printing money right now.
apparently neocloud has officially reached the mac mini layer of the compute stack.
i get it though. every time i travel to asia or china for events, i keep my claude code hosted on a box in the us. way too paranoid to run it locally on the ground. the demand is very real.
3
1
12
1,629
this is actually insane.
just built a playable gta-7 online mini game and this trailer using the new @deepseek_ai -v4.1-flash-1017 + @threejs. the quality easily matches opus 5.5 at 1/30 the cost.
the video speaks for itself.
honestly feels like gpt is cooked. overpriced, nerfed intelligence, throttled speeds, and slashed usage caps. nobody wants to build on that anymore.
5
5
202
12,383
spent 3 minutes yesterday making an ai video about the roadster. woke up to 20 million views.
my timeline is flooded with memes, but scrolling through them made me stop and think about something deeper. about @elonmusk, the power of delusion, and what it means for the rest of us.
elon is the ultimate daydreamer. his playbook is simple. set an impossible goal, lock in for years, ignore the noise, and force an asymmetric outcome into existence.
as a founder watching him run this loop over and over, i used to think this mode of operating was reserved for outliers with infinite capital.
then i read a piece by @felixrieseberg at anthropic on how ai lets us become wildly ambitious. and it all clicked.
ai collapses the cost of ambition. you no longer need an empire to chase impossible things. with limited resources and the right leverage, we can all run the elon loop now.
be more delusional.
1
396
spent 3 minutes yesterday making an ai video about the roadster. woke up to 20 million views.
my timeline is flooded with memes, but scrolling through them made me stop and think about something deeper. about @elonmusk, the power of delusion, and what it means for the rest of us.
elon is the ultimate daydreamer. his playbook is simple. set an impossible goal, lock in for years, ignore the noise, and force an asymmetric outcome into existence.
as a founder watching him run this loop over and over, i used to think this mode of operating was reserved for outliers with infinite capital.
then i read a piece by @felixrieseberg at anthropic on how ai lets us become wildly ambitious. and it all clicked.
ai collapses the cost of ambition. you no longer need an empire to chase impossible things. with limited resources and the right leverage, we can all run the elon loop now.
be more delusional.
1
2
13
1,974
got some early tester feedback on gemini 4 argon.
massive jump in raw reasoning. the only problem is it is way too cautious, spends too much time over-validating, which hurts efficiency compared to opus 5.5.
anthropic is still ahead at the frontier. early checkpoint caveats apply of course, final weights might feel very different.
6
2
183
12,317
spent the last two hours running the exact same video generation prompt on sonnet 5.5 vs opus 5.5.
the result is pretty counterintuitive: OPUS 5.5 was both faster and cheaper.
side-by-side comparison in the video. curious which output you think is actually better.
Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family.
It鈥檚 a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work.
3
1
18
3,136
introducing flowith canvas powered by opus 5.5.
we used to spend $10,000 on studio work for this kind of piece. we did it twice.
opus 5.5 just made it for $10 in tokens in about an hour. one shot, zero extra apis.
hard to overstate how fast things are moving.
11
6
61
9,761
spent the whole day playing with @typesafeai jev. built some pretty wild demos that would easily farm engagement if i posted them right now.
instead i tried actually shipping it to production. swapped out gemini 3.8 flash/lite for jev on non-visual DOM execution in matrix browser use and modified our harness. real websites break it almost instantly, it's general reasoning just isn鈥檛 ready for browser use yet.
the @browser_use jev demo is super cool, but that page was basically an ai-tailored sandbox, not the real messy web.
the timeline is increasingly optimized for cherry-picked shock value over real utility, which ends up misleading a lot of builders trying to ship actual products.
jev is a great concept and an impressive model. genuinely looking forward to the open source weights and multimodal release.
9
5
78
7,310