263 Photos and videos
i’ve tried to consistently use at least two or three different model harnesses every day.
i’m quick to pick favorites, but it’s also quite interesting to see the variation across different models and harnesses for different tasks.
3
240
May 27
just did my weekly rep of searching for the apple tv remote between the couch cushions
7
184
May 26
i truly believe there has never been a better time to study computer science/engineering
which is why this is crazy to see
College CS enrollment is declining
Charts of the Week: a16z.news/p/charts-of-the-we…
3
8
569
May 20
if you reason from fundamental realities and incentives, you arrive at conclusions that are both more optimistic and more probable
6
143
May 17
this is what the future of engineering looks like.
if you’re resourced to do this, i think you absolutely should be. if you don’t have the resources to operate this way yet, figuring out how to get them should be a p1 priority. you are now competing against teams with massively leveraged engineering output.
we are no longer just building software. we are building software that autonomously builds more software.
3
1
8
893
May 7
just wait until the masses figure out they can just use a dev vm instead of walking around with their laptop half open
1
5
222
May 5
peek into what we have been working on!
May 5
Agents have redefined the modern inference workload with multiplying complexity:
1. scaling test-time token usage
2. increasing long-context usage
3. introducing high entropy inference patterns as agents themselves drive decisions to orchestrate sub-agents and invoke tool executions.
In this latest article, I collaborated with Eduardo Alvarez and Graham Steele to break down the anatomy of these workloads and explain how NVIDIA’s extreme co-design approach translates into building the best inference stack in the world for agents.
Give the article a read! 👇
9
771
May 5
my favorite color to see in ci is green
May 4
my favorite kind of revenue graph is the kind that goes up and to the right
5
289
May 4
it’s been my experience that a lot of students and others in academia, outside of the bubble at a few prestigious schools, are struggling to keep up with the frontier
as a student, it's worth doing whatever you can to stay close to it by using the latest tools, reading, and carving out time every day, even if it’s just 15 minutes, to learn something new
chatgpt plus is $20/month, same as netflix
nah my classmates really dont know about gpt 5.5
they're still all on 5.3 instant and complaining that it's shit
1
5
386
May 4
my favorite kind of revenue graph is the kind that goes up and to the right
1
5
472
Dylan Mitic retweeted
May 3
My friends at provenance are hiring a founding engineer!!
They are building the infrastructure for AI agents in finance and already have contracts with top tier banks. Work in SF, housing and living expenses paid for, equity, and more. Dm me or email uskudar@provenancexl.com for more info!
2
1
7
675
May 3
i don’t buy the moat erosion narrative
seems to me that if everyone has the same tools, all software will become equally more complex and powerful meaning the whole distribution will shift up but that doesn’t mean the outcomes equalize. if anything, the variance between companies expands as better operators leverage the tools more effectively.
i think its true that the burden of innovation will be on the incumbents but it’s always been that way so ai doesn’t seem much different in that regard.
i could see this changing if access to models is gated but that hasn’t been the reality. everyone’s using roughly the same tools, or close enough
5
202