tinkering and loitering | intern @ SF1.vc | prev @cornellCIS

Joined August 2018
60 Photos and videos
May 30
chart of the weekend! vibe shift of the past week about token spend has been surreal
The latest finding in the LangSmith Signal: Open Models are having a moment. 1 in 3 AI teams ran an open-weights model in April 2026, up from 1 in 5 nine months ago. The overall number of teams using open weights grew 3x. We’re seeing newer users choose open models at a higher rate than those who came before.
3
167
May 30
In addition to data being the moat, isn’t the router / optimization of the LLM router the moat for many of these businesses? Why would a company out-source that kind of data
3
75
May 25
I have multiple friends that still use desmos.com/ as their on computer calculator, always was and always will be a beautifully made product

99
May 21
Absolute rockstar and rocket ship
Today we're announcing our Series C funding: $355M at a $4.65B valuation, led by some great investors @generalcatalyst and @Redpoint. We've had insane growth in the last year, but we're still very early. So proud of the team and what we have built so far!
2
243
May 20
Think this compounds more than anyone expects over the next few years
The younger and poorer the generation, the more they eat out and DoorDash This is part of the problem
118
May 19
wow guess they are just letting in anyone now...
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
1
2
204
May 19
“we only invest in startups that use claude or codex but no, we do not use claude or codex”
2
198
May 18
Was Tuesday always the fundraise announcement day or did that just become a thing?
2
124
May 15
my prediction is that most of the neolabs will start releasing research previews of various products in the next 6 months new models with cutting edge research ideas will create hyper specific capabilities that end up crushing generalized benchmarks lot of interesting ramifications if so
1
151
May 15
lot of alpha here: the vc-ification of public market trading accounts is definitely here: just need some of these asymmetric bets to work out, which as we've seen can return entire funds
1
200
May 15
Leopold buying $fig and $team would go so hard
1
3
340
Canvas hack, not sure this is the best messaging
3
1,161
Figma. That's the tweet
Ramp TOP SAAS VENDORS MAY 2026 @tryramp 1. FIGMA is a fastest growing vendor, and the only publicly traded company on our list. Now competing with Claude Design. 2. AI inference platforms grew as businesses opted for cheaper models to combat exploding token budgets. 3. It's a great time to be in web deployment (Vercel, Netlify, Lovable)
1
184
for anyone who has to take a multiple choice test in the future, the models confirmed after many tests, really like to put the correct answer as letter B... so do with that what you will
2
106
First off, beautifully written stuff. Second off, feels good to use my brain to digest, understand and attempt to internalize Will's writing. Genuinely difficult after working with the clankers all day every day
3
151
Apr 21
I've been posting less and writing more. I've found it helpful to slow everything down and take the time for proper reflection. Today I'm putting out a review of the last quarter: Growing AI Anxiety Turning LLMs into Computers Anthropic and the Department of War Iranian Conflict The Incoming Techlash It's easy to feel like the response to this is for tech to "get tough" But I want to encourage the next generation to do something simple: keep your soul
2
575
Apr 20
find myself doing the same and have been giving this a lot of thought as to why: chat won memory first. i think this instilled in me a lot of personal interactions, questions, and long-term goals, so i knew my go-to “guy” would end up being chat. almost like having a childhood friend, whereas claude is more of a “work” friend who knows what i am currently working on. claude code cowork design the default of using the desktop app makes me much less likely to use claude for quick personal tasks on my computer (still figuring out the exact reasoning, but again, i think it is the general association with work that prevents me from asking random questions like “who holds the longest jeopardy streak” to claude). the mobile experience on chat feels cleaner. the haptics, the audio, the polish of the swift ui just wins.
i have no idea why or how this happened, but i use ChatGPT for personal and Claude for work.
1
3
357
Apr 20
pretty wild that @tbpn is just dead...
4
198
Logan retweeted
To check if your Google Workspace has been compromised by the same tool that compromised Vercel: 1. Go to admin.google.com/ac/owl/list… - This is Google Admin Console > Security > Access and Data Control > API Controls > Manage app access > Accessed Apps 2. Filter by ID = 110671459871-30f1spbu0hptbs6… - This is the ID of the compromised OAuth app If you see an app after filtering, you have potentially been compromised
62
745
4,449
1,055,077
Apr 15
Matt Levine about to be ALL OVER THIS
Allbirds just pivoted from shoes to AI compute and the stock is up 200% on the news This is the new economy
3
272