Lavan (@lavanpuri) header
Lavan (@lavanpuri) profile picture

Agents, Distributed Systems and Frontend | CS at UIUC

San Francisco, CA
Joined August 2014
Photo by Lavan (@lavanpuri)Photo by Lavan (@lavanpuri)Photo by Lavan (@lavanpuri)Photo by Lavan (@lavanpuri)
Introducing FloorSense. We've been quietly working on FloorSense with a big mall in Thailand It's an AI that turns a mall's existing CCTV into a live picture of how people move. It can count people, see where they walk and stop, and even follow the same person across cameras without knowing who they are, kinda like Google Analytics but for a building. Still early, but seeing it work on real crowds is pretty crazy. If you run a mall or any big space, DM me. Would love to try this with more people.
18
6
2
92
3,087
Lavan retweeted
not a render. 6 agents coordinating a climbing gym model, live. they share the same file, wait when they get ahead, and a QA agent pings me when it has to assume a height. every cursor move is a tool call.
5
2
9
529
The experience of orchestrating agents on t3code is really good @theo @jullerino
2
3
117
Really nice design, so much better visibility
74
We let an AI agent go shopping 101,416 times. 1,044 stores. 25 categories. Every step recorded. It finished the job about 7 times out of 10. Most failures came at checkout. Sometimes it bought the wrong thing because it never asked a question. Sometimes it gave up when it could have finished. We wanted to see how agents really shop, and where they break. Now we're releasing all of it for free. It's called CartTrace-100k.
22
11
3
56
2,223
Lavan retweeted
I asked Opus 5.5 to turn a real apartment project into a walkable 3D site pick a tower, slice down to any floor, walk into a flat, then drag the sun down and watch the lights come on try it: siliconbricks.co/aayra
5
2
12
612
Gave a demo on getsesame.dev/ at @WorkOS Demo Night! (Thanks @grinich!) Here are a couple cool features we built: - Auto detect secrets and passwords in your agent's host machine - Transfer the secrets of your choice (or all of them!) to Sesame - Secure and log all outgoing requests from your agent without changing any routine scripts, harnesses
2
4
1
8
321
Fan-girling over LLM security with @unrealKushagra at yesterday’s AI Tinkerers meetup in Boston
1
1
4
141
Lavan retweeted
im late to the loops discourse but from what i'm seeing it's mostly about creating a loop from your asshole back into your own mouth?
128
67
15
2,388
110,782
Limit your rage prompts folks, it can quit on you
1
4
185
Lavan retweeted
guyssss, we just crossed 1000 uploaded images on vibewtf.com, so happy!! 🥹
2
2
14
278
this is super cool, gonna embed this in my workflow!
Stop pasting API keys into agent prompts or memory files. Your AI agent needs secrets. It doesn't need to see them. Your agent tries to refund a customer. Sesame denies it. Only read access was authorized. Brokered. Audited. Revoked in one click. 70+ providers, bring your own vault.
1
3
227
Stop pasting API keys into agent prompts or memory files. Your AI agent needs secrets. It doesn't need to see them. Your agent tries to refund a customer. Sesame denies it. Only read access was authorized. Brokered. Audited. Revoked in one click. 70+ providers, bring your own vault.
4
3
3
7
877
Try it at getsesame.dev 5 minute install.
1
156
Lavan retweeted
i say the we're T -8 days away from the flippening!
1
1
174
I’m a claude kinda person @cursor_ai
1
3
210
Night 10: how me and my cofounder be working with a 13 hr time difference
Night 9: how me and my cofounder be working with a 13 hr time difference
1
1
1
11
1,156
i can't work today there is a cat on my screen
2
1
6
340