Diptanu Choudhury @diptanu

Founder @tensorlake, building sandbox infra for agents. Past - Built Nomad @hashicorp, AI Infra @meta, Container Scheduler tech lead @linkedin and @netlfix tensorlake.ai San Francisco, CA Joined December 2007

Tweets

9K
Followers

5K
Following

927
Likes

4K

Diptanu Choudhury @diptanu

2 hours ago

@rockwotj I think in the next 3-6 months we can remove all limits and just live migrate our workloads to bigger machines and claim back memory when they are not using them. Asking users to specify limits never work in practice

0 0 1 17 0

View Details

Diptanu Choudhury @diptanu

3 hours ago

Active CPU usage based billing makes it possible to use sandboxes at scale for background agents. I was looking into resource usage of some of our users. More often than not, most agents use between < 10% of CPU in aggregate of what users request for. I expect every platform to offer this shortly if they are not doing it already.

3 0 18 941 5

View Details

Diptanu Choudhury @diptanu

2 hours ago

@erans Yeah

0 0 0 26 0

View Details

Diptanu Choudhury @diptanu

3 hours ago

@HeyGarrison Nah, not yet

0 0 0 61 0

View Details

Diptanu Choudhury @diptanu

5 hours ago

@aniketmaurya @michielmv 💯

1 0 4 134 0

View Details

Diptanu Choudhury @diptanu

5 hours ago

@d_philla @armon Yeah dude, come home! And if you are in New York you can visit Armon’s new amazing cafe :)

0 0 0 22 0

View Details

/sesh/null @nerdsane

22 hours ago

Unikon.dev uses katagami.ai for design ingredients and runs on hosted temper in @tensorlake cloud. It's basically my own lovable with temper as it's universal constructor (that's why unikon)

/sesh/null @nerdsane

22 hours ago

Built this for my 9yo today - (took 10 min on @bot on phone).

1 0 18 972 1

0 1 5 400 3

View Details

Diptanu Choudhury @diptanu

a day ago

@w3ewen which durable fs are you using? :D

1 0 0 65 0

View Details

Diptanu Choudhury @diptanu

a day ago

It would be interesting if coding agent harnesses defined some sort of "portable session format". Sometimes I am working on a problem on my laptop but I will need to go out, and at that point I want to move the session off to a Tensorlake sandbox. The workaround for me right now is using TLFS, which can be shared between laptops and many sandboxes as a shared file system. I spin up a sandbox with the shared file system and resume the agent session there.

10 0 17 2K 6

View Details

Diptanu Choudhury @diptanu

a day ago

@damian_b How would this work? You would tell the currently running agent session to use a cloud sandbox and ask it to package it's own session history and code?

2 0 1 154 0

View Details

Diptanu Choudhury @diptanu

a day ago

@TreeCityWes That is what the TLFS workaround is lol

0 0 0 93 0

View Details

Diptanu Choudhury @diptanu

2 days ago

I miss writing code manually. There was a lot of joy in learning a new programming language, learn the APIs of a new database, go through the journey of coming up with a cool idea and seeing it work after a couple of days/weeks of hacking.

5 2 52 3K 2

View Details

Diptanu Choudhury @diptanu

2 days ago

@dotpem What do you think comes after Rust? I miss writing C++ tbh. If that community had figured out how to build something like Cargo and packaged it with the standard toolchain, life would be nice.

1 0 1 210 0

View Details

Diptanu Choudhury @diptanu

2 days ago

Pretty excited to be partnering with @datadoghq! We have a solid roadmap to support all the cool products their teams are working on.

/sesh/null @nerdsane

2 days ago

@GeoffreyHuntley @ibuildthecloud @AnthropicAI This is pretty neat first principles design composing boring hardened building blocks (elixir, cedar, nixos, firecracker, zfs) Pretty close to what we run at @datadoghq with temper + tensorlake. Once you opensource it, I’ll try to play with it by replacing your elixir with our

1 0 15 2K 7

0 0 16 1K 1

View Details

Diptanu Choudhury @diptanu

2 days ago

@techpoto Nope not fuse. Who else is doing it?

0 0 0 50 0

View Details

Diptanu Choudhury @diptanu

3 days ago

This is a huge issue. Our solution to this problem is a new content addressed file system that can share file chunks across sandboxes on the same machine thereby downloads very little file chunks in most cases. It was inspired by work from AWS Lambda a couple of years back. docs.tensorlake.ai/sandboxes/cas-… In most scenarios our new VMM resumes in a few seconds from sleep even if the sandbox snapshot/image is not cached. A provider we know asks their users to keep refreshing their sandboxes every couple of hours so that the images are cached on disk, etc. If you are evaluating a platform ask about p95 and p99 latency of real world sandbox create/resume latency without involving hacks that you would have to pay for.

Kush @kushbhuwalka

3 days ago

also, alot of providers pride themselves on a <100ms sandbox spin up. it doesn't matter if the sandbox spins up in Planck time if sticking my image on / launching the desktop takes 10-20 seconds.