karan (@thekaranchawla) 's Twitter Profile
karan

@thekaranchawla

everything's computer | interested in hci, ai, systems, robotics | autonomy @zipline

ID: 906884103710167040

linkhttp://karanchawla.io calendar_today10-09-2017 14:16:34

1,1K Tweet

1,1K Takipçi

462 Takip Edilen

karan (@thekaranchawla) 's Twitter Profile Photo

agents without persistence are NPCs, agents with persistence are players working on vvm memory semantics tonight, a short thread on what's actually hard:

Nihanth (@nhnt11) 's Twitter Profile Photo

Love what Karan is exploring with VVM. I see it as formalising a runtime agnostic platform for building structured agent workflows.

🕊️ (@lichthauch) 's Twitter Profile Photo

Will is the rarest skill on earth right now. you can mass produce intelligence you can buy beauty you can fake status but you cannot download will from anywhere it has to be grown in suffering and nobody wants to suffer so nobody has any. one man with will is worth ten thousand

karan (@thekaranchawla) 's Twitter Profile Photo

agents that learn your codebase. remember your preferences. pick up where they left off. vvm now supports persistent memory across sessions vvm.run

agents that learn your codebase. remember your preferences. pick up where they left off. 

vvm now supports persistent memory across sessions

vvm.run
karan (@thekaranchawla) 's Twitter Profile Photo

context holds handles. disks hold payloads. this was my core insight when building the new filesystem mode for VVM and it changes the math on multi-agent pipelines when you orchestrate an agentic pipeline such as: research → analysis → synthesis you don't want to pass

karan (@thekaranchawla) 's Twitter Profile Photo

here's a demo of a code review system built with ~100 lines of VVM code. this runs 9 specialized reviewers running in parallel. each writes a full review to disk. the orchestrator just tracks refs. the synthesizer pulls what it needs. context stays bounded.

Reggie James (@hipcityreg) 's Twitter Profile Photo

Theodore: Do you talk to someone else while we're talking? Samantha: Yes. Theodore: Are you talking with someone else right now? People, OS, whatever... Samantha: Yeah. Theodore: How many others? Samantha: 8,316.

Theodore: Do you talk to someone else while we're talking?

Samantha: Yes.

Theodore: Are you talking with someone else right now? People, OS, whatever...

Samantha: Yeah.

Theodore: How many others?

Samantha: 8,316.
Marc Simonetti (@kemar74) 's Twitter Profile Photo

"All governments suffer a recurring problem: Power attracts pathological personalities. It is not that power corrupts but that it is magnetic to the corruptible. " -Frank Herbert, DUNE

"All governments suffer a recurring problem: Power attracts pathological  personalities. It is not that power corrupts but that it is magnetic to  the corruptible. "
-Frank Herbert, DUNE
j4orz (@j4orz) 's Twitter Profile Photo

> Some companies hire heavily out of Twitter, some hire from communities such as GPU Mode or NanoGPT speedrunning. To Nathan's point, I am leading an open source workgroup within the GPU MODE community (#teenygrad channel) in order to develop a deep learning systems course

> Some companies hire heavily out of Twitter, some hire from communities such as GPU Mode or NanoGPT speedrunning.

To Nathan's point, I am leading an open source workgroup within the <a href="/GPU_MODE/">GPU MODE</a>  community (#teenygrad channel) in order to develop a deep learning systems course