Nilesh Trivedi
banner
nilesh.trivedi.link
Nilesh Trivedi
@nilesh.trivedi.link
Building AIs at snowmountain.ai

I love machines, mathematics and music.

Earlier: Meta, Foresight Institute, Clear, Gupshup, IIML, IITKGP


❤️ http://isfixable.com

https://nilesh.trivedi.link/
https://x.com/nileshtrivedi
https://fosstodon.org/@nilesh
Checking in after many many weeks.

What's been happening on #bluesky - both the network and the software?
November 6, 2025 at 7:42 AM
We're having the first edit-a-thon for the Fixable hardware Wiki ( isfixable.com )

3PM today on Google Meet: meet.google.com/ybx-eose-jbi

#openhardware #RightToRepair
Fixable products for India | fixable
isfixable.com
October 18, 2025 at 7:10 AM
I'm putting together this online mixer this Sunday for those who’re building coding agents. #ai

luma.com/o26hrnb4
Coding Agent Builders Mixer · Luma
If you're building Coding Agents using LLMs, this online meetup should help you connect with similar folks and discover ideas and opportunities.
luma.com
September 5, 2025 at 10:30 AM
How come client-side calls to LLMs not more popular?

OpenAI, Anthropic, Gemini all support it. But no client-side library has taken a foothold for this pattern.

@simonwillison.net had written about this a year ago.
August 24, 2025 at 4:42 PM
#math notation you didn't know you wanted!
August 22, 2025 at 9:31 AM
Floorp v12 got Vertical tabs, AI chatbots in sidebar with multiple LLMs without data leakage.

Is there anything I'm missing from Comet / Dia / Chrome? 🤔
August 21, 2025 at 2:44 AM
This is depressing but only if you THINK about it. 😁
August 13, 2025 at 7:13 AM
Reading survey papers across disciplines is insanely effective.

Not only do you learn almost state-of-the-art knowledge that has not yet made to textbooks, but you also build a mental map of the field and how different ideas relate to each other. I find this scaffolding very useful for retention.
August 12, 2025 at 9:42 AM
Topology is so underrated.

I used to think it's just a corner of abstract mathematics, but now I see EVERY question in terms of what the space of possible answers is and what the properties of its shape are because they inform approaches for design, search & optimization.

#mathematics
August 12, 2025 at 6:56 AM
I have been thinking about distributed reactive computation. There is way too much plumbing involved in building apps.

Have landed on an elegant model, with which, expressing a multi-user, secure, high-performance todo app takes 20 lines of code (including a basic reactive UI).

#computing
August 12, 2025 at 6:54 AM
Duty calls.
July 18, 2025 at 10:18 AM
Reposted by Nilesh Trivedi
Well, here we go. After well over two years, here's a new version of my programming languages book, PLAI (v3.2.5). As always, free of cost! Can't thank enough all the people named in the acknowledgments. Enjoy!
www.plai.org
Programming Languages: Application and Interpretation
Site for the PLAI book
www.plai.org
July 14, 2025 at 5:20 AM
This is such a good site:
July 2, 2025 at 1:26 AM
Reposted by Nilesh Trivedi
All these categories seem fuzzy to me. I recommend modeling everything as agents with declared attributes in case it happens to be non-intelligent and non-longrunning, i.e. tool-like.
June 10, 2025 at 4:55 AM
This was a mind-bending read!

Argues that intelligence might originate outside of physics and actually come from the platonic space of mathematical patterns.

For eg: the familiar "bubble sort" algo shows cognitive skills like delayed gratification and clustering.

osf.io/preprints/ps...

#ai #cs
June 7, 2025 at 4:26 PM
On landscape displays, horizontal scrolling seems best for reading ebooks (until we get a really tall/portrait display that can show 3 pages at once). They make it easiest when pages refer to previous or next page, especially with visuals. #design #ebook
June 7, 2025 at 4:31 AM
I wrote a new blog post.

Coding Agents are supersets of not just programs and ML models, but also for the first time, PROGRAMMERS!.

But Type Theory and Tests only covers programs. We need a theory of Evals.

Link to full post: nilesh.trivedi.link/thoughts/we-...

#ComputerScience #AI #typetheory
We need a formal theory of Agent Evals | Nilesh Trivedi
nilesh.trivedi.link
June 4, 2025 at 1:45 PM
The world does not seem to care about the incorrigibility of #Pakistan. And #India always bears the cost of it. ☹️

Last image shows Pak military officials giving state funeral to a terrorist, attending along side HAFIZ ABDUR RAUF who himself is a Specially Designated Terrorist under OFAC sanctions.
May 12, 2025 at 12:03 PM
This is such a good idea!! Early days (prototype is based on Fairphone 2 only), but this shows the power of open hardware.
April 9, 2025 at 7:04 AM
What are the chances that the hardest implication remaining to be proved or disproved in Terence Tao's Equational Theories project happens to be indexed at Ramanujan's Number 1729?
April 9, 2025 at 3:33 AM
Announcing VibeLang 1.0 - a new #programming language for the AI age.

Designed for professional vibe-coders.
April 1, 2025 at 3:09 PM
Happy to see progress on the interoperability of infinite canvas tools: canvasprotocol.org/

Will add support for this in my Graph editing tool if it supports nesting properly.
Open Canvas Working Group | OCWG
Working group for establishing interoperability between different infinite canvas tools.
canvasprotocol.org
March 31, 2025 at 5:50 AM
Reposted by Nilesh Trivedi
Where have all the tech creators gone?
March 22, 2025 at 4:28 AM
Just gave a lightning talk about the #repairability wiki for India
March 22, 2025 at 9:05 AM
I am at #MisinfoCon2025 today.
March 22, 2025 at 6:56 AM