Atharva’s Summary
This community was started after Nilesh, Arvind, Prabhanshu etc met on the FoC global slack and decided to do these meetups. Nilesh does pretty much all the organising work where he pulls in people from his network. This whole thing is invite-only and happens in Blume Villa. Everyone present usually has to give a 5-minute demo of their own.
The demos were a mix of long-term projects, scrappy side projects and research that people do. One common theme that I felt tied the meetup together generally was that there was a streak of “alt-computing” going on, ie, not super common or corporatey type of work.
Some unordered highlights to convey the vibe
- KS Sreeram was an interesting person who writes his own programming languages and has this vision of “breaking the tyranny of the text editor” for programming. He demo’d an educational tool that worked like the editor that was in the Bret Victor inventing on principle demo. But had some other interesting add-ons around snapping to expressions/reactivity and avoiding failure states that blow up the program.
- Abhinav Tushar demo’d a multimodal emacs editing mode exploration where he tries to correct code or add comments via dictation while also simultaneously typing in an attempt to increase his bandwidth. Uses a mix of whisper and LLMs to interpret the instructions and stream to the buffer.
- Paras Chopra demo’d an exploration where he tried to generate (scheme-like) valid programs based on a given set of inputs and outputs. He showed ways in which he tried to make it generate more and more complex programs progressively, by using grammars and then progressively adding in genetic algorithms.
- Ruturaj is a core maintainer of circuitverse, a digital logic simulator, and he showed how people make circuits and put them into components. Showed how the community implemented Conway’s Game of Life with circuits and even fully working CPUs.
- Abhay aka Nemo showed a demo where he tried to hack an OpenAPI-based workflow orchestration system called Arazzo to implement Stephen Wolfram’s Rule 110 automaton, since he figured that Arazzo is turing-complete. He hasn’t got there yet, but it was a fun scrappy demo.
- Unnati showed her in-progress demo which tries to analyse political comedy routines and tries to annotate the bits having laughter as some rough measure for humour. Spoke about interesting properties of laughter sounds and the surprising result of how the LLM’s non-english language understanding influences the laughter captured. Also mentioned random tidbits of trying to analyse humour with these models, effects of code-switching and swear words. Also some shoutouts to keyboard-based accessibility standard called WebAIM.
- Arvind showed off remix, a work-in-progress node and wire based visual programming language. We spoke a bit later about how he’s running the project, how it’s funded and how the literal spaghetti of wires can be managed, and the challenges with getting it right.
- Nilesh spoke about some algebras that are in partial order (like booleans and numbers) and how they can be unified to implement generic event-management system that could cover everything from JS-style event loops to inboxes to order management systems.
- Shubhojit showed a graphical simulation of teams of ants trying to get food and eliminate other ants. Basically an agentic sim demo thing. Then spoke about how they used this idea to figure out optimal battery point locations(?) for Ola electric drivers. He could’t give any interesting detail of that part unfortunately because proprietary corporate secrets.
- Advait showed off his Mars computer startup which is a thin client thing which lets people stream powerful hardware to their laptops for video editing/3D modelling etc
- Ishan showed his research on performing automatic smart refactors by examining the types, along with some math/proofy things to show why it worked
- Aditya showed off his custom-built CPU that’s specialised for running AI inference/training.
- Satpal (also from Turing’s Dream) showed an Open Source library he’s building to generate synthetic data sets.
- Prabhanshu showed the tables model on bean along with the LLM table placement inferencing.