The AI-First Manifesto
Six principles for work where a person selects the problem and judges the result, and the machine does the execution — with every step auditable. It argues that the architecture is no longer the bottleneck, so a small team can now take on problems that used to need an institution. This is the position the whole series is arguing from; start here.
Code at First Light
Three teams took an agent harness outdoors — running a small model on Android phones, on battery, with patchy signal — to find out what the constraint actually does to the design. The conclusion it ends on: the harness is the product, not the model.
Anatomy of an Agent
Fifteen builders read pi-mono and Claude Code side by side, looking for the patterns both had to solve: the tool loop, the permission model, streaming, and how state survives a restart. An agent harness is not the model. It is everything around the model that decides what the model gets to see, do and remember.
How to take part. Sessions run in Pune and the write-ups are open to anyone. If you want to be in the room, or you have a domain problem worth building a session around, mail parth@varahitechnologies.com.
Code from the sessions is on GitHub.
If you are thinking about working here, this is the fastest way for us to meet you. Turning up, reading the code and disagreeing with us is a stronger application than a cover letter — several people here arrived that way.
Open roles →