yash@jain:~$ ls -la /projects
projects
Things that are live and running. Each one is documented somewhere in the archive.
The business. Self-hosted, Slack-native agentic AI for 10–100 person companies. Deployments the client owns outright — infrastructure, knowledge, and all.
What do your agents actually cost? Live LLM API pricing across 1,230 models and 81 providers — model your token mix or set a budget, including image and video generation. Refreshed daily from provider APIs, with a public pricing API.
Who stores your prompts, who trains on them, and who offers zero data retention. 43 providers tracked across US frontier labs, Chinese providers, inference APIs, and AI coding tools.
The reference for AI deployments that have to work. Eight failure modes, ten core concepts, recovery playbooks — written for the solo implementer at a small business, not the enterprise architect.
Live usage stats from my own agent fleet — $3,963 in tracked spend, 242.5K messages, 2.73B input and 80.6M output tokens, 11.65B cached. The public receipts behind the cost-math posts in the archive.
The nerdy side. AI-assisted music videos on philosophy, psychology, science, and history — We Are Stardust, The Sixth Extinction, State of Flow, Jung's Shadow.
more on github ↗