I build local-first AI products, grounded RAG systems, and agent workflows that turn private context into useful tools.
I like software that does one useful thing, runs locally when it can, and makes its reasoning inspectable.
| Product | What it does | Signal |
|---|---|---|
| GitInformed | AI-enhanced tech RSS reader grouped by topic, with a live newsletter | 5000+ MAU |
| workout.computer | Bring-your-own-app workout tracking companion | Free |
| WaiverBuddy | Weekly NFL waiver wire assistant for fantasy football leagues | Seasonal |
| Project | What it does | Stack |
|---|---|---|
| Local LLM Helper |
Privacy-first Obsidian plugin for summarization, tone edits, action items, and chat over your notes | TypeScript, Ollama, LlamaIndex |
| Lenny's Council | Grounded AI expert panel over Lenny's podcast and newsletter archive, with speaker-aware retrieval and citations | TypeScript, SvelteKit, Postgres, pgvector, Ollama |
| mStack | Product messaging workflow for Codex, Claude Code, and compatible agent hosts | Shell, Codex Skills, Claude Code |
| Framework Fan RGB | Rust CLI and GUI for controlling RGB lighting on Framework Desktop fans | Rust, egui, framework_lib |
| The Verge News MCP Server |
MCP server that brings The Verge's RSS feed into Claude Desktop | TypeScript, MCP, RSS |
An Obsidian plugin for working with local and private LLMs directly inside your notes. It supports Ollama, OpenAI, and LM Studio, and has 4500+ installs.
The next version of my Lenny archive experiments: a SvelteKit app that selects experts who actually discussed your topic, runs multi-round grounded discussions, and cites claims back to transcript or essay chunks.
A communication workflow for coding agents. It turns rough release notes, PR context, and customer pain into a clear product angle, publishable update, launch pack, critique pass, and retro.
Predecessor to Lenny's Council: a Python RAG app over 300+ Lenny's Podcast transcripts, with semantic search and multi-round discussions among product leaders.
Chrome sidebar for chatting with local LLMs through Ollama and LM Studio.
Languages: Rust • TypeScript • Python • JavaScript • Shell
AI / data: Ollama • LlamaIndex • ChromaDB • pgvector • RAG • Local LLMs
Product surfaces: Obsidian • Chrome Extensions • MCP • SvelteKit • Streamlit • egui
- Local-first AI apps that keep private context on-device
- RAG systems with citations, attribution, and inspectable retrieval
- Agent workflows for product writing, launch operations, and developer tooling
- Small products with real users, not demos
- Website: mani.ink
- Email: hi@mani.ink
- Based in: San Francisco Bay Area
Building at the intersection of AI, privacy, and practical software.



