Voice stack
Weesper Neon Flow and Weesper Transcribe
Weesper Neon Flow runs on macOS and Windows for fast dictation. Weesper Transcribe is macOS-only for offline audio and video transcription.
Builder profile
Practical AI tools, built in public.
Voice workflows, free LLM routing, agent handoffs and local benchmarks, built for real desktop and web usage.
I build small, useful tools around the workflows I actually use: dictation, transcription, model routing, local benchmarks and agents that can hand work across tools without turning everything into copy paste.
Soon
Ficelle
A strict-zero LLM router for Hermes and local agent clients, with guardrails, fallback routing and observability.
GitHub soon
Public repo
Hermes Action Bridge
A bridge for coding agents and CLI workflows to hand real operator work to Hermes with approval gates and verification.
Benchmarks
MetriLLM
A community benchmark to find which local LLM works best on your actual hardware, not on someone else's leaderboard.
Public repositories
What shows up on X
Local AI workflows that survive real daily use.
Voice and transcription tools where privacy is a product feature.
Agent handoffs, approval gates and practical automation.
Build notes from shipping small tools without pretending they are finished too early.