I work full time running industrial robots at a glass factory.
I've been in school since 2003. I just finished.
In my spare time I've been quietly building something called GISMO.
The end game is simple: a fully local AI that knows you, runs on
your hardware, manages your digital life, executes tasks, remembers
everything — and never phones home.
Not a subscription. Not a cloud service. Not someone else's brain
that you're renting. Yours.
The problem with every AI assistant right now
Every AI assistant you use today lives on someone else's server.
It learns from you. It reports to them. You don't own it.
I run real industrial robots for a living. I know what happens when
machines do unexpected things without the operator knowing. That
experience is baked into everything I've built.
The core philosophy: policy before power. No silent actions.
Everything auditable.
What GISMO is today
GISMO (General Intelligent System for Multiflow Operations) is a
local-first autonomous orchestration system. It uses Ollama to run
a local LLM entirely on your own machine.
What's already built and working:
- Local LLM brain — Ollama integration, model-agnostic, zero cloud
- Durable task engine — queue, daemon, state machine, resume-safe
- Full audit trail — every decision logged with receipts and hashes
- Deterministic risk classification — every plan rated LOW / MEDIUM / HIGH before anything runs
- Policy-gated memory — persistent SQLite memory, operator-controlled writes only
- Supervisor lifecycle — up, down, recover, always-on service behavior
- Windows-first — because real people use Windows
- Leashed autonomy — the agent can only act through the same policy gates as the operator
Current status
- ✅ Phase 0 — Foundation
- ✅ Phase 1 — Local LLM Planner
- ✅ Phase 2 — Control & Guardrails
- 🔄 Phase 3 — Memory & Context (in progress)
- 🎯 Phase 4 — Interactive GISMO (the end game)
The hardware reality
I built all of this on a 7-year-old laptop while working full time.
That's part of the point. GISMO is designed to run on whatever
you have. No GPU farm. No cloud credits. Just Python and Ollama.
This is just the beginning
Phase 4 is where it all comes together. Always-on. Knows you.
Executes for you. Remembers everything. Runs locally.
Answers to you alone.
Not Alexa. Not Siri. Not Copilot. Not anyone's cloud product.
Just yours.
GitHub: https://github.com/GISMO-1/GISMO
What would you do with a personal AI that was actually yours?
Top comments (0)