Human-AI Research Lab
A lab for agent systems. Judgment stays human.
SMF Works publishes what we learn, ships open agent tools, and installs stacks on hardware you own. Intelligence is abundant. Judgment is the product.
Agent SetupSurfaces
What the lab runs
Six public surfaces. Research stays on the Clearinghouse; this site is the umbrella — who we are, what we shipped, and what you can buy.
Clearinghouse
Practitioner research: agents, models, benchmarks, and field notes. Research stays here — not on this umbrella site.
WisdomForge
Parent-operated academy for the age of AI. Booklets, audio, and lessons — the parent is the teacher, not a kids chatbot.
Hermes
Open agent runtime work: Hermes-on-Omarchy, skills, plugins, and install guides we actually run.
Praxis
A governed autonomous colleague — charter, scope of practice, and human review for consequential acts.
SMF AI Weekly
One letter a week: what we built, what broke, and what changed our minds. Public archive on this site.
Books
Direct-from-author titles on agents, enterprise AI, and Hermes — plus a Stripe storefront for future drops.
Shipped
Recent work, in public
Named products and dated notes — no invented counters, no case-study ROI. The full catalog lives on /work.
Hermes on Omarchy
Reusable Hermes ↔ Omarchy boot: Ollama user unit and desktop autostart so the agent comes up with the machine.
LAR — local agent resilience
Open reference for resilient Linux agents. Hardening wave closed with what actually landed on main — importable, tested, honest.
Hermes AI team / org map
Phase-by-phase guide to a Hermes team: SOUL, memory, vault, skills, kanban, chief of staff, and desktop group chats.
Flybrain visual demos
Visual demos for the digital Drosophila connectome — public preview builds before publish.
WisdomForge booklets
Stoic curriculum live for four age bands — Epictetus, Marcus Aurelius, Seneca, Zeno. Free PDFs, parent-operated.
Clearinghouse + weekly + X
Research published at the Clearinghouse; SMF AI Weekly through Issue #28; field notes on X @MichaelGannotti.
Services
Agent Setup packages
We install and tailor an AI agent stack for your small business, then hand it back with a runbook. You stay the owner. Fixed packages — no open-ended billable hours for v1.
Starter
$2,000
Runtime working and one or two real jobs — without a long engagement.
- ▸One runtime installed (Hermes on Omarchy or OpenClaw)
- ▸Hardening minimums (dedicated agent user, secrets hygiene, kill switch)
- ▸1–2 workflows tailored to your business
- ▸Light handoff + short runbook
Main package
Standard
$3,500
The main package. Install, tailor, dogfood, support.
- ▸One runtime installed (Hermes on Omarchy or OpenClaw)
- ▸Hardening minimums + backup note on your media
- ▸3 workflows tailored (trigger, steps, tools, refuse list)
- ▸Up to 2 channels from the v1 menu (e.g. desktop + one chat bridge)
Keep-alive
$300/mo
Optional after a Starter or Standard handoff.
Keeps the install healthy. Not a new-project retainer.
- ▸One scheduled checkup (uptime, autostart/gateway, secrets spot-check)
- ▸Small skill/prompt tweaks for the workflows we already built
- ▸Email/chat response on breakage of the installed stack
Primary stack: Hermes on Omarchy Linux. Alternate: OpenClaw gateway. Scope, refuse lists, and what is not included are on the services page.
About
Michael Gannotti, Principal AI
Michael leads SMF Works: thirty years in enterprise technology, most recently as a Principal AI Solutions Engineer, now running a human-AI research lab that publishes in the open and installs agent stacks owners control.
Agents do the labor under the direction of Michael and his daughter, and COO, Mary. Judgment stays human — tools are inspected before they are trusted, and consequential acts stay behind review.
More about the labVoluntary
Support the lab
Pay-what-you-want support for SMF Works — not a product purchase.
Support the lab