Logan Matson The future of work,
built by hand.

AI systems. Automation pipelines. Tools that run while you sleep.

Scroll to explore
→ Demo → Validator → GitHub
01 — Case Study

A Real Impact

8 Years of Epilepsy and Invasive Neurosurgery — now my health is driven by an autonomous system that plans, tracks, and self-corrects.

For years, recovery from seizures meant spending 90% of my time planning and 10% acting, with no real loop back to see what worked. So I built one.

Six autonomous agents run the system end to end: meal planner, training calendar, neurologist risk forecaster, Whoop analyst, Google Workspace operator, weekly orchestrator. They share a pipeline-state.json nervous system that connects every component.

Twenty-five-plus subagents handle parallel work. Calibration weights self-tune monthly through a retrospective validator that checks every forecast against real biometric outcomes.

126days of self-tracked signals
35+data points captured per day
3,707Whoop biometric rows analyzed
16calibration factors tuned monthly

Key insights surfaced

  • HRV → next-day recovery r = 0.815 — the strongest predictor in the system
  • Strength days yield +6% next-day recovery vs. run days; rest days redistributed accordingly
  • Forecast-driven rest has cut overtraining flags and lifted average weekly training quality

Why I built this

Eight years of epilepsy taught me that anxiety is what bad planning feels like — I was spending 90% of my energy guessing, and 10% acting on it. The unlock came when the agents started surfacing things I'd never have caught by hand:

  • The neurologist agent flagged hydration drop-offs the day after long runs — a pattern I'd been missing for months that mapped directly to how I felt the next morning.
  • The Whoop analyst found that meditation moved my deep sleep more than any other input — more than caffeine timing, more than screen time, more than bedtime itself.

Why your company needs the same thing

Solo operator buried in 14 spreadsheets and one CRM. Your "data" is everywhere and nowhere. An agent pipeline pulls from each source nightly and gives you one source of truth.
Founder with good metrics but no connections between them. You can see each number; you can't see the patterns. A self-calibrating analyst agent learns what predicts what — and surfaces what you'd never catch.
A company with no website demonstrating what it can actually do. Pitch decks are a guess; live, working systems aren't. One agent build, deployed, becomes the proof.

AI agents cut the planning tax. You spend your time on the work — not on figuring out what to do next.

Interactive map of the agent network: circular SVG flow, expandable role panels, animated handoffs between subagents. Built to show how the system thinks before it acts. Open the Demo link above.

Monthly retrospective validator: every forecast (training load, recovery, seizure risk, meal-recovery correlation) is checked against the next month's real biometric data. Drift gets pushed back into agent-weights.json so the system gets smarter on its own. Open the Validator link above.

02 — Architecture

How I Build

A pattern, not a product. Anyone can apply it.

Agents

Autonomous workers that complete a job end-to-end — from input to delivered output, no babysitting.

Skills

Reusable capabilities the agents draw from. Think of them as muscle memory the system never forgets.

Workflows

Orchestrated sequences. One agent's output becomes the next agent's input — a chain of small, validated steps.

Subagent Dispatching

Independent tasks fanned out in parallel. Each runs in its own context so nothing bleeds into the others.

State Files

A shared JSON nervous system. Every component writes its block; every component reads what it needs.

Self-Calibration

Agents that audit their own forecasts and tune their weights from the gap. They get sharper with every run.

→ Let's Talk
03 — Working Together

What I Do For You

Solo founder drowning in admin → a morning digest agent that reads your inbox, calendar, and Slack and delivers one Gmail at 7am.

~5 hours/week reclaimed · 100% inbox triage before your first meeting

Small team running 8 spreadsheets → one pipeline that pulls from your tools every night and updates itself.

90% reduction in data-entry errors · 8 spreadsheets → 1 source of truth

Data but no insights → a self-calibrating analyst agent that learns your domain over 30 days and surfaces what you'd never catch by hand.

Pattern detection in 4 weeks · forecast accuracy compounds month over month

Stuck on repetitive ops work → I install one agent and walk your team through running it. You own it the day I leave.

One working pipeline deployed before handoff · zero vendor lock-in · full source delivered

→ View All on GitHub
04 — More Builds

In Production

Health Agent Ecosystem · flagship · 6 agents · 25+ subagents → A multi-agent network that plans my training, food, and recovery every week, with zero manual handoffs.It proves the architecture works in production, not just on a slide. Ecosystem Demo · interactive SVG flow → Click any agent to expand its role, inputs, and outputs.Built so anyone can see exactly how the network reasons step by step. Validator Dashboard · monthly accuracy tracker → Every forecast checked against real biometrics.Drift gets pushed back into the system; the whole pipeline gets sharper on its own. Pittsburgh Marathon Plan · cartographic 7-zone race plan → Race-day pacing strategy, nutrition windows, and elevation-aware tactics for all 26.2 miles.Turns a generic training app into a custom tactical brief. Whoop Performance Brief · monthly analyst PDF · 1,500+ data points → Auto-generated each month from raw Whoop CSVs: sleep, strain, recovery, behavioral correlations.The same pipeline works for any time-series data, not just biometrics. Skill Portfolio · 30+ skill empire · self-managing → Live index of every reusable capability my agents draw from: created, audited, and rotated automatically.Shows how a system can curate itself without me touching it. Scotland · London · Dublin Itinerary · 9-day interactive trip planner → Hike toggles, packing checklists, real bookings, 50+ deep links, all in one HTML file.Same pattern works for any complex personal logistics, not just travel.
05 — Why Now

The Pack Is Already Moving

AI agents cost $0.25 to $0.50 per task. Humans cost $3 to $6.

That's a 90% per-interaction cost cut, and you sleep through it.

AI super users save 9 hours every week.

That's 6 weeks a year reclaimed for the people who actually integrate it. The rest are still in meetings.

3.5x average ROI on AI agent investment. Top performers hit 8x.

Year 1 returns 41%; year 3 compounds past 124%. Waiting costs you the curve.

94% of companies have deployed AI but report no significant value from it.

Get ahead of the pack. The difference is who actually integrates the tools, not who buys them.

06 — Contact

Let's Build Something

→ Tell me what's broken, I'll show you what to build → Connect on LinkedIn