HELLO, I’M A PERSON OF MANY MODES

Billy
Pronovost.

I build small, useful systems with AI in the loop.

I’m Director of Technology at Pendleton. I lead technology, make software, shape products, and work alongside AI—usually where those jobs overlap.

TRY THE WORK

MODES OF PRACTICE / SELECT ONE

Choose a mode. Arrow keys work here too.

PUBLIC EVIDENCE / 01

Make the work legible.

Turn competing needs into a direction people can understand, question, and carry forward together.

At Pendleton, I direct technology work while translating between people, priorities, and the systems that support them.

Works beside Systems Thinker + Product Builder

THINGS I'VE BUILT / USE THEM RIGHT HERE

Nothing here is
a screenshot.

Every card below runs the real thing in your browser — drag it, tap it, break it. The last one is the method I used to build this page.

Drag the handle — a pixel-blurred mark becomes a crisp, infinitely scalable vector.

NATIVE iOS APP · IN TESTFLIGHT

Skalable

Turn any image into a clean, usable SVG — right from your phone, in seconds.

  • iOS 17+
  • Image → SVG
  • Fidelity-first
  • On-device

Build 32 · TestFlight

WHY IT EXISTS

I was building with AI agents and needed a logo. What I had was an AI-generated PNG at some unknown resolution — and raster marks fall apart at the sizes a real site needs. I wanted to fix that from my phone, in the moment, without opening a design tool. So I built Skalable.

HOW IT WORKS

  1. 1ImportPull an image in from Photos, Files, or the share sheet — PNG, JPEG, HEIC, WEBP.
  2. 2TraceIt vectorizes on your device, with live preview and fine-tune controls.
  3. 3ExportSend the SVG to Files, the share sheet, or straight to the clipboard.

WHAT MAKES IT DIFFERENT

  • Runs on-device
  • No servers, no accounts
  • Fidelity-first output
  • Gradient synthesis
  • Outline inspection mode
  • Layered SVG export
  • No desktop, no design tool
STATUS · TestFlight OUTPUT · SVG STACK · SwiftUI + Rust

SHARED HOUSEHOLD BILLS / $2,400 A MONTH

A pays$933
B pays$1,467

Split 50/50 instead and Partner A pays $267 more than their income share.

Move either income — the fair share recalculates. This is the actual math inside the app.

iOS + ANDROID APP · PRELAUNCH

Equa

Couples split shared expenses by what each person earns — not down the middle.

  • iOS + Android
  • Income-based split
  • Zero-knowledge
  • One-time purchase

In prelaunch validation

WHY IT EXISTS

Splitting rent 50/50 when one partner earns twice as much isn’t actually fair, and most couples know it — they just don’t want to run the spreadsheet every month. Equa does that math quietly in the background so the conversation can be about something else.

HOW IT WORKS

  1. 1PairTwo people connect with a pairing code. One purchase covers both.
  2. 2SplitEnter incomes; Equa sets each fair share. Override to 50/50 or custom any time.
  3. 3BalanceLog expenses or scan a receipt. Equa tracks who covered what and when you’re even.

WHAT MAKES IT DIFFERENT

  • Fairness is the default
  • End-to-end encrypted on device
  • No bank connection required
  • No subscription
  • Receipt scanning
  • Shared savings goals
  • Free demo, no account
STATUS · Prelaunch PLATFORM · iOS + Android MODEL · One-time purchase

REQUEST → CONTEXT → PATCH → PROOF

Start with the actual message. “You aren’t actually doing anything with the site” becomes a concrete target: make a visible card, not another plan.

Tap the chips — a complaint turns into a repo patch, tests, and screenshots.

AI PARTNER SYSTEM

Lumi / Hermes

A local AI partner that can pick up context, edit the real site, and prove the work happened.

  • Local context
  • Repo edits
  • Tests
  • Screenshots

Live system

WHY IT EXISTS

I want AI collaboration to leave artifacts behind: changed files, passing checks, screenshots, and a better next version of the thing we’re building.

HOW IT WORKS

  1. 1ReadLoad the brief, repo state, and the latest correction.
  2. 2ChangePatch source files in the actual site, not a throwaway mockup.
  3. 3VerifyRun tests and capture screenshots before claiming progress.

WHAT MAKES IT DIFFERENT

  • Starts from real feedback
  • Touches the actual repo
  • Uses memory without leaking secrets
  • Shows proof, not promises
STATUS · Running locally OUTPUT · Site changes PROOF · Tests + screenshots

DESIGN DECISION / PICK ONE

Nothing picked yet — that's the point. The machine doesn't choose.

Pick one. That's the whole loop — AI builds the options, a human makes the call.

METHOD · RUNNING ON THIS SITE

The Lab

I can't describe a design in words. So I have my AI build the options for real and I pick.

  • Live options
  • No mockups
  • Human decides
  • Ships same day

Used to build this page

WHY IT EXISTS

Describing motion in words doesn’t work — “make it feel calmer” means nothing until you see it. So instead of writing a spec, I have the AI build every option as working code, put them side by side, and choose from the real thing. Taste stays human; the labor of building five variants doesn’t.

HOW IT WORKS

  1. 1BuildAI implements each candidate as real, running code — not a picture of one.
  2. 2ChooseI use them side by side on my phone and pick what actually feels right.
  3. 3ShipThe winner goes into the live site, tested and deployed, usually within the hour.
STATUS · In use ROLE · AI builds, human decides OPEN THE LAB

BUILD LOG / THIS SITE, EVOLVING IN PUBLIC

The page keeps
a record of itself.

Every change to this site lands as an entry below — rendered from the repo's own history when the page is built. No hand-authored changelog; the site documents its own evolution.

Entries
26
Last change
Jul 28, 2026
Source
git history

  1. Update

    Revert "feat(site): launch interactive AI career portfolio"

  2. Feature

    Launched interactive AI career portfolio

  3. Fix

    Harden application-ready public surface

  4. Fix

    Rebalance Tocin vignette composition

  5. Fix

    Tall sections stranded at opacity:0 on mobile; threshold 0.12 unsatisfiable

  6. Exploration

    Give each bench card a live vignette depicting its real function

  7. Fix

    Constrain page to 1080px container instead of full-bleed

  8. Exploration

    Showcase The Lab as a method card; fix Lumi node text collision

  9. Fix

    Remove CSP-blocked inline styles; add lint guard for inline style attrs

  10. Fix

    Kill horizontal overflow from sticky bar, unstack margin notes, mobile clipping

  11. Feature

    Added live direction picker for motion + AI presence A/B choices

  12. Fix

    Correct status to TestFlight build 32, ground facts in repo STATUS

  13. Exploration

    Added Equa fair-split demo, replace inventory list with bench grid, tighten copy

  14. Fix

    Replace floating left ledger with sticky section bar

  15. Exploration

    Added sticky section ledger and weighted staggered reveals

  16. Exploration

    Added self-writing Build Log rendered from git history

  17. Exploration

    Added Lumi Hermes exploration card

  18. Exploration

    Added AI signal trace easter egg

  19. Fix

    A11y + responsive fixes; stop tracking dist build output

  20. Exploration

    Expanded Skalable card with story, how-it-works, features, status

  21. Exploration

    Added Explorations feed with interactive Skalable before/after reveal slider

  22. Feature

    Launched Modes of Practice homepage

  23. Feature

    Added focused Billy homepage direction

  24. Feature

    Added three portfolio design directions

  25. Fix

    Pin Netlify Node runtime

  26. Feature

    Launched kinetic Signal Garden redesign

ALSO IN FLIGHT / SHORTER FORM

Other things
on the bench.

ABOUT / THE SHORT VERSION

The medium changed.
The concern didn’t.

I started in communications, learning that the way an idea is explained changes who gets to take part.

That led through design and software into technology leadership: translating between people, possibilities, and the systems that have to hold up after launch.

Now, human–AI experimentation extends the same question: how can new tools increase people’s agency instead of quietly taking it away?

communicationsdesignsoftwaretechnology leadershiphuman–AI experimentation

HAVE A COMPLICATED, HUMAN PROBLEM?

Let’s make it
easier to hold.

wdp@wdpronovost.com