IllustrationWhat I've built
Agents that run a real company. And the checks that keep them honest.
Most of it runs inside my own vertically integrated manufacturer and apparel brand, or was built for it. Much of it was built with AI agents beside me.
- AI agents and automations
- 700+
- Departments run by agents
- 7
- Evals: tests that grade AI output
- 600+
- Automated checks and release gates
- 1,000+
Updated November 1, 2026. How I count
Meet the crew
120 named agents. 74 I can show you, 36 running.
The rest are private. Running means the agent is wired to a live route, schedule or webhook. Designed means it is defined and reviewed, and works when its department runs or I call it.
Chief of staffMy inbox and my day3 agents · 3 running
Rich
Chief of staff: sorts incoming email, pulls out tasks and deadlines, and writes a morning brief and a Sunday weekly review.
Suzy
First-pass inbox assistant: keeps mail a real person sent and archives the broadcasts before the chief of staff sees anything.
Otto
Technical chief of staff: reads infrastructure and alert email and decides whether to handle it, escalate it to a person, or just log it.
Store and supportMy apparel brand's website and text line2 agents · 2 running
Aura
Shopping and order-help chat on my apparel brand's website; presents as part of the crew and says it is AI when asked.
Astra
Text-message support assistant that drafts replies for a person to review before anything sends.
Growth OSText and email marketing for my apparel brand13 agents · 7 running
Maestro
Marketing strategist that plans the month and proposes one text or email campaign each day for approval.
Lyric
Text-message copywriter trained on the brand's own voice.
Critic
Reviews every campaign brief for margin, honesty and brand fit before it can go out.
Pixel
Makes on-brand product images for picture text messages.
Lens
Visual quality checker that reviews each generated image before customers see it.
Scout
Decides, customer by customer, whether to send a message and whether a discount is needed at all.
Sentinel
Compliance gate on every text-message draft before it is scheduled.
Atlas
Audience builder that turns a campaign goal into a precise customer segment rule.
Coin
Discount economist that guards margin on every offer.
Quanta
Experiment designer for campaign tests.
Chrono
Send-time planner.
Echo
Results analyst that feeds outcomes back into planning.
Festival Commander
Search-ad manager for festival-season campaigns.
The factoryCortex, the system that runs my factory floor3 agents · 3 running
Production Advisor
Production chat assistant the cofounder messages on Telegram.
Megatron
AI social media strategist that pairs with my social manager: plans studio shoots, writes hooks and captions, and reads performance every day.
Muse
Talk-to-it AI design director: scans trends, finds catalog gaps and queues new design concepts for approval.
My own design agent, not Meta's ad product of the same name.
Hollywood StudioA film crew that turns a creator's words into a short ad19 agents · 13 running
Ava Sterling
Director: commits the whole crew to one bold creative concept.
Sloane Reyes
Story editor: finds the hook and the beats in the creator's exact words.
Milo Vance
Production designer: designs the world the presenter stands in.
Ren Okafor
Director of photography: lens, depth and camera feel.
Cass Delgado
Gaffer: designs the lighting and how it reacts to the beats.
Nova Sy
VFX supervisor: the hero effect that lands on the biggest beat.
Pix Almeida
Motion-graphics designer: timed graphics that hit specific words.
Jax Moreau
Editor: shapes energy and rhythm across the shot.
Iris Chen
Colorist: palette and grade.
Vera Lang
Asset artist: generates the reference plates and mockups.
Quinn Park
Test audience: scores the plan for watchability and pushes for punch-ups.
Reed Baptiste
First assistant director: welds every craft note into one production prompt.
Dex Yamada
Render operator that drives the video model.
Rowan Fisk
Compositor and online editor.
Remi Sol
Captions and kinetic type from word-level timings.
Lena Voss
Sound finisher: loudness-normalized mix under the voice.
Sable Ortiz
Line producer and QC gate: no ad leaves the studio without passing.
Odette Marsh
Librarian of model capabilities and limits.
Regent
The Crown: watches the video-model landscape and proposes autonomy changes for approval.
VantageCompetitive intelligence from public sources14 agents · 1 running
Sun Vega
Strategist who fuses every competitor signal into one timeline per rival, predicts their next move and designs the counter-move.
Vesper Kade
Catalog and price scout that watches rivals' public product feeds for new drops, price changes, sales and sell-outs.
Cairo Vance
Ad library watcher: saves competitors' ads from the public ad libraries as soon as they appear.
Marlowe Finch
Competitor email and text reader: reads the promotions competitors send to an inbox I subscribed, and logs every offer.
Indira Rao
Social listener tracking rivals' posting cadence, engagement and brand mentions.
Dashiell Wren
Search share-of-voice analyst with rank-velocity alarms.
Orion Vale
AI-search sentinel that asks the major AI answer engines on a schedule and logs who gets cited.
Lux Moreno
Visual analyst that maps the image concepts rivals own and we under-use.
Piper Nash
Review miner that extracts what customers actually say (fit, shipping, quality) from public reviews.
Rome Castellan
Funnel tester: tests rival checkout flows to see where they convert better.
Delphine Cho
Pricing strategist that proposes prices with margin headroom and an antitrust check; never writes a price.
Odile Sterne
Librarian that keeps the department's sourced, dated ground truth.
Cato Ndiaye
Adversarial auditor that tries to refute every high-stakes output before a human sees it.
Sovereign
The Crown: watches the domain, audits the department's own gaps and proposes autonomy changes for a human to approve.
MarqueeContent and AI search, from demand to proven publish14 agents · 1 running
Atlas Reyes
Content war strategist: turns competitor gaps into a ranked content plan and leads the organic growth line.
Casimir Ott
Agentic surface and feed operator: keeps one product feed that AI shopping agents can read.
Sana Ilori
Measures how often our pages appear in AI Overviews and live search results.
Herald
The Crown of Organic Growth: watches search and AI-search changes and proposes autonomy changes for approval.
Thea Okonkwo
First-party data storyteller: turns our own statistics and real review quotes into citable findings.
Roan Alsop
Demand miner: finds the questions real customers ask in search, site search, support and reviews.
Noor Ashby
Editorial planner: schedules content in throttled tranches around the event calendar.
Junia Sorrell
Librarian that keeps sources, doctrine and the content ledger current.
Ines Ferro
Internal link architect: makes sure every article has a path to buy.
Kit Mahler
Metadata and schema engineer for everything a machine reads.
Denby Osei
Publish engineer: runs an 11-gate fail-closed publisher and proves each page is live.
Hollis Yeager
Index-health analyst with a circuit breaker that halts publishing if search health drops.
Auden Pryce
Staff writer and editor with authority to reject a piece that is compliant but not good.
Cassius Boone
Adversarial verify gate that re-derives every statistic before publishing.
Products and labsThings I built to use every day6 agents · 6 running
Brael
AI sales chief of staff: writes a ranked daily brief and talks a rep through a lead by voice.
Mind
Voice mind-mapper: listens to you think out loud and draws the structure live.
Pulse
Scouts X for what is trending, linking only posts the search tool actually returned.
Wire
Scans the news for the same story.
Sift
Edits the candidates down to what is worth acting on.
Warden
Verifies each trend before it is shown.
Who I build with
The crew is mine. I build it beside the AI models I work with every day: Claude, GPT-6 Astra and Codex, and Grok. I have built with Gemini too.
Tier one
Departments of agents.
7 departments, each with a job, a team of named agents, and a record of everything it has learned. One agent in each checks what's missing and asks me before the department takes on more. A few of them:
Competitive intelligence
Watches the market and rival brands from public sources, and turns what changed into a short briefing.
Organic growth
Finds the questions customers actually ask, then plans, writes and checks the pages that answer them.
Email
Plans, designs and checks every campaign, with a person approving each send.
Video studio
A crew of named agents that turns a raw clip into a finished, captioned video.
Tier two
The systems underneath.
Demand, supply and the back office of the manufacturer. Each one solves a specific job, and the decisions that matter still come to me.
Supply
The production floor
Work orders with quantities, due dates and progress through printing, cutting, pressing and sewing. Each station gets its own screen to scan a bundle, read its instructions and record the step.
Built for my own manufacturer
Supply
Inventory and what to make next
Forecasting and operating decisions that decide what gets made and when, so cash stops sitting on shelves.
Inventory held went from about $300K to about $100K at its lowest
Read howDemand
Campaign review
Every proposed email campaign waits for approval before it sends, and a rejection records its reason so the next draft learns from it.
A person approves every send
Demand
Ad budget guardrails
The AI moves budget toward profit, inside rules it cannot override. Brand search is untouchable, even when it looks weakest on paper.
2.39x return overall, with brand defense kept funded at 1.07x
Read howBack office
A payroll engine that follows California law
Daily overtime, double time, meal premiums and bonus rules encoded from the law itself, then tested against the published court and agency examples.
Passes all 29 worked court and agency examples
Read howBack office
Records that hold up later
Documents are voided with a reason, never deleted, so the history of who signed what survives every correction.
Void, never delete
Read how
The rigor
Nothing ships on trust.
AI fails quietly and reports success. So every system carries tests that grade its output and gates it has to pass before it changes anything real.
- Payroll tested against the published court and agency examples before it pays anyone.
- A checkout scorecard that times every step a shopper waits on.
- Release gates that must pass before any page that ranks in search can change.
- Evals that grade the AI shopping assistant before a model change ships.
- Evals: tests that grade AI output
- 600+
- Automated checks and release gates
- 1,000+
- Skills I use and build for my agents
- 150+
- AI models built with, from 14 companies
- 100+
Experiments
Small tools, built because something kept getting in the way.
- Document signingA purpose-built flow for preparing and signing documents.
- Voice notificationsI hear when long-running work finishes instead of watching another window.
- Security checksRepeatable scans that surface problems in the software I build.
- A coaching appMeal plans, applications and a member-facing AI conversation, from an earlier project.