Strategy is the prompt.
Someone has to run the graph.

Ten lines from the posting. Seven receipts and three partials.

A self-scored ten out of ten would be worth nothing, so this is scored the way he would score someone else. The posting runs three layers in parallel. Pick one and the graph below dims every node that does not prove it. Click any row to jump to its proof.

No filter. All thirteen nodes are live.

Nineteen agents in four squads. One person built all of them.

The posting asks for AI-powered automations that raise the operational ceiling without growing headcount. Warren built one to find out how far that idea actually goes, on his own time, with his own project work as the test case. Nobody commissioned it and nothing depends on it commercially. It exists because he wanted to know the answer, and it is the most complete thing he can show you about how he takes a problem apart.

These are the real names, in the order the system lists them, and every one of them appears in the screenshot below. Each agent runs on its own schedule. Hand-offs between them are traced. Cost is tracked per agent, which is the part that matters: the work has a unit price, not just an output, so the question of whether an automation is worth keeping has an answer rather than an opinion.

This is the closest thing in his record to the posting's fourth bullet: dashboards, reviews and inspection mechanisms that turn data into decisions, including the underlying queries, models and tooling. He did not write that tooling by hand, he directed it into existence and corrected it until it held, which is the skill the posting is actually describing when it says reach for code or AI before another meeting. It is also the reason he can claim daily hands-on AI use without hedging. What is on show here is a method, not a venture: how he decomposes a set of functions, decides which are worth automating, gives each one a schedule and a unit cost, and then builds the view that tells him whether it is working.

Six teams sent requests in six formats. Relay reads them before anyone commits.

Relay is the intake node Warren built at Dow Jones. It reads brief health before work is promised: which specs are missing, which background is absent, who has not been named. Then it maps the request against roadmap and capacity, so the conversation with a partner team starts from a complete picture instead of a partial brief. The graph below is its actual shape: six inlets, one triage step, one plan.

The posting asks for someone who identifies a high-friction workflow, prototypes the solution, and rolls it out to the team. Relay is that loop completed inside his employer. Vizor is the same loop run outside it: built on his own time, in real weekly use by him and by a few agencies run by friends. That is a smaller claim than an enterprise rollout and the page is not going to inflate it into one.

He reaches for code or AI before he reaches for a meeting.

Three built solo in roughly a year of nights and weekends. To be exact about how: he is not an engineer and he does not hand-write the code. He specifies the system, directs AI agents to build it, reviews what comes back, and sends it round again, with staff engineer friends at Meta and Google advising. The skill on show is knowing what to ask for and recognising when the answer is structurally wrong. He built them to pressure-test his own thinking end to end, not to found anything. Two of the three are in real production use today: Relay inside Dow Jones, adopted by a team that had other options, and Vizor with him and a handful of agencies run by friends. The third is on private staging, and this page says so rather than rounding up.

PACTO is the deepest of the three and the least glamorous underneath: ESIGN and UETA compliant signing, SHA-256 tamper-evident custody, identity verification and a reusable clause library. None of that is the interesting part and none of it gets built by accident. It is there because a tool people sign things inside has to hold up, which is the same reason a design system needs governance and not just components.

Ninety-four is a countable number because the work was sequenced into phases rather than generated in one pass. That is the difference between building software and prompting it. The point of showing these is not the products. It is the thinking: what to build, what to reuse, what to leave alone, and how to hold a sequence together for a year without a team to hold it with you.

Two years at Google. Two programs with opposite constraints.

The posting asks for problem framing and option analysis through to shipped rollout. Those are two different muscles and most program managers only have one. He ran both at the same employer at the same time.

He was Huge's senior project manager in the Android pod, embedded at Google. He has also sat in the other chair, before and after: stewarding the APS design system into production with Infosys and 28 or more engineers at Vertic, then owning the Monaverse vendor relationship end to end at Everyrealm and bringing in Moving Brands for the NYU Langone audit on a later contract. Being the agency is how he learned what the client sounds like from the other end of the call.

Acquisition, onboarding, upsell, and eight properties becoming one.

UX Program Manager for Brand. He partners with the commerce and acquisition product design teams on all brand touchpoints: customer acquisition and onboarding flows, upsell experiences, and event websites across WSJ.com, Barron's.com and MarketWatch.com. Day to day he operates across marketing, subscription strategy, commerce, product design and event marketing.

A design system is a program, not a library. The two unmigrated properties are the next two milestones with owners and dates, not a rounding error, and this page would rather say six of eight than round to the nearest story. The 80 percent is his own measure of build time before and after, not an instrumented metric, and he would rather label it that way than have it fall apart in an interview. Cross-property design inconsistencies on the migrated properties are gone.

Every output fed the next one.

Oldest first, the way a graph reads. Click a node to load it into the inspector.

M.A. Digital Media Management, Hyper Island and Teesside University. B.S. Advertising Communications and Marketing Management, Sacred Heart University. Clients across the years include Microsoft, Google, SAP, Vodafone, Viacom, GE, Merck, NYU Langone, Novartis and Purina. Thirteen years. New York.

Four people have already written this down.

Everything else on this page is Warren describing Warren. These are the only sentences here written by someone else, so they are reproduced in full rather than trimmed to the flattering part.

Titles are as each person gave them. Contact details on request.