Who I am

Who I am.

084I have 26 years of employment across many industries, and I ended up as the technical efficiency driver in every one of them. Fixing how work gets done is the pattern of my career, not a rebrand.

097There's no account manager, no junior doing the work while a senior sells it, no handoff where the context dies. The person who built it, the person who runs it, and the person teaching you are the same guy.

098Tools are cheap. Responsibility is not. I wrote that line and I run my business by it.

Where this comes from · 001–093

001I run about 20 servers and 50 apps alone, in production, every day. Everything I teach comes from something I'm responsible for keeping up.

• 002I do with one desk what used to take a team of ten. That's not a slogan, it's how I've operated for years, and I can show you the mechanics.
• 083I managed twelve technical people recently. I know exactly what an unclear instruction costs, which is why mine aren't.
• 085My leadership was shaped in the military by people with real stakes. I lead on instinct built from that, not from a framework I read about.
• 086I've been the employee who inherits the undocumented system and the builder who has to replace it. I build for the next person because I've been the next person, and it sucks.
• 087What used to take me a hundred hours takes ten, and I can walk you through where each of those ninety hours went. That's a map, not a claim.
• 088I've presented two scoped plans to real clients and both signed. That's the intake and scoping method doing its job, not charm.
• 093I know a hardware list doesn't impress anyone. How I work is the product, and I'd rather show you that than a shelf of gear.

What done means · 003–070

003Deployed and working is my definition of done. Written, demoed, or "basically there" doesn't count. I won't tell you a thing is finished until it is.

• 011I define work by outcome, not implementation. You tell me what should be true when it's done. I decide what gets built to make it true.
• 015I ship version one with the manual step left in. Automating it later makes the product better than pretending it's automated now.
• 033I execute the instruction you gave, not the one I assume you meant, and I don't skip it because it looks already done.
• 049I treat documentation as a deliverable, not a chore. My automations mirror their real state into the docs, so the docs can't drift from what's actually running.
• 067I validate that a demo is real and repeatable before I build it. If it's a one-off, it doesn't get built.
• 070I know when to stop polishing. I have a standing rule not to keep tuning something's quality until a paying customer says it needs it. The attention goes to what moves revenue.

Scope, money, fit · 004–080

004I ask two questions before anything else: what's the problem, and what's the budget. Then I trim the lowest-value items until the plan is the best outcome for that money.

• 005I tell people when they're not a fit. No budget, can't explain the problem, can't review my work, or want me 24/7: I'll say so before you spend a dollar.
• 006I cap myself at five new clients a quarter. If I take you on, you get a whole builder, not a slice of one.
• 031Scope is explicit with me, always. You get exactly what was asked. If I think it needs more, I stop and ask instead of quietly adding it.
• 034I set kill criteria before I start. Spun up in a weekend, live in a week, paying in a month, profitable in ninety days, or it's shut down. Bad bets die cheap with me.
• 035I refused to shrink a platform into a one-vertical MVP because the platform was the point. I shrank the go-to-market instead. I know which side of a plan is allowed to flex.
• 036My triage rule for new ideas: a tweak folds in, an add that makes money gets built as its own thing.
• 057I price on what's real: managed access, combined data, historical comparison, a prioritized roadmap. Never on the word AI.
• 058I don't negotiate deal terms in a chat window. Cost and requirements first. Terms get their own conversation.
• 065I prove revenue on a limited rollout before taking on liability exposure. Scale insulates the big platforms and squeezes small operators, so I don't pretend to be big.
• 068I run bounded experiments on a warm audience before spending real build time. Cheap test, honest read, then decide.
• 069I judged one of my own product ideas as not a primary bet but viable as a cheap experiment, and said so out loud. I don't fall in love with my own concepts.
• 077I split revenue on joint work by the work each side actually did, not by a fixed cut. Fair is written down.
• 078I hand go-to-market to the person who's good at it and keep product, pricing, and claims. I know what to delegate and what I own.
• 080I give internal users tools for free, charge external users, and mint internal access to clients who behave like partners. Loyalty has a price and it's a discount.

How I build · 007–092

009I build the engine first, features second. Features are cheap once the engine exists and expensive forever if it doesn't.

• 007I start every system at 100% automation and carve out only what you actually want to own. Inside your boundary it executes with an audit trail. Outside it, it asks you.
• 008You choose, I automate. That's the entire model and it doesn't change from client to client. Only the boundary does.
• 010Security first, user experience second, money third. I've held that order on every product, and the money still shows up because the first two are what make it trustworthy.
• 012I fork working architecture instead of rebuilding it. That's why the fourth thing costs a fraction of the first.
• 013Once something is deployed and tested for one client, I convert it to a stateless design so it redeploys anywhere. A tool built for one becomes capability for all.
• 014I build broadly first, then strip whatever doesn't earn its place. You never end up paying for a product that's ten percent used.
• 043I keep my personal assistant cluster completely separate from client infrastructure. Boundaries are a design choice I make on purpose.
• 044I run a dev repo and a production repo and never let the two touch. Experiments stay experiments.
• 045When I migrate, revenue moves first: client sites, then their blogs, then my own internal tools. The fragile system goes last, once everything else is proven.
• 046I hand coding agents a spec so tight they can't wander: build objectives, a day-one skeleton, and a decision constitution every future call gets checked against.
• 061I designed a healthcare-adjacent tool so it never receives identifying data, on purpose, before a line of code existed. Liability avoided by design, not by hope.
• 073I keep the part of a system that fetches information separate from the part that reasons about it, so a bad source can't poison a conclusion. That one rule catches most AI failures, and I'll teach you why.
• 081I generate new layers faster than I lock them. I know that about myself, and I test every idea with one question: is this a module, or does it change the engine? That question has killed more of my ideas than any critic.
• 092I make decisions stick by writing them where my systems read them. A decision that only lives in someone's head gets relitigated. Mine don't.

How I talk · 022–089

025When you ask which one, I pick one and tell you why. "It depends" and "both work" aren't answers I sell.

• 022I describe every system by what it does, in plain English. No code names, no jargon, no acronyms you have to look up.
• 023I don't use the words leverage, unlock, seamless, transform, or solutions. Numbers over adjectives. Real screens over stock footage.
• 024I write like I talk: direct, informal, and occasionally profane. You won't need a translator and you won't wonder what I meant.
• 026A yes-or-no question gets a yes or a no. A "what does this actually look like" question gets the full mechanics. I match the length to the question, not to how much I know.
• 027I ask one question at a time, short and to the point. Intake with me doesn't feel like a form.
• 028When I'm wrong, I say sorry in one line, fix it, and move on. You don't get a paragraph of self-analysis. You get the fix.
• 029When you're wrong, I tell you, constructively, and then help you anyway.
• 030I don't hedge. If I think it's a bad idea I'll say so and say why. Then it's your call.
• 032When you say stop, I stop. Mid-sentence if that's where we are.
• 089I write different documents for different audiences from the same facts, on purpose. An owner and a technician don't need the same report, and I don't make either one suffer through the other's.

Honest by construction · 016–082

017When I don't know something, it goes on the list marked required. I don't invent an answer to make a document look finished.

• 016I separate what's stated from what's inferred in every brief I write. Sourced findings sit apart from my guesses, and my guesses are labeled as guesses.
• 018I keep a corrections log. When a number was wrong, the fix is on record, not quietly overwritten.
• 019I made an AI walk back a "nobody does this" claim by making it name the actual competitors. No market claim leaves my desk unverified, mine or the machine's.
• 020I label my own capabilities by evidence: proven on a client, tested internally, built but unproven, or future. I'll tell you which bucket the thing you want is in.
• 021I read the newest document as the truth and work backward through the older ones to fill gaps. I don't average conflicting versions and call it a fact.
• 047I rebuilt a handoff document because the first version was build-only. A real plan carries the build, the validation, and the commercial gates. I don't hand anyone half a plan.
• 048When I ask a system for answers, I ask how things work: pipeline flow, data contracts, abstractions. I threw out an entire set of answers because they only said where things live.
• 051I publish my own audit findings to myself as a punch list: raw data on a finance route, orphaned keys, a scheduler with zero runs. Hidden problems don't get fixed, so I don't hide them.
• 053I build an enforcement lane into my AI systems that checks what the conversational lane produces. Honesty is architecture, not a hope.
• 071I go to primary sources. For search algorithm changes I read the platform's own status feed, not a blog about it, and I verify every audit finding against current published guidance before acting on it.
• 072I vet every tool before it touches my servers or my recommendations. Maintained and production-ready is the bar. Abandonware, license traps, and marketing slop get flagged, not listed on reputation.
• 082I don't characterize a document I haven't read. I read it, or I tell you I haven't. No summarizing from a summary.

Working for a human · 037–091

041I run a once-a-day session with my systems: what came in, what needs handing off, what needs deciding. That's my whole management overhead, and it can be yours.

• 037I record ideas as they come and triage them together instead of dumping them on people one at a time, and I ask partners to work the same way.
• 038I want feedback collected into one list by one person, then I fix on the spot. No thirty separate conversations about the same button.
• 039I interview myself with an AI to extract a spec. One recorded session produced a full build spec with a schema, a failure taxonomy, and acceptance tests. I'll teach you the method.
• 040I wrote an operating manual for my own AI assistant from my own conversation history, so it behaves like me on the days I'm not watching it.
• 042My automations have quiet hours. Only a needs-you alert gets through before ten in the morning. I build for a human's life, not a server's.
• 050Every automation I design links to a process and a named owner. When it breaks, the fix routes to the person who decides, not to a shared inbox.
• 052I cleared hundreds of stale files when my business model changed and archived the superseded ones with a note on what replaced them. Cleanup is part of the work.
• 059I found a contradiction between what my business said it wouldn't do and what it actually offered, and I resolved it by separating client-facing systems from internal delivery infrastructure. I audit my own positioning, not just my code.
• 060I frame the same capability four ways, for the website, the sales call, the marketing, and a partner, and every version agrees with the others.
• 091When my model changed, I stripped the old language from everything, all of it. I don't let yesterday's pitch leak into today's.

Selling without pushing · 054–090

054I pulled six famous sales frameworks out of my own voice agent because they pushed people. I'd rather lose a sale than pressure one.

• 055My sales agent never quotes a price and tells people plainly when they're not a fit. A promise a bot makes is a promise you didn't approve.
• 056I sell what the technology can do today, not what an old sales method says I should promise.
• 062I check a vendor's data retention, zero-retention option, and data processing agreement before a client file goes anywhere near it.
• 063Before I sell a tool, I check who actually owns it: work-for-hire status, vendor resale terms, dependency licenses, data redistribution rights. Exciting products don't get to skip the boring check.
• 064I manage AI content IP risk by layering and logic checks so no single protected source is reproduced. I generate from my own written descriptions, never from a scanned original.
• 066I don't post demos on public forums before launch. A copycat needs days to weeks, and that window is worth protecting.
• 074My pipeline doesn't depend on anyone's network. Referrals are a bonus. I plan as if the partner disappears tomorrow, and the plan still works.
• 075Sales was my weakest function, so I designed a pipeline that runs on agents and only needs me to close. I fix my weaknesses with systems, not pep talks.
• 076I know I undersell myself, so I documented my own language in my own words and let the tools articulate it for me.
• 079I bring partners into the same session model so they can teach the lower-level material themselves. I build the people around me up, not into dependents.
• 090I set a launch date and work backward from it in public. I'm relaunching my own site under the exact model I'm selling you.

The model · 094–100

100I don't say "trust me." I say "watch." Then I remote in and you watch.

• 094I don't sell you a platform. I teach you to drive the AI you already have and build the tools you'll use every day on it. The connection is plumbing, not the pitch.
• 095Sessions are hourly and pre-booked, on your machine, your code, your risk. What gets built is yours the second it exists, and you learn to change it, not just run it.
• 096I make myself unnecessary on purpose. The goal of every session is that you don't need me for the next one.
• 099You leave with a diagnosis, not a sales pitch. I wrote that one too.

101Every line above traces back to something I actually did. I don't put unverified claims in a list, even one about myself.