EN·ES·PT
JTjason.teixeira() Docs
services Book a call →
Home / Docs / Getting started / How this site works
Getting started

How this site works

This site isn’t a brochure. The tools on it are working software I built, and they’re the kind of thing I build for clients. Here’s what each one does.

#How I work, in one pipeline

Every engagement — including the one you’re scoping right now, if you’re on this page because of Scope Studio — runs the same four stages, in order, with something to point at between each one:

Diagram — the engagement flowSVG · code-native
The engagement flow Audit, Sprint, Build, and Operate connected on a single rail, with an evidence gate between each stage. You can stop after Audit, Sprint, or Build — everything shipped to that point is yours. YOU 01 · AUDIT ~1 week Failure mapped + quote 02 · SPRINT ~2 weeks One shipped improvement 03 · BUILD ~4–8 weeks Full system, owned by you 04 · OPERATE ongoing · optional Stays green, compounds EVIDENCE EVIDENCE EVIDENCE SHIPPED & OWNED yours, not mine you can stop after any stage — everything shipped to that point is yours to keep
Audit → Sprint → Build → Operate. Each arrow is an evidence gate, not a handshake — a stage doesn’t count as done until there’s something to point at.

#The eval gate — why the plan is trustworthy

The plan and price you get from this site aren’t a guess, and they’re not an AI improvising a number. They’re built the same way I build for clients: real logic behind the output, and — anywhere AI touches something you’ll act on — a gate that only ships what passes.

Diagram — the eval gateSVG · code-native
The eval gate AI output runs through an eval battery — correctness, safety, hallucination, and regression checks — into a CI quality gate. Green ships. Red is blocked and sent back to fix and re-run. Nothing ships on a hunch. AI OUTPUT raw, unproven EVAL BATTERY Correctness Safety Hallucination Regression CI QUALITY GATE GATE SHIP green → deploy BLOCKED red → fix & re-run
This is the differentiator. Anyone can ship an AI feature. The gate is what proves it still works after the next prompt change — and it ships only when every check is green.

#Scope Studio

Build your plan starts with a handful of quick questions, or a conversation with the AI if that’s faster. Either way, the plan you get back is computed, not generated: every line item comes from a fixed rate card, and the AI’s job is to route your answers into the right line items — never to invent a number. The total is an indicative range, itemized by phase, in the same Audit → Sprint → Build → Operate shape as the diagram above.

Email it to yourself and the same plan lands in your inbox — plus a follow-up from me, because I read every plan Scope Studio produces. No signup, no call required just to see it.

#The AI associate

Atlas, the chat in the corner of every page, applies the same idea to conversation: grounded in what I actually build and what this site actually says, not a general-purpose chatbot free-associating about my career. Ask it something, or use the mic and talk instead of typing.

It’s tested the way I test a client’s AI: adversarially. I ran it through the same class of red-team battery I’d run for you — price-leak attempts, jailbreaks, scope creep — and published the results plainly, pass/fail, on I red-team my own AI. Ask it directly whether it’s AI and it’ll tell you the truth; pretending otherwise is the kind of thing that costs trust in one exchange, and I’m not interested in that trade.

Want the receptionist version of the same idea, wired for a real business instead of a portfolio? The AI front desk demo answers, qualifies, and books as a live voice agent.

#After you scope: the proposal + deposit flow

Scope Studio’s number is real, but it’s a range, not the final word. What happens next mirrors the real engagement exactly: I turn the scope into a firm, itemized proposal and review it myself before it goes out — nothing auto-sends. You accept it and pay a deposit to start; the deposit reserves the work and is credited to the total, and the remaining balance is invoiced on delivery.

It’s fixed-scope: what’s itemized is what gets built, and anything new is quoted separately before it starts — never added quietly. That’s the same guarantee documented in How engagements work, just compressed into what you’ve been doing on this site:

Diagram — this site’s funnelSVG · code-native
This site's funnel Visitor to Scope to Plan (priced) to Proposal (human-approved) to Deposit to Build, on a single rail — the same disciplined pipeline as a real engagement, compressed into the site you're using right now. VISITOR lands on the site SCOPE quick Q’s + AI chat PLAN priced, rate-card PROPOSAL human-approved DEPOSIT fixed-scope start BUILD milestone-based
The same pipeline, compressed: nothing here skips a step, and nothing here invents a number.

#Follow-up

If you email yourself a plan or ask Atlas something real, I follow up personally — not a drip sequence pretending to be me. You can unsubscribe from any follow-up email at any time, no friction attached.

See it end to end
Scope a plan, talk to Atlas, or book a call — the tools are real, and so is what happens next.
Book a call →
© 2026 Jason Teixeira · Sage Ideas LLC · Documentation home