PRISM ONLINE · BUILD WEEK 10

Real products, real checks, no hand-waving.

I build AI
products.
They ship tested

I'm Rodney John. I build AI products with PRISM, a rules-first build system that turns plain-English decisions into working software. The proof is public here: shipped products, live build records, and checks tested in the browser before anything ships.

PLATE 01THE ENGINE

How the products get built.

five stages, one rule: nothing ships until the checks pass.

PRISM is the build system behind the products on this site. I start with the decision: what the thing should do, who it is for, and what would make it worth shipping. The system turns that into research, design, code, and tests.

I wrote my own rules because most AI build workflows assume the person driving can read code and spot problems early. I can't. So verification is part of the system: PRISM opens the product, clicks through it, tests it on desktop and phone, and nothing ships while a check is failing.

Every product here moves through the same 5 stations:

01
Principles
Research first, then a working prototype. I approve the direction or the idea stops here.
02
Requirements
The approved prototype becomes an exact list of what the product must do. A feature that can't be stated precisely doesn't get built.
03
Interface
The look gets locked in: exact colors, exact type, down to the spacing. The build can't drift away from the design I signed off on.
04
System
The unseen part gets planned: where data lives, how the pieces connect, and what the product does when something goes wrong.
05
Make
The code gets written, then the product gets used for real: buttons clicked, pages opened, desktop and phone.

The panel below is the receipt. Every number comes from the build records, not from a marketing deck.

PRISM // TELEMETRYSYNCED
P
Principles
what must be true to feel right
R
Requirements
what the product needs to do
I
Interface
how people will experience it
S
System
what needs to exist underneath
M
Make
turn the plan into the product
Builds shipped
11
verified with gates
counted from .prism.json stamps
Automated checks passed
672
tests a build must pass to ship
summed from stamp gates blocks
Project types built
5
saas · games · tools · sites
counted from .prism.json stamps
Integrations wired
5
auth · stripe · email · db · sentry
counted from .prism.json stamps
Lessons captured
388
fed back into the system
counted from the improvement backlog
Last shipped
JUL 26
plumb · 407 checks green
counted from .prism.json stamps
Between builds
IDLE

last shipped: plumb, 407 checks green. the next build appears here when it starts.

Build streak
10 weeks
every week since the engine turned on
26 weeks■ = a week with a ship or a fix
PLATE 02THE SHOWROOM

Products you can inspect.

3 builds, every one open to inspect. 1 has a price. start anywhere.

Affix signature editor with a populated live preview and healthy Gmail size gauge
LIVE · TOOL

Affix

"Gmail kept breaking polished email signatures. Affix fixes the paste."

shipped: 2026-05
Single HTML fileNo accountNo subscription
Program Needs Assessment Starter welcome screen with ready-made, past assessment, and custom assessment choices
LIVE · FREE TOOL

Program Needs Assessment Starter

"Program teams can see several concerns at once but still leave a review without a clear priority, owner, or deadline."

checks: 26 greenshipped: 2026-07
Tailwind CSSLocal browser storageLocal PDF exportNo account
Plumb performance dashboard as an administrator, showing goals on track, at risk and off track
CLICKABLE DEMO · EXPERIMENT

Plumb

"Google says its tools are everything an organization needs to run. I took that as a dare."

checks: 407 greenshipped: 2026-07
Google Apps ScriptGoogle SheetsGoogle DocsNo serverNo database
Free to exploreOpen Plumb
PLATE 03THE SHIP LOG

The receipts stay visible.

ships, fixes, mistakes, and lessons. useful beats polished.

JUL 26 00:00SHIPPublished the Plumb clickable demo: intake and monthly performance for a program with several sites, running on Google Apps Script with no server and no database. 288 offline checks and 119 browser journeys green. build page →the real one stays private, so the demo has to carry the proof.
JUL 18 00:16SHIPShipped the redesigned Program Needs Assessment Starter with an evergreen decision workspace, local PDF export, and 26 verified checks. build page →The breakthrough was treating the interface as a product system, not another styling pass.
JUL 03 17:24NOTERodneyBuilds v2 passed all compiler gates: 71 automated checks green across unit, behavioral, accessibility, performance, links, and visual regression. Live on a preview URL for the owner's review. build page →the site verified itself before asking to ship.
JUL 03 09:41STARTRodneyBuilds v2 entered the pipeline. This site is rebuilding itself in the open. build page →the page you are reading is the build log.
JUL 02 14:20SHIPNextRound, a full SaaS with auth, Stripe, and email, passed 118 automated checks and stamped verified.6 repair loops. the compiler earned its keep.
JUL 02 11:05FIXStatic analysis was flagging generated output folders as source code. The build template learned the exclusion permanently.
All entries
PLATE 04THE FRIDAY BUILD LOG

One real AI build.
Every Friday.

See the plan, decisions, mistakes, fixes, and working result. Issue #02 follows a real program-improvement build and includes a free needs assessment you can use locally.

one email each Friday · unsubscribe anytime · the free starter needs no account