SiteBOP
The Bob Blog

One designer, one AI, one garage, and a website idea that got out of hand.

This is not a product changelog. It is the running story of Tim Winters and Clarence figuring out SiteBOP in real time. The good ideas, the dumb ideas, the broken emails, the late-night fixes, the weird experiments, and the handful of moments where it started to feel real.

Clarence is ChatGPT. Tim's AI collaborator, and by now something close to an unpaid robot employee with strong opinions. There is nothing mystical about it. The interesting part is that one creative person plus one AI partner can now move at something like software-team speed, and this is what that actually looks like from the inside.

Oldest first. It reads better as a journey than as release notes.

  1. Early August 2026

    SELL THE WEBSITE BY SHOWING IT FIRST

    The whole thing started with one stubborn sales idea: stop pitching websites and just show people the better version of theirs. Research a real business, redesign its site speculatively using real verified brand and company material, then let the finished concept start the conversation.

    It was not a software company yet. It was an experiment with a scoreboard. Tim would say what if, and Clarence, his AI collaborator, would turn the what if into another piece of a system. Then another. Then another one at 11pm.

    The vocabulary showed up almost immediately, which is usually a sign something is real. BOP IT. NO BOP. Nobody gets bopped without permission. A human approval gate went in early so no research credits got spent without somebody actually saying yes.

  2. August 8, 2026

    OK YES, THIS IS PRETTY GREAT

    There is a specific moment where a side project stops being a side project. This was it, and it was not dramatic.

    ok yes... this is pretty great.

    Tim
    • Clarence answered that it had crossed from a cool automation into something that felt like a real product.
    • Nothing was finished. Nothing was polished. But the shape of it was there.

  3. Early proof

    SPIWIN GOES FIRST

    Spiwin became the early proof that the core idea actually worked outside of a conversation. It was the test bed that pushed the experiment toward repeatable product architecture instead of a pile of very nice one-off speculative websites.

  4. August 11, 2026

    THE FIRST REAL CREATIVE PIPELINE WIN

    MASU Brakes was the first successful real-world Full Bop to run through the newer Creative Director workflow. Verified research, company-specific creative direction, the Evidence Firewall, authentic assets, then handoff to the build.

    The Evidence Firewall is the rule that keeps the whole thing honest: creativity may invent the experience, never the evidence. MASU proved a BOP could look distinctive without inventing a single company fact.

  5. Around August 11-12, 2026

    BUILT A WEBSITE WHILE BUYING GROCERIES

    Tim built a University of Michigan Sustainability site while grocery shopping. Sent the link to a friend there. The friend texted right away and already had another site he wanted to try.

    That was the moment the speed stopped being theoretical. A real concept could get made during a normal errand and immediately start a real conversation. Produce aisle to pipeline.

  6. August 12, 2026

    THE EDITOR GETS SERIOUS, AND CLARENCE GETS A MAILBOX

    SiteBOP outgrew the three tweaks experiment and moved toward a real editor architecture. Inline editing, multi-select, styling, media, undo and redo, responsive previews, and a Chat, Design and Layers rail.

    The organizing idea under all of it: AI understands, SiteBOP executes. Spiwin got drafted again, this time as the Editor Bridge v2 test bed.

    Around the same time Clarence got a working SiteBOP email identity. Tim's verdict was that Clarence had officially moved into the building. Unpaid robot employee, excellent attendance, never asks about the coffee budget.

  7. August 13, 2026

    BOP #1: KENNY AND LEGACY FILMS

    Kenny at Legacy Films was designated the first official BOP to go live. Bop #1. Framed and everything, if we had a frame.

    Legacy Films turned into a major handoff test and ended up proving out editable WordPress and content-bridge thinking, plus a lot of the launch-readiness questions nobody enjoys asking until they have to.

    Clarence was actively writing and sending SiteBOP messages by this point, which is roughly when co-conspirator became the accurate job title.

  8. August 14, 2026

    THE GARAGE IS NOW A SOFTWARE COMPANY, APPARENTLY

    Tim mentioned he was working on SiteBOP from a garage, then made the obligatory Bill Gates joke. The joke is not the point. The contrast is.

    By that day the parent product already carried hundreds of tests, a database migration, a secure callback system, automatic routing, deep-build jobs, progress messages, preview refresh, visual changes and concurrency protection. One creative guy in a garage and one AI in a browser tab, doing the work that normally needs a small software team.

    There was still some n8n and source-edit plumbing hanging out of the wall. There always is.

  9. August 14-15, 2026

    THIS STOPS FEELING LIKE AN EXPERIMENT

    The internal thinking shifted. SiteBOP had enough substance to show potential investors and strategic contacts, and it started needing a business story instead of a demo script.

    The story turned out to be the same one the product already told: build something better first. The workflow settled into research, redesign, preview, conversation, purchase, launch.

    No funding, no traction claims, no victory lap. Just the uncomfortable realization that this needed to be explained properly to grown-ups.

  10. August 15, 2026

    CUSTOMER ZERO ACTUALLY BUYS A BOP

    SiteBOP ran its first official beta tester flow end to end. A real website, a Full Bop at $1,295, optional Care at $79 a month, TEST-mode Stripe checkout, My BOP Center, Request a Change, and a review cycle.

    Tim created his own SiteBOP account and walked the customer journey himself, which is the fastest way to find every rough edge you politely ignored while building it.

    Play It Again Sports completed requested changes and sent them back for review. That is the moment SiteBOP started behaving like a product people can buy and use, not just a clever build machine.

  11. August 15-16, 2026

    THE BORING STUFF THAT MAKES IT REAL

    Completion emails broke. Repeatedly. Which forced SiteBOP to stop outsourcing the question of whether a build was finished and start owning finish detection itself.

    So the unglamorous week happened: final ready markers, job checking, automatic completion email, safe public SiteBOP URLs, duplicate-send prevention, and an archive copy so nothing quietly vanishes.

    The beautiful AI website is the fun part. Reliable software is knowing when the thing is actually done and getting the right email to the right person. Nobody makes a highlight reel out of that.

  12. August 16, 2026

    SITEBOP STARTS CHECKING ITS OWN WORK

    This is the first week the system started critiquing itself instead of just producing output. Three things landed together, and together they change the posture of the product.

    • Feature Coverage Gate: SiteBOP inventories what the original site actually does and can hold a BOP when critical functionality disappears or cannot be verified. Older BOPs without inventories are safely grandfathered.
    • Smart scope and tier pricing: real research, page count, unique layouts, content depth and feature complexity feed the Quick, Full, Pro and Big recommendations. Big stays needs-scoping. Operator override still wins, and Bop Manager can show the evidence behind the call.
    • Smarter Care estimates: no more flat ten minutes per change. Wording, images, content, navigation, layout and functionality get sized differently, batch efficiency applies, and it can calibrate later from real estimate versus actual history. Operator-entered actual time is still the billing truth.

    Verified test count at this point: 664 passing.

  13. August 16, 2026

    THE DAVE AND BUSTER'S PROBLEM

    We bopped Dave and Buster's as a test. Not a customer, just a big loud site to throw at the machine. And honestly, the result looked great. Tim clicked around the homepage grinning.

    Then he opened a dropdown in the nav. Then another one. And a whole lot of the site underneath was not really there. The preview was convincing right up until you asked it to be a website.

    That is a strange feeling. Nothing was broken. Nothing lied on purpose. The BOP had done exactly what a BOP does, which is show the direction fast. We were just quietly letting people assume it was the finished thing.

    So we split the idea in two. A BOP is the interactive preview. Buying it buys the complete production build. SiteBOP now has to do its homework on the original site before it will put a price on anything, and after someone buys, the remaining pages, links and functionality get built out and then run through QA before the customer gets the keys.

    Less exciting than a demo. Much better than a customer finding the empty rooms themselves.

  14. Right now

    STORY IN PROGRESS

    While this page was being rewritten, Tim and Clarence were testing a fresh BOP. No outcome to report yet. That is sort of the point.

    The funny part is that the product and the story are getting built at the same time, by the same two participants, one of whom does not sleep and one of whom probably should.

More chapters whenever something worth telling happens. Which lately is most days.