I build things. Products, teams, and now AI systems.

Founder to Builder: What I Learned Building My Own Website in a Weekend

Last weekend (July 25th and 26th) I spent two days in a room in Allentown (plus a live Zoom room full of people dialing in from as far as Los Angeles and Mobile, Alabama) for the first-ever "Founder to Builder" workshop, run by Dr. Jeff Bullock. The premise was simple and a little intimidating: show up non-technical, leave with actual working software for your business. Not a plan. Not a mockup. A real system, built by you, by Sunday.

I'd been working with Dr. Jeff for about two months at that point, and this personal site — the one you're reading this on — is the direct result of what I built that weekend using Claude Code and VS Code.

It wasn't a lecture — it was a room full of people building live

The format was the whole point. A little bit of setup, then hours of hands-on building, with breakout rooms for anyone who hit a technical snag. If your install broke or you got lost, you got pulled into a side room with a coach one-on-one until you were unstuck, then dropped back in. Nobody sat and watched a slideshow for two days.

That structure mattered more than I expected. A room full of complete beginners — a pharmacist, a restaurant owner, a photographer — all building in parallel, all hitting different walls, all getting unstuck in real time. Watching someone else's build succeed (or watching them ask the exact question you were too embarrassed to ask) turned out to be half the value.

The wins in the room were the real pitch

Before anyone touched a keyboard, we went around and talked about what people had already built in the weeks leading up to the workshop. A few stuck with me:

• One team member had cut an audiobook quality-check process from roughly 20 hours down to about 8 minutes.
• Another had built what amounted to a book-publishing assistant — taking a manuscript from draft to a formatted, published book in about an hour.
• Someone else had built a tool that turns a folder of photos into a finished short video in seconds, instead of the hour it used to take bouncing between apps.

My own answer, when it came around to me: using Claude, I'd shaved about an hour and a half off every single BNI one-to-one meeting I do. That's not a hypothetical — that's real time back, every week, from one small workflow change.

Watching a build happen from zero

The core of the day was watching one participant build their business site from scratch, live, screen-shared for the whole room. A few things from that stuck with me:

Read what it tells you. The instruction that came up again and again: don't just let it run and skip past the summary when it finishes a task. Read the plain-English explanation of what it did and why. You don't need to understand the technical commands — but you do need to read the reasoning, because that's how you catch mistakes and actually learn what's happening to your own project.

Memory is what makes it feel like it knows you. A recurring frustration with AI tools is having to re-explain yourself every session. The workshop spent real time on this: giving the tool a written memory file with your goals, your business, and your context, so it stops treating every conversation like the first one. Once that's set up, the difference is immediate.

Security doesn't happen by default — it happens because you ask. Every project, more than once: is this secure, is it exposing anything it shouldn't? Ask directly and it gets fixed. That question is now a standing part of how I build.

Then we put music on our own websites

Near the end of the day, the group worked through adding original AI-generated music to their new sites — download the track, tell the tool where to find it in your downloads folder, ask it to wire up a player, then ask for it to actually work when clicked instead of just sitting there. Small thing, but it's the same idea as everything else that day: describe what you want in plain language, read what comes back, ask again until it's right.

If you've clicked play on anything in the music section of this site, that feature exists because of that exact exercise.

What I'm taking into the rest of my build

The workshop closed with a simple piece of advice that's stuck with me since: you don't know what you don't know, and that's fine — ask the question when you hit the wall instead of pretending you understand. That's basically how this entire website got built, one honest "wait, what does that mean?" at a time.

PRISM has set the dates for the next Founder to Builder: September 22–23, 9:00 AM to 1:00 PM each day. I'd highly recommend it — and it's worth repeating: you don't need any prior AI skills to attend. That's the entire point. Founder to builder, not coder to builder.

If you're a small business owner sitting on an idea for a website, a booking tool, or some workflow that's eating hours out of your week every month — this is exactly the kind of gap PRISM AI Consultants exists to close. Book a free session and let's talk about what's possible for your business.

I've been working through an ad account audit for a client lately — the kind of project where you're reading reports and proposals full of terms that sound intimidating but usually aren't, once someone explains them once. I asked Claude to walk me through a handful of them as they came up, and it seemed worth writing down here in case you run into the same terms and want the plain-English version instead of the marketing-speak one.

Meta Ads audit

A review of a Facebook/Instagram ad account to see how it's actually performing and where money might be getting wasted — things like account structure, who the ads are being shown to, whether the creative is stale, and whether the tracking is even set up correctly to measure results in the first place. The output is usually a written report: what's working, what's broken, and what to do about it.

One thing worth knowing: a "free Meta Ads audit" offer is also a common way agencies get a foot in the door before pitching ongoing ad management. The audit itself can still be genuinely useful — just worth knowing the incentive behind it.

Native bot filtering

Shopify now tags every visit to a store as either a real human or a bot — crawlers, scrapers, and other automated traffic that was never going to buy anything. Turning that filter on means your traffic and conversion numbers reflect real visitors instead of noise. It's a built-in setting, not a paid app, and it only applies going forward, not retroactively.

Microsoft Clarity

A free tool that shows how people actually behave on a site, not just how many showed up. It includes session recordings (literal replays of a visitor's mouse movement, scrolling, and clicks), heatmaps, and flags for "rage clicks" — when someone clicks repeatedly out of frustration because something's broken or unresponsive.

Heatmap

A visual overlay on a webpage showing where people click, scroll, and look, using color instead of raw numbers — red or orange for a lot of activity, blue or cool colors for little to none. Instead of watching visitor sessions one at a time, a heatmap aggregates many visitors into one picture, which is a fast way to spot something like a button nobody's clicking or a section of the page most people never scroll far enough to see.

None of these are complicated once you've heard them once — that's most of what I keep relearning about this AI-assisted way of working. Ask the plain question the moment you hit an unfamiliar term, and it stops being intimidating. If you're staring down a pile of marketing jargon of your own, book a free session and we'll work through it together.

One framework I keep coming back to when I'm prompting Claude is what I call the five stages for prompting. It's not a fill-in-the-blank template so much as a way of thinking through what you're actually asking for before you type it. Here's the breakdown, mostly the way I say it out loud when I'm working through it myself.

1. Give it a role. Which expert (or experts) do you need? Think of Claude as a team of PhDs waiting on your instructions — real experts on just about everything.
2. Give relevant context. Introduce yourself and the situation. Your team of experts doesn't know who you are yet, so this is where the files, the background, and the situation go.
3. State the problem. Be clear on the what and the why — explicitly not the how. People have been solving problems their own way for so long that they unconsciously constrain the model with their own process assumptions. There might be a hundred ways to do it — don't constrain it to yours.
4. State the goal. The big goal, not the next small step. People default to thinking small and describing one step toward the goal instead of naming the actual destination. Think big.
5. Give the prompt or command. The actual deliverable — a chart, a table, a strategy, whatever you're expecting to get back.

Not all five carry equal weight. Role is the one you can most often skip without hurting the output — leave it out and you'll probably still get what you need. But I still say it out loud every time I'm working through this with someone new. The point isn't that every step is equally load-bearing — it's building the muscle memory of the full framework, even on the step that matters least.

If you want to run through this on your own prompts instead of just reading about it, book a free session and we'll work through it together.