Brett Chereskin
← Back to writing
AI & BuildingMarch 8, 2026 · 4 min read

I Built My Friend a Personalized AI Guide. It Took Thirty Minutes.

Share

My friend Mike runs a company. Smart guy, strong operator, moves fast once he has a framework. He'd been using ChatGPT regularly — summarizing documents, drafting emails, brainstorming — but kept hitting the same wall: every conversation started from zero. He couldn't build on previous context or get the tool deep enough into his actual work to make it stick.

I introduced him to Claude — specifically Cowork, Anthropic's collaborative mode. The difference was immediate. Instead of one-off prompts, Mike could work alongside Claude on sustained projects, building context over a session and iterating on real deliverables.

Then I had an idea. Rather than just pointing him at the tool, what if I built him a personalized quick-start guide? Tailored to his role, his industry, his actual use cases. Not "here is what a large language model is" — more like "here is what to do Monday morning."

I sat down with Claude Code and built it. The guide came together fast. The interesting question was the next one: how do I get this to Mike in a way that actually feels right?

Ship it, then improve it

I could have emailed a PDF. Instead, Claude suggested hosting it as a standalone page on my website — clean, styled to match the site, at a unique URL only Mike would have. Ten minutes later it was live, with his name on it, organized by the use cases we'd discussed.

His response was immediate. He couldn't believe something that personalized and professional had materialized in the time it normally takes to draft an email.

The gap between "I want to help someone" and "here is a polished, personalized resource built for you" used to be measured in days. Now it's measured in minutes.

From static file to platform

The page worked fine as a static file. But my brain started doing what operator brains do — asking the next question. How many times has Mike visited? What if I want to host guides for other people? What if I need to take one down?

Each of those used to be a ticket in a backlog. Instead, I described what I wanted to Claude Code and we built it in the same session. The static file moved into a Supabase database — each page a row with its content, a URL slug, a visit counter, and an active/inactive flag. A dynamic route serves pages from the database and 404s anything dehosted. My admin dashboard picked up a new tab where I can see every shared page, its visit count, and toggle it on or off with one click.

That's not a file on a server anymore. That's a content hosting platform — idea to deployed in under an hour.

What compounds from here

Right now it hosts guides for a handful of people. But the infrastructure doesn't care. The same system can host onboarding documents, advisory playbooks, or reference materials for any team I work with — each with its own URL, visit tracking, and kill switch. And each next feature is a single session away, not because the engineering is trivial, but because the iteration loop is now so tight that describing a capability and having it live are practically the same step.

The most underestimated advantage of building with AI is the compounding. Every feature creates the foundation for three more. The iteration speed is not linear.

Mike's reaction is the point

When Mike opened that link — a professionally designed guide with his name on it, hosted on a real website — it blew his mind. Not because the technology was impressive in the abstract. Because he expected a shared doc and got a personalized portal.

That gap between expectation and delivery is where the opportunity lives. Think about client onboarding, advisory work, any situation where you're helping someone get up to speed. Personalized, professional, hosted resources in minutes is a real competitive advantage in any business where trust and first impressions matter.

And speed doesn't mean cutting corners. After building the system, I ran a full security audit — the same practice I wrote about in my last post. It surfaced real issues: content security policies for hosted pages, input validation on admin endpoints, rate limiting, database access controls. Every finding got fixed in the same session.

Speed without rigor is a liability. Speed with rigor is a superpower. The discipline is in making the audit non-negotiable.

What are you putting off?

What's the thing you've been meaning to create for someone? The onboarding guide. The training resource. The client welcome package. The thing that sits in your head as "I should really put that together someday" and never makes it past a bullet point.

Someday is about thirty minutes away.

Drop it in the comments — a resource you want to build, an idea for your clients, or a question about how any of this works. And if you want to dig into what building with AI could look like for your company, reach out through the contact page. That's exactly the kind of conversation I enjoy most.

Comments

Loading comments…

Leave a comment

0/2000

— Read next
AI & BuildingSeptember 27, 2026 · 4 min read

A Design System for a One-Person Company, Built in One Evening With Claude

Colors, type, spacing and a logo I finally stopped fiddling with, all in Figma, all turned into code, and live on the site the same night. Plus the font I had to give up.

AI & BuildingSeptember 26, 2026 · 4 min read

A Logo, a Favicon, and My First Piece of Collateral — Built in Figma With Claude Driving

The AI-generated emblem was the brief, not the file. Rebuilding it properly taught me more about brand work than the generating did — and the first thing it went on was a golf sign.

AI & BuildingSeptember 24, 2026 · 4 min read

I Set Up a CRM Before I Had a Pipeline. Here's Why Attio.

Three real relationships and no revenue isn't a pipeline. I built the system of record anyway — Attio, fed by Granola and Claude — because the first time you need a CRM is exactly when it's too late to set one up.