I just redesigned my website.
Eight main pillar pages. Multiple rounds of iteration on each.
New typography system, rewritten copy, restructured layouts, refined testimonials, updated resource cards with real imagery.
Total cost for design: $0 in agency fees.
Zero Figma files. No 6-week timeline with a creative director.
I used Claude's Cowork mode, and here’s how I did it.
What I Actually Built
This wasn't a logo swap or a color tweak. I redesigned:
Homepage (completely new structure, universal new footers, top nav)
About page (added new horizontal credentialing bar)
Advisory page (new testimonial layout with photos, 2x2 services grid)
Newsletter/subscribe page (stripped nav, new byline, rebuilt testimonial section)
Resources page (pulled real hero images from each deck's landing page)
Blog page (new tagging structure)
Hiring OS page (completely new)
Mixtape page (completely new)
Each page went through 3-5 versions.
It is absolutely not perfect.
Will I find something inevitable f*cked up on mobile, a responsiveness issue or something broken?
I’m certain of it.
But it’s 90% there, and that’s good enough for me.
The Process
Here's what the workflow actually looked like:
1. I started with what I had.
I dropped my existing site HTML into Cowork along with my brand guidelines PDF and a mood board of stuff I liked. Gave it context on fonts (DM Serif Display for headings, DM Serif Display for subheads, Inter for body, Space Mono for UI elements), colors, and general vibe.
2. Cowork generated full-page HTML prototypes.
Not wireframes. Not mockups in some design tool I'd need to learn. Fully styled, single-file HTML pages I could open in a browser and see immediately. Every CSS variable, every font weight, every layout decision — rendered and clickable.
3. I iterated like I was sitting next to a designer.
This is the part that surprised me. The conversation was natural:
"Make the byline font 0.5 smaller so it competes less with the subhead."
"These CTA buttons need to be bottom-aligned across all cards."
"The testimonial names are floating at different heights — fix that."
"Cut this copy down by 40%."
Each time, I got back a working page in a minute or two. Not a revision request. Not a Loom video explaining what I wanted. A working page.
4. I used it for things I didn't expect.
Font comparison sheets — side-by-side renderings of different font pairings at various weights so I could make typography decisions visually. A/B comparisons of byline options displayed as a visual widget so I could see them in context. Pulling real hero images from my resource deck landing pages by extracting URLs from the live DOM.
I even used the built-in design-critique skill to get structured feedback on pages — hierarchy, spacing, consistency issues — the kind of stuff a designer would flag in a review. It's like having a second set of eyes when your actual designer isn't in the loop yet.
5. My ops guy implemented the final HTML into Beehiiv.
This was the handoff. I had production-ready HTML prototypes with all the design decisions baked in. My ops person took those files and implemented them into Beehiiv (my site platform). No ambiguity about spacing, fonts, colors, or layout. The prototype was the spec.
What I'd Tell Someone Who Wants to Try This
You don't need to know how to code. But you do need taste.
I'm not a developer. But I've spent 20 years learning from, collabing with, or managing incredible designers. I've picked up a lot about responsiveness, UX/UI, direct response design, and like to think I have a pretty good aesthetic in the first place.
That matters here, because the AI isn't making the design decisions. You are. It's executing your vision at speed. If you can look at a page and say "that spacing is off" or "that font weight is wrong" or "those CTAs need to be bottom-aligned" — you'll fly.
If you can recognize good design but can't verbalize what's wrong or right, you might hit a wall. But if you're a marketer with great taste and no coding skills (like me), this is a game changer.
Start with your existing site.
Don't start from scratch. Drop your current HTML (view source, save page) into Cowork with your brand guidelines. Iterate from what exists.
Be specific in your feedback.
"Make it better" gets you nowhere. "The CTA buttons aren't bottom-aligned and the card heights are inconsistent" gets you a fix in 10 seconds. The more specific you are, the faster you move.
Treat it like a conversation, not a brief.
I didn't write a creative brief and wait. I worked through decisions in real-time. "Should the byline say 'former' or 'Fmr'? Show me both." That kind of thing.
Have someone for implementation.
Unless you're building on a platform where you can paste HTML directly, you'll need someone to take the prototypes and put them into your CMS or site builder. But that's a much cheaper, faster job than "design + build from scratch."
The Math
A proper website redesign — 8 pillar pages, multiple rounds, custom typography, copy rewrites, etc, could cost $10-25K with a decent agency.
More if you're picky (and I'm picky).
Timeline: 6-12 weeks.
I did this across a handful of sessions. The implementation cost was my ops guy's time, which was dramatically cheaper because he had exact prototypes to match rather than interpreting a designer's mockup.
But the 80% of design work that's "try this layout, no move that, make this bigger, align these" — that's exactly what Cowork handled.
The Bigger Point
The unlock isn't "AI can make websites." The unlock is that the feedback loop between idea and visual output collapses from days to seconds.
Every design decision I made — from font pairings to testimonial card layouts to byline copy — I could see rendered in context immediately. No waiting for a designer's calendar. No playing telephone through a project management tool. No three-day turnaround on "can you move this 10 pixels to the left."
That speed changes how you think about design. You explore more options. You iterate more aggressively. You make better decisions because you can actually see the tradeoffs instead of imagining them.
If you're a founder or operator sitting on a website you know needs work but you've been putting off because the agency quotes are insane and you don't have time for a 3-month project — this is worth trying.
Drop your site into Cowork. Start talking to it like you'd talk to a designer sitting next to you. See what happens.

If you liked this blog post, you might enjoy this other one on i why i buy high & then buy even higher.
Or sign up for my weekly-ish newsletter, Signal // Noise, where I share my honest notes on work, startups, life, and money. No bullshit or spam!

Looking For More?
—> Here are 3 ways I can help you:




