We paid $25k for our website. I vibe-coded a new one in 2 days.
Last year we hired a design agency to build our marketing site for @Basedash. They did an incredible job. The headline makes it sound like I'm dunking on them, but I'm not. The site was genuinely great. They built it in Framer so we could manage content ourselves, which was a completely reasonable bet at the time (and something we explicitly asked for).
But over the past few months it’s been begging me to migrate it over to code. Every time we wanted to make a change beyond just editing text, we hit the limit of our Framer capabilities. I’m a coder, I don’t know how to use Framer. The friction was too high, so the site didn’t get updated for months.
Last week I made a bet. I'd take our entire marketing site, all 865 pages including blog posts, case studies, and landing pages, and rebuild it in code from scratch. I gave myself two days.
Day one was structure and content. I had the AI scrape our existing site, extract the page hierarchy and collections, and recreate the skeleton. Then it pulled in all the text and assets. By the end of day one I had 865 unstyled pages that were structurally correct and content-complete.
Day two was styling. I had a good base to work from, so I just iterated on the homepage until it matched our brand, then pointed the AI at all the other pages and told it to propagate. Lots of iteration later and it was done.
The $25k comparison is the clickbait version of this story. The real value was after we hit parity with the old site. The first thing I did was build a new page showcasing our embedded BI feature. This alone will probably close us $25k in new MRR this month. Then we knocked out a bunch of SEO, accessibility, and performance improvements that had been sitting in the backlog for months. Vibe coding that kind of stuff is 1000x easier in code than a visual builder.
A year ago, buying a Framer site from an agency was the smart move because code meant maintenance and developer bottlenecks. But now code is the low-friction option. Visual builders, which were supposed to be the easy, accessible option, are now the thing that requires specialized knowledge to use.
865 pages. 2 days. And for the first time in months, I'm actually excited to make changes to our marketing site.
All vibe coding was done in @Cursor with GPT-5.3 Codex.
The new site is built in @Astro.
Check out the old site here: old.basedash.com
And the new site here: www.basedash.com
Replies
Oh bro...both old and new site look great! Curious what you used to build that. Most mockup tools leave emojis inside haha. If im not careful I end up spending hours tweaking design elements. a complete waste of time. But yea basedash is the look and feel of the future. Need something to be able to produce a deign like that consistently.
Basedash
Appreciate it! I wish I could say the AI did all the design work but I still had to guide it pretty closely and describe how I wanted it to look. Now that it has some references it's a little better at building new pages in the same style, but still needs hand-holding.
@maxmusing Agreed. Have you messed around with MCP servers yet? Thoughts?
This is super impressive. Not just because of the visual design but the new site loads faster and performs better. I ran a quick SEO checker (using a free tool) and the new site scored a 86/100 while the old one scored 54/100 - I didn't bother looking at the details but it's still a massive jump.
Clearly @Astro did a fab job.
Definitely interesting.
Basedash
Performance was a big goal of mine! Visual website builders don't give you the control you need to really optimize performance, accessibility, SEO. Much easier to do in code (especially now with AI).
TinyCommand
@maxmusing Did it also impact your AEO? I'm curious.
Basedash
The entire web development ecosystem is about to experience a fundamental change.
hey @maxmusing This is such a clear example of how the bottleneck has shifted.
A year ago, no-code/visual builders reduced dependency on developers. Now, with AI in the loop, code is the faster iteration layer. The friction isn’t writing code anymore — it’s being constrained by a tool’s abstraction.
What stands out isn’t the 865 pages in 2 days. It’s what happened after parity: new revenue pages, SEO wins, accessibility fixes. That’s the real leverage.
Feels less like “no-code vs code” and more like: which layer gives you the fastest path to continuous improvement?
I had the same realization but with a mobile app. Zero coding background, built a full iOS + Android app with AI in a few months. The old bottleneck was "you need to learn to code first." Now the bottleneck is just clarity of thought - knowing what you want to build.
Your point about what happens after parity is key. That's where the real value compounds.
@virtualviki Which AI did you use for building the app and do you mind sharing your costs and build time with the AI? I have a web background but looking to build a mobile app for a new idea for the first time.
@custominstall Claude Max ($120/mo) + Figma AI ($30/mo). Around $150/month total.
Timeline: 4 months of building, now 2 months live in both stores.
Stack: Capacitor - one codebase for iOS and Android. If you're coming from web, it's a smooth transition.
Honestly the hardest part isn't the code - it's marketing and getting users.
@virtualviki Thanks. Saving this one. For now, I am hoping for feedback on my landing page, before I spend more time actually building the app.
@custominstall Happy to take a look! Drop the link
I believe that although there are tools to speed up how we educate ourselves, those who think they are insignificant are using AI to create websites. Those starting businesses for the first time, myself included, are overwhelmed by numerous ways to promote ourselves. Not all businesses have people who can code as their primary form of communication. I have met friends who want to set up bakeries and nurseries, but who fear using these tools. I would encourage the community to share, not just here but also within our own communities, the tools we find or are discovering, especially since for many, keeping up with this rapidly evolving landscape can be quite overwhelming.
@maxmusing, I admire you bringing this to the forefront of conversation. How can we create a framework chamber for those who could be left behind? I am not a coder, but a researcher and design practitioner.
UXDesigner.top
Huge congrats on this, seriously. The new site looks fantastic.
I’m in a very similar phase right now. This weekend I finally carved out time to pick back up a side project (BenchCanvas) I’d been sitting on, and in just 3 days I managed to: research the space and competitors, write a PRD for an MVP, put together a launch plan, and draft branding + domain + SEO notes, then design, build and ship a waitlist landing page
A year ago this would’ve looked totally different for me. I probably would’ve defaulted to Framer (I use it on other projects) and it would’ve taken weeks, not days. The vibe-coding workflow really changes the pace.
I’m aiming to hit 200 people on the waitlist before I start building the actual product, and I’m already close to 70 in the first 2 days. Let's see if I can get to 200 😅
Where did you promote the landing page ? I have a landing page that I am hoping to test out myself.
BenchCanvas is a nice idea. I can see it cutting down the time for web agencies sharing thoughts on websites. Is that your primary market?
UXDesigner.top
@custominstall I shared my landing page on Reddit, a few Discord and Slack communities, and my LinkedIn.
My primary target market is PMs and product designers, but the tool is also useful for marketers, researchers, and founders.
@david_martin_suarez Thanks. I would appreciate if you can share what Discord and Slack communities you utilised. Thanks again.
UXDesigner.top
@custominstall On Discord: Friends of Figma, Design buddies, Lovable... On Slack: Mindtheproduct, Product School... Communities like those ones
I think a lot of agencies and consultancies will be in trouble with the progress in AI. I recently built my landing page ( https://zest-ai.netlify.app/ ) with Claude. Everything from the logo to design was built with Claude and I am amazed by the results. Although, you have to know what you are looking for. Some of the early logo designs were too complex and horrendous.