Hacker News Reader

Comment by nibab | Hacker News Reader

Great work! I think the barrier to entry here is going to gradually become smaller and smaller (as models become better and we see more multi-modal models) and the big differentiator here is that you will have lots of open source developers building functionality that is relevant to them (vs keeping this closed-source like Vercel).

I am curious how much of the functionality in this area is "write code that does scaffolding and prompt engineering" vs "finetuning and model-level improvements".

View on HN

I think this kills the CMS market—-companies like Contentful & Sanity that charge $400/mo+ just for a team to be able to edit their own website. There’s been a lot of money made (and web stack complexity added) because of how much domain-specific skill is required to created and edit html content.

The only missing piece is direct framework integration (Next, Astro, etc) but at the rate of development I've seen for these projects we'll have that by next week. Then you have an intuitive site builder that operates directly on the page code just like another dev on the team.

One thing Vercel got really right is being able to click on any element in the page for it to become the context of your prompt. That way, you don't have to type "change the top icon inside the third pricing box to red". Also you'll get faster responses because you don't have to feed the AI the entire page / component each time.

Yeah - this space is moving really fast. Codesandbox (https://codesandbox.io/blog/meet-boxy-ai-coding-assistant) released the click on any element functionality back in May.

Funny because my original idea was to build a browser extension to modify existing pages with "click on any element and prompt" type of workflow. That approach also has certain limitations/gotchas too. eg if you click on a list element and ask it to push it down the list, it won't work. It'll fail as expected but it may be counterintuitive for someone playing with it for the first time. Having said that, I still believe that there's value in this UX and will most likely implement it soon.

There's one huge not-immediately-obvious element there: Vercel bandwidth is expensive. My company switched our CMS content over to Prismic in part because serving all media content directly from them was significantly cheaper than doing so through Vercel, even at a fairly low volume.

Vercel/Guillermo Rauch are really onto something

It's established that cloud is paying a premium for not having to hire: but they've gone a step further and started testing the limits of how much you can overcharge on top of cloud pricing for good DX

Clerk.dev (another investment of his often pitched along side Vercel) follows a similar model, and almost without fail if I see a Vercel sample pushing a SaaS, Rauch is usually listed as an investor

> Clerk.dev (another investment of his often pitched along side Vercel) follows a similar model

Not when there’s auth0 / Okta charging 5x more. Completely different.

Clerk.dev charges like Okta and then gets pushed to people who should be using something like Cognito which costs ~1/100th as much.

If you're not building some $100/per/month/seat enterprise tool there's no way you should be reaching for something like Clerk.dev, but obviously that'd be a constrained market.

So now it's pushed as a general bandaid for Next.js having a poor auth story, complete with the "official authentication partner" badge (but of course, no mention of the investment)

I mean Guillermo invests on a lot of Products, so do so other vercel team members, many of those end up being FOSS or Vercel friendly, hardly anything wrong there.

Pumping your own bags is legal

You're right. I've noticed this pattern, too.

I wonder how it'll actually work out for them though, their revenues don't seem to justify their valuation, especially since the Zero Interest Rate Phenomenon is now over. For example, in 2022 they only hit $25 million on a $2.5 billion valuation [0].

[0] https://getlatka.com/companies/vercel

Hard disagree, this ain't gonna kill CMSs, what are you talking about?! Content management goes beyond just a website, especially in the headless cms world. Also those products can be as low as free.

Been involved with a lot of Contentful projects large and small that only existed to publish content to web properties. I guess it kills that use case.

Depends imo. If it's purely just a few static pages then yeah I can see this happening but as soon as you have anything like a blog you're going to want some sort of content management system surely to manage it?

What I see this more doing is eliminating the need for someone to code the glue between the CMS and the website -- you can just ask the AI to add code to get your blogs from Contentful

100%!! You need one model to understand the initial prompt and figure out a development plan. Once the plan is created, the subtasks can be delegated to specialized smaller models (I think OSS will play a big role in this!).

have you explored using something more lightweight than react ? for example, wouldn't it be easier to generate vanilla html with tailwindcss ?

Reply on news.ycombinator.com