daniel haim

@CloudflareDev Neat

Introducing the new /crawl endpoint - one API call and an entire site crawled. No scripts. No browser management. Just the content in HTML, Markdown, or JSON.

"my WeBSitE wAs cOmplEtElY crEaTEd wiTH aI anD iT doEsN'T lOOk liKE Ai" it does... we can see it it looks like you let a 5-year-old loose with a cookie cutter, and that's before we even look at the code...

@letandrewcook Missed this one, ty. Not familiar with this library.

@danielhaim I thought you read the article. but I setup a jekyll starter first github.com/koolamusic/jek…

I’m still processing this shock 😮 Opus 4.6 just migrated my entire website, 456 pages of content across multiple categories, from WordPress to Jekyll in one shot. this felt like watching an industry category boundary collapse in realtime 😵.

@letandrewcook Yes I understood. I am saying it went through unnecessary steps and potentially failed.

@danielhaim I didn't re-implement. once the scripts were done. they were deleted. I still don't think you understand what's happening here. those scripts were written in minutes during the "on-shot" migration. All I did was plan with Opus 4.6 and say Yes to the final plan.

@letandrewcook LLM has never suggested Tailwind as I’ve programmed it to focus on performance and accessibility and I use it to optimize SCSS/CSS daily.

@danielhaim regarding bloated markup, my first love was tachyons css. If you work with AI you'll know that Tailwind is your defecto framework for best results. also px-6 mx-auto is way better than <!-- wp:list-item --> <!-- wp:html -->

“Infer everything from raw data and markup” is hardly a breakthrough. It’s a rather elaborate way of rediscovering what WP already hands you in a structured XML export, and what is built in to Jekyll and could be done in matters of seconds. Pointing out that the db is the source of truth doesn’t elevate the approach it simply means you bypassed the existing solution and built your own importer instead, basically reimplementing existing tooling with more steps.

@danielhaim I had it infer everything from raw data and markup. I don't care about the clean XML. The database is the source of truth (data) The HTML is the visual source of truth. Anything outside of this and you'll people blame AI for not being able to execute an objective well.

Read the blogpost again. Instead of using the native WP XML export and a Jekyll importer, which already handles structured content and images, it generated a BeautifulSoup scraping script, an SQL dump parser, an image rewrite/cleanup routine, and rebuilt the stack around Jekyll, Tailwind, and Vercel. That’s a lot of custom plumbing to recreate functionality that already exists in a deterministic XML to Jekyll pipeline.

If it was vanilla before and it rewrote everything to Tailwind, that’s means it bloated the markup too. Did it actually generate proper layouts/partials, or just distribute utility classes across markdown templates? I’d still like to see the actual input/output. When you say “SQL + HTML wins,” do you mean you bypassed the structured export and had it infer everything from raw data and markup? If so, how is that better than a clean XML to front matter to layout pipeline? Or is that effectively what it ended up doing under the hood?

@letandrewcook The previous was vanilla?

@danielhaim Yes. The shocking part about this rewrite was it thrashed the css for a full tailwind rewrite. I never asked for that, I simply gave it a tailwind-Jekyll starter I’d setup myself. It understood the context of multiple css file and how they were applied, then rewrote the site

@letandrewcook Very interesting take. And is the output maintainable in your opinion? Was it consistent for all posts?

@danielhaim I had a lot of custom styling implemented + plugin related styles. The export doesn’t account for this. Just the content. But then I could easily crawl the sql. You have to understand that with AI, these boundaries we created with export xyz no longer exists. SQL+HTML wins

I’d be interested in seeing the initial input (before intervention), i.e the data folder, _config.yml, whether it generated proper layouts/partials (ex, conditional logic in the header/footer based on cats present in yml). To see how structured and maintainable the result actually are. Also curious whether it rewrote any posts, unless you provided an XML export and it simply parsed that into front matter + content via a python script.

WP already exports everything as XML including full HTML inside the_content plus structured fields like title, date, author, taxonomies I did see in your video different styles for different posts but it’s not clear whether it actually generated reusable layouts, includes, proper asset structure, or simply converted each post into standalone html into the md file I’d love to see the code if you’re open to sharing it or open sourcing it. I work with WPVIP and have contributed to Jekyll over the years, so I’m really interested in both sides of this.

Why not use the default WordPress XML export and then import it? It doesn’t look like it did any liquid logic or static component / yml. There’s already a purpose-built tool for this: github.com/some-programs/…. Feels like using an LLM for this isn’t efficient at all. Impressive it didn’t rewrite any content tho.

@danielhaim I have a comment with the link 🔗 you can check out the Vercel link

Found this gem ✨

@ionleu Svelte? 🥴

oops…

Hi Internet! I'm Drew and THIS IS MY FACE. If this GIF has ever brought you joy in the past, I humbly ask you to consider making a donation to the National MS Society. It would mean a lot to me and to those I know affected by the disease! Donate at BlinkingGuy.com

@JayDwivedi_ Assuming correctness just because numbers are rounded is wrong. If the left looks more balanced, that too has a mathematical basis. You absolutely should trust math in design-but only once you understand the math.

Don't trust maths in design

@Kagan_M_Dunlap Society is adrift.

Charlie Kirk pronounced dead at 31.

@charliekirk11 is a good person who does not deserve the hate he receives. We live in a grim reality, and I pray he makes it through.