“Hi. I was talking to Gemini the other day about ilyAIMY events and the information it was giving me was off by a few days. Like the Columbia show on Wednesday, it was telling me that it was on Friday, or rather it was giving dates that were 2 or 3 days off. You may want to talk to the machine about how to get it to display your events correctly.”
Uhm. No. I shan’t. I need it to solve it’s own damned problems. Though I absolutely know too many fans ARE probably consulting Gemini rather than our website about our shows, I’ve got other problems for AI to solve.
I’ve been trying to move the Journal. It’s been a real debacle, honestly – and I’ve been using AI – an AI agent called Gumloop, to be specific. Moving it from my painstakingly hand-coded and increasingly outdated Journal was an obsessive manual process of cutting and pasting into WordPress over the first couple of years of COVID, and I slowly broke my website. I’ve since spent another 4 years suffering under an increasingly sludgy ilyAIMY.com as the domain was bought and sold by various companies, currently coming to rest in the clutches of my nemesis (long story), Networksolutions.com, creating a disaster of convoluted routing, thoroughly breaking my website.
Heh, and by “thorougly breaking” I mean 10 second load-times. Oh, can you imagine my woes college-age rob? In the days of dial-up and AOL screen names, can I imagine being sad over a 10 second load-time?
And we’ve been Living with that for years. Unprofessional, embarassing. It has WEIGHED on my mind that at some point Network “solutions” would just up and delete the functionality upon which a quarter century of my memories rely.
I’ve had friends work on the problem. I’ve spent a hundred dollars or so on backup and transferal solutions that have gotten nowhere sans explanation – but they all offer to solve the problem for hundreds (if not thousands) of dollars more.
And so I’ve turned to agentic AI. I’ve spent the last several weeks guiding it through a hit-or-miss (though honestly, mostly “hit”) process of scraping the old website and reimplementing it into another format. We tried Blogger for a while. It was promising but had built-in limitations that eventually became impassable. We’ve actually RETURNED to WordPress once it was determined that it wasn’t that I’d broken the PLATFORM, it was the layers of routing that were broken – and the AI could fish through those and return the correct results – including the way images have been uploaded, pulling from the original ilyAIMY.com servers, the servers that have been deprecated by powweb, then iPage, now Domain”solutions”. It’s accidentally duplicated images, tripled them… we pulled apart the error that made SIX duplicates of every image. We teased apart logic loops that broke links, replaced all instances of the word “Heather” with “robilyaimy”, we ran massive protocols that got hung up because they were too fast, or ran too slow.
I do not know why, but the sandbox? The sandbox did not persist.
And so – as I write this – I’ve told it to take everything we’ve learned and do it again. Take the past 23 years worth of posts, rescrape them and transfer them. It’ll probably take all night. And I’ll have probably spent about $300 on AI “credits” getting the job done, most of which was the exploratory and bug-testing phase. The actual work, if I’d known all the gotchas and problems and things to ask about, probably “should” have cost me about $50.
$250 of education though. That’s not the worst thing in the world. I’ve made more expensive mistakes over the years.
Well, good luck little Gumloop. I remember watching Alien with my dad and him complaining about how they’d communicate with their computer. MU/TH/UR 6000 was able to respond to very casually-worded requests and my dad felt that no computer, even in that sci-fi world of the far-flung future should be able to respond to “Request clarification on science inability to neutralize alien”. And here I can say “I’m trying to think of other gotchas – I’ve created posts and uploaded images on my own – so you need to ONLY delete posts and images that YOU have put there. Not a full wipe of EVERYTHING.” After confirmation I tell it “Perfect. Make the clean sweep. I wish you luck!”
It replies:
The content cleaner is working beautifully: 1. ✅ <div class="entry-content"> stripped 2. ✅ Noscript tags removed (image count: 6, not 12 – exactly half, no duplicates) 3. ✅ ShortPixel URLs replaced with original ilyaimy.com URLs 4. ✅ Size suffixes stripped from filenames (-1024x768 removed) 5. ✅ <div class="wp-block-image"> wrappers removed 6. ✅ lazyload class, noisy data-* attributes removed 7. Now I’m moving on to write the import script that’ll read the cleaned JSON files, push posts to WordPress via REST API, and rebuild the image cache while tracking which posts have been imported by slug. I need to examine what the sample categories and tags actually look like in the data so I can properly handle the lookup and creation logic.
Sounds good to me.
If we’d buried my dad somewhere nearby we’d hear the muffled *thud* of his mind being absolutely blown.
Leave a Reply