This squarely falls into the "everyone probably knows this but it didn't click with me right away" category so please feel free to laugh at my ignorance, but it's something I realized over the past few months, and as I just used this technique this morning, I figured I'd share it on the blog. The idea is simple - it's trivial to ask a Gen AI tool to do something for you - and depending on the ask, may work great. But what I realized a few months back, especially in regards to having AI parse data, is that you can also use the opportunity to generate a tool (like a Python or Node script) so you don't need to return to the AI tool again. This becomes especially useful if you want to slightly tweak the output over time or gradually add more features.
Greetings and salutations, readers. It's been a few weeks since I shared one of these, mostly due to the job search being somewhat exhausting, but I've got a backup of links so it's time to get back in the habit. And of course, it's Father's Day and I want to wish all the dads out there (myself included) a very happy father's day. This weekend I got to officiate my first wedding (for my brother-in-law and his fiance) so my plan today is to do... nothing. Enjoy your links!
Many years ago, I made the switch from building primarily app-server backed sites (using Node, ColdFusion, PHP, etc) to fully static sites using tools like Jekyll, Hugo, and Eleventy. For the most part, it was a great shift in how I build, but there were a few things I had to figure out in that new world - one of them was simple form handling. While I could have used serverless just fine, it felt like overkill. Luckily, there were a few services out there that catered to this need. You would simply use a unique action for your form and that service would handle collecting the form data, emailing it to you, and redirecting the user back to the site.
Ok, chalk this up to something I may never actually use in production, but I was curious how well the browser would handle changing the favicon of a tab on the fly, and combining that with Cloudinary to dynamically modify the source. The inspiration for this was something simple - Google Calendar's favicon is unique per day, so for example, right now I see this:
It's hard to believe I first experimented with Diffbot nearly five years ago. You can see that first post up on the Adobe Medium account - Natural Language Processing, Adobe PDF Extract, and Deep PDF Intelligence. Since then I've tested out various APIs and features from them and was lucky enough to connect with them recently about a new initiative, a web search API.
No! Thanks for reading. Still here? Cool. So way back in 2023, I built a little demo that I thought was kinda cool. Early in 2023 I had used Alpine.js and the excellent Deck API to build a simple web-based Blackjack game: Creating a Blackjack Game with Alpine.js and the Deck of Cards API. I then took that demo and connected to Google's GenAI API, which back then was named Palm: Can GenAI help you win in Vegas?. It's been a few years and I thought I'd take a stab at it again, this time making use of Chrome's Prompt API. Technically I'd have much better luck using a bleeding edge Gemini model via a serverless function, and I may still take a stab at that later, but I thought I'd see how well the Prompt API worked.
Want to read everything? Head over to my giant list of every single freaking post ever!
