are there people out there who just want to refactor every day?
just wake up and find the worst code and just chip away at it and clean it up
wake up the next day do it again, infinitely improving things with zero external impact?
The thing is, you don’t need an LLM for this. A few years back I needed just this functionality to parse addresses from the business cards, and there’s a plenty of existing services doing this that give you, and this is important, an address that really exists in the world.
Maybe the third time's a charm 😄. Yesterday, I posted that we could eliminate things like country-name dropdowns and replace them with free text, perhaps using an LLM to correct errors. Too many people came back with ways to improve dropdowns, but missed my main point: an LLM, when used judiciously, can both save you work and improve the UX.
Instead of focusing on country-name dropdowns, how about we replace the entire name-and-address dialog with a single free-text "name and address" field and let the LLM handle everything else, including JSON conversion? No complex forms at all, so it's a win in the UX department, and there's way less work for you. There's a simple prompt and response that demonstrates how this would work in the attached image. Note that the LLM has corrected all significant misspellings in the original and added the correct ZIP code.
Of course, I would still verify the generated JSON and present the data to the customer for approval (in a format more appropriate than JSON) before accepting it, but that's just one "ok" button.
I'm actually using this as an example in a class I'm putting together, implemented as a microservice-sized agent that uses a micro-front-end architecture to create a form you can embed on any page (in an iframe, but there are other possibilities, of course).
The UI entirely eliminates the need for a complicated address-entry screen (and associated JSON-coversion code), replacing it with a single free-text field and an "okay" button. I'm using a local LLM instance, so there's no per-token cost. Not too shabby.
@ClaireMax You look great! I’ve been following you for quite some time now, and I’m in a completely awe of what you’ve achieved (and I hope I can use the inspiration to finally get my own lazy self to train)
@wutairoses@biliouris_amber Huh. I thought her chest looked bigger in the audition scene, where she’s wearing a dress, but I thought it was just my imagination. Compression bra for fighting outfit makes for much more logical in-game explanation!
I get what this is going for but it is exactly backwards.
TV in the 90s:
- buy a guide, scout out everything in advance
- plan a schedule, track channels and times
- manage space on your DVR, lose recordings, cry
- 30% ads
- often actually watch live, with ads and everything
- if you miss it, you miss it
- huh, that's on, okay, sure
- you think it's good because you grew up without DVR
TV 2025:
- watch anything anyone has ever made
- any time, on demand, it tracks your place in everything
- even if you buy everything you want, less total cost
- if they make you watch ads it's like 2 minutes per hour
- even those 2 minutes make you angry
- okay, sure, it takes 30 seconds to navigate sometimes
- you complain because of paradox of choice
Software Engineers are not paid for writing code.
They’re paid for solving problems.
The faster you accept this, the better your life and career will be.
@ClaireMax I also find that this particular screenshot of Aloy is in a bad faith. Her face looks very differently throughout the game, and I think there are some less than perfect _camera_ decisions by the devs that create this illusion. No-one looks good in a low-angle short-lens shot.