The Allure of Immediate Feedback
Tao Bojlén wrote a very interesting post, comparing AI-assisted coding with the short-term pleasure one gets from slot machines.
Tao Bojlén wrote a very interesting post, comparing AI-assisted coding with the short-term pleasure one gets from slot machines.
“The reason that tech generally — and coders in particular — see L.L.M.s differently than everyone else is that in the creative disciplines, L.L.M.s take away the most soulful human parts of the work and leave the drudgery to you,” Dash says. “And in coding, L.L.M.s take away the drudgery and leave the human, soulful parts to you.”
So far, this has been more experimentation on my end. I’ve got a personal project in mind, and I’m doing most of the work via Claude Code (or rather, Claude Code is doing most of the work). I’m mostly just providing guidance and follow-up steps.
I think the comparison folks have made has been to think of Claude Code as a very competent Junior Developer, who sometimes goes wildly off the rails. And needs guidance back to sanity.
youraislopbores.me is a fun online project by mihir maroju, that allows you to ask questions of humans who are pretending to be AI. It also allows you to answer questions from other humans, while you pretend to be AI.
You can submit a prompt that asks for an answer (text) or image (drawing). Each request costs 1 credit. In order to earn credits, you can switch to the other tab (the AI tab) and answer questions from other people.
“There are many arguments against vibe coding through A.I. It is an ecological disaster, with data centers consuming billions of gallons of water for cooling each year; it can generate bad, insecure code; it creates cookie-cutter apps instead of real, thoughtful solutions; the real value is in people, not software. All of these are true and valid. But I’ve been around too long. The web wasn’t “real” software until it was. Blogging wasn’t publishing. Big, serious companies weren’t going to migrate to the cloud, and then one day they did.”
In addition to exploring Vercel’s AI SDK, I started to actively use Claude Code.
Many said the trouble had started when their loved ones engaged a chatbot in discussions about mysticism, conspiracy theories or other fringe topics; because systems like ChatGPT are designed to encourage and riff on what users say, they seem to have gotten sucked into dizzying rabbit holes in which the AI acts as an always-on cheerleader and brainstorming partner for increasingly bizarre delusions.
Gandalf is a fun game that challenges you to write prompts to elicit a password from an AI bot. With each level, the task becomes more difficult… requiring newer and less direct approaches.
The level of detail is pretty incredible, even if there are issues here and there in the background. Difficult to wrap my head around the fact that all of the videos here don’t actually exist, and were generated by text commands interpreted by a machine.
I’m a bit excited and terrified, to see where this will go.
After getting a pretty gruesome example (left), I asked DALL-E to make something a bit more family friendly. And it produced the image on the right. I’m honestly not sure which image is more terrifying.
It’s been fun getting back into this project, after setting it aside for so long. It was more of a proof of concept project, but lately it’s been enjoyable to try to improve it a bit.
While I’ve had so-so success with text prompts adding things to images, I’ve found it pretty amazing when I just leave the prompt blank. And what Firefly does is more or less try to fill in the blanks (fun thread on expanding album covers).
Found out today that I’ve actually got access to the Browsing and Plugin features, for ChatGPT. Tried out a few things, and ended up asking it about the type of content on my site/blog. The answer was… fine. Not great, but a decent summary of the most recent posts. But the thing I’ve been most…
It’s hard to explain, without giving it away. So I’ll just share another screenshot and say that I’m having a blast, trying to make this weird idea happen. Which is always a good sign, whenever I have a weird idea I’m trying to code out.
I have what I think is a great idea. I have what I think is a terrible idea. We will see.