I just had to go check out your site with that comment. Nicely done! Was waiting for the physics simulation to either fully settle or explode, but really neither of those happened.
I've done this. Ask Claude to permutate a prompt and run Claude Code or a subagent to observe effects until success. Claude will iterate on the prompt faster than I could and likely catch more edge cases, too. It's one of those tasks where the end goal is very clear and the agent just needs to iterate on permutations. Arguably this is a perfect match, too, because prompt engineering is really just text generation, so why not have a text generation machine do it.
Similar thoughts here. My gut tells me "if I wanted AI responses I would just ask AI", but in truth LLMs are not like Google Search. LLMs are more like a slot machine where you put your prompt in and out comes a combination of insights that are personalized to the inputs.
"I don't know" isn't in the training data. Nobody writes engineering books, science papers and blog posts that end with "well, I don't really know, the end".
So it's better than 3.6 Flash, at half the price. I've been pretty excited about Gemini models recently, they just feel so fast after spending most of the day at work waiting for Opus 5.
Yes, 3.6 Flash is very fast. I used to get a fair amount of usage of the Gemini Flash models on the free tier. I signed up for their $4.99/month tier (includes 400GB of Google space which was also enticing) and it turns out I only get about 15 to 20 minutes of usage before I get a come-back-in-7-days message. Comically low usage limits on that plan.
For comparison tho, 400GB of cloud storage for $5/mo is actually quite amazing even if it didn’t include anything else.
I know, it’s cloud storage, a NAS lets you own your data, but for non techies who have a bunch of photos and videos; it seems like an easy recommendation.
You can get Google AI Pro through verizon fios or mobile in the US and pay only 10/mo. For 5TB cloud storage, and decent gemini usage its a compelling option
I'm super curious how you make the development process interesting for him. I get that a 5yo would be good at generating ideas but how do you then go "okay, now I'll need a week in front of a screen to build all these ideas" and not have him lose interest. Do you accelerate using LLMs?
Yeah, good question! We can do a lot of the creative process on paper, with his styrofoam models, and audio effects etc without even being near the computer. Then when I need to go ahead and write code, it's not going to be that interesting for him, but I've shown him how it works and how we can change stuff.
A lot of the very technical parts are done during the evenings after he's asleep, and I think that's okay, as it'll end up in a better result the next day so he can go and test and give further feedback.
Oh yes, I eventually did less hand coding and relied more on Claude Code to get the job done. I understand the game loop, the architecture and the algorithms, which I think is important. Don't hide that under the hood.
I also attempted to make a game with my eldest daughter initially. Unsurprisingly, she didn't want to sit there watching me code or drag things around in Godot's level editor.
Admittedly, I really need to make a new blog post showing the latest improvements brought about from the in school pilot I'm running weekly at the moment.
> web app I had launched using Claude was strikingly similar to his open source project, even reproducing a bug he had later fixed,
How? How do I prompt to get a literal copy of another app? When Claude/Gemini builds anything for me I have a thousand ideas for functionality improvements and UI/UX changes. How do you just stop at v1 of what Claude gives you?
