[jurgenaut23]

What used to be a life time project that would inspire awe and respect and make OP an instant hire for most managers is now a fun 1-week stunt that makes you go “cool, how many tokens?” Of course, the result is cool and maybe even useful (I wouldn’t dare say _correct_, being ignorant of the topic), but I cannot help but think that this would have been tremendously better if done the old (proper?) way.

Also, I suspect that OP would have learned so much more on the topic.

[stavros]

Yeah but the choice isn't "do I spend two weeks on this, or do I spend a lifetime?". It's "I have two weeks, do I spend them making the whole thing with AI, or 5% of it without?".

[codingminds]

If it's inaccurate it's worthless anyhow.

It's absolutely no issue to build all the technical and design stuff with AI, but science facts must be science facts and not just AI gibberish presented like facts.

What's the benefit of presenting not trustworthy data?

[stavros]

That's an entirely different argument, though. Let's not move the goalposts.

[jurgenaut23]

It must be me getting old, but definitely 5% of it without, if your goal is to learn something. 5% of thinking about a hard problem is more valuable than 100% of letting it being done by someone else.

Especially when the value of the end result is close to zero, since it is now so easy to do and replicate.

[stavros]

Eh, my goal is to make something, personally.

[tgv]

I'm pretty much on the no-AI side (learning, art, decision making, etc.), but this is the kind of thing I can appreciate. I suppose OP didn't want to learn more about coding this kind of visualization, but rather learn from the visualization. Any tool that can help with that is acceptable, whether it produces code or not. That the tool produced code has the advantage that OP can share this with us. I only hope it doesn't contain fundamental errors, because that would make this project a negative contribution.

[jurgenaut23]

Sure, I don’t discard the contribution altogether, but I am dubious that it is possible to properly draw the line between what to vibe code and what to do it “by hand” to make sure you get the benefits of building.

[noduerme]

Errors are just nonsense that shows up in the console until you spend more tokens to make them go bye bye, right? I think you're talking more about the idea of true and false information, which is such a human bias. Will it really matter to anyone in 5 years whether this accurately depicts phages? By then AI will have solved everything. /s

[eigenvalue]

Look at the commit history. I’ve been working on this essentially every single day for over a month.