Post
Watching people try to "debug" coding agent behavior is so unbelievably painful. It's just fucking magic incantations
Did you use this specific super power?
Did you ask it to explain itself?
Did you read it's thought process to see where it's getting overwhelmed?
No because none of that is real. It does not do any of those things. It confabulates statistically probable text, and it does that over and over until a control loop stops asking for more. That is the whole entire thing.
@jenniferplusplus "Any sufficiently advanced garbage is indistinguishable from magic" (Sterling's corollary to Clarke's 3:rd law)
@jenniferplusplus Have you actually tried current-gen coding agents like https://antigravity.google?
What you wrote sounds like you should have tried them. But what you wrote also sounds like you don't know what you're talking about.
@jenniferplusplus
Magic >>>>(…)>> AI
(Even if the earlier doesn't exist.)
@jenniferplusplus Society is building towards these guys https://wh40k.lexicanum.com/wiki/Tech-priest
But I wrote "DO NOT HALLUCINATE" at the end of the prompt
@jenniferplusplus you cannot control the stochastic parrot's random number generator
Well, no, but you can nudge it into a different place in the latent space which might be closer to your desired output. The problem with wanting to do this is that it requires you to have a mental model of the shape of that space, which is a high-dimensionality space with things located nearby based on a similarity metric inferred from the training set. And even if you did somehow create a mental model of how this works, that mental model won’t be portable across different ML models (and often not across different training runs of the same model). You can build a bunch of heuristics (‘if I put X in the prompt, it increases the probability that the output will look like Y’) but these are not going to be consistent and reproducible.
But that’s fine, because now you’re being charged per token, wrong answers that look close to the desired output are a feature, because they encourage you to try again.
@jenniferplusplus yes, exactly this, and exactly why it's burning down the forests. just ... a supercharged Monte Carlo routine
@jenniferplusplus As I'm fond of saying, you did not 'wear down' the AI until it gave you the right answer. It wore you down, because it's inexhaustible, until you were finally willing to accept its answer.
@jenniferplusplus "I guess this art looks kinda like what I asked for. Close enough, anyway."
Install bonfire.cafe
Get the full app experience
