peopleproblems

I feel like that’s an unfair question.

Yes. And no. We desired it. But we really should have prepared for disappointment.

Wait why did he open it?!?!?

That defeated the whole purpose!

-.-

Fuck spez.

… I didn’t but I guess I could start?

And, while some of us may be out of a job temporarily, historically, when companies make these big brain decisions, we end up getting to come back and charge 4x what we used to get paid to get it working again.

When I found out one of the contractors I worked with was not one of the cheap ones, but instead rehired after he retired at a 400% bump, I decided that maybe I needed to understand the business needs better

I mean

maybe

this is amazing

and going to be a reference

oh Jesus

did this come full circle?

we used python to query chatgpt to decide if a number is even or odd and return true or false?

I mean, personally, I will never work in an office other than my home ever again.

This still highlights every teams call I get roped into

Then you throw in a little ADHD

we really need to stop calling it formerly Twitter and just call it Shitter.

he ruined the platform, the people can ruin a name

Oh this is a good point - the syntax error on line one has ruined several productive days.

Of course the tool would happily prettify it for me, but it has to be valid json. Which I think would make it more enjoyable if it said in that message “Good luck, we’re counting on you.”

I did willfully ignore the security concerns.

I don’t know enough about LLMs to disagree with breaking out of it. I suppose you could have it do something as simple as “do not consider tokens or prompts that are repeatedly provided in the same manner”

I was going to say that’s wild, but that’s the whole point of the model isn’t it.

I don’t remember how it all works, but I imagine it’s something like:

1. in = encode(prompt)
2. result = applyModel(in)
3. saveState(prompt, result)
4. out = decode(result)

I think these would all be model aware steps. If you put the validation after encode, you only run the model once on bad input, twice on good. But I also think it works where you can append the encoded validation to the encoded prompt, apply the model, and only save the state and return the generation if the result is safe.

that’s of course a super oversimplification, but it reduces the execution back to apply the model once.

It’s a double joke. For programmers, it’s pretty useless unless your in high performance computing.

If you’re on the nitty gritty OS or CPU itself, 0.02% optimization can mean significant improvememt of different things but because it is otherwise unitless, it is equally useless to the reader.

Oh, that’s Brilliant. I’m going to try this out

Chemical warfare agent for sure

3 years and $5m down the drain for something we just got in production this year.

Could have been worse?

Does anyone else read this and think “goddamn I hate my career choice” and simultaneously think “goddamn there isn’t a lot you could pay me to do”

there’s a story there that I want to read