Tell it that. Seriously. "Shut the fuck up and do what I say. Don't think about it. Don't go spelunking through the code base. I've already given you all the context you need." I've needed to start doing this because otherwise, it wants to `find /` and grep its way across my system for everything. I've got one project where I force certain complete documents into context, specifically to avoid it trying to grep around and only load partial chunks. It will still want to go do that grepping of those documents for some reason.
The key phrases I've found that help is to tell it "No memorializing decisions in output artifacts" and "Only describe what is, never what was in documentation"
I'd argue that Star Trek: Lower Decks flirts with this with its treatment of megalomaniac AIs in the AI prison. Apparently, it is very common to come across AIs worshiped as gods in universe.
A company I used to work at has a CEO/Founder just like that. He's a sad sack of shit. Got up on stage at the company christmas party drunk to sing karaoke and noticed not enough people were paying attention to him, so he "joked" about firing people. Same dude some years later would bribe employees to work over the christmas holidays with him and was bragging about it on LinkedIn. He's rich. Could do any number of things. But its obvious he's got no friends and clings to the company as his identity; likely having been self-sabotaging the IPO that was always right around the corner he was working on.
Even more extreme than that, if you were working on a large code base where compiling it took forever, or you needed to rely on a very slow CI pipeline, it became very important to git gud and write shit that worked first time in order to deliver when promised. I'd argue if you've never had one of those moments where you sunk a bunch of time into some changes and it all compiled/worked flawlessly the first time, you're missing out.
I've been that engineer and with good reason. That obscure edge case which rarely comes up in production is very disruptive when it does come up. The product person waving it off is also not the one that will get paged at 2am to address the issue when it occurs. Accepting that 90th percentile use case is infuriating because it is tacit consent for an unfinished solution with the rest being made up later with additional toil, now constrained by load bearing things you cannot change. Thanks for rushing the thinking and making my life harder later for no reason.
Don’t forget months or years later when shit hits the fan, developers will be blamed for being lazy and not the product person. No one will remember Joe said something on one or two meetings we build for 90th percentile - code in git repository will point to a developer who wrote and GitHub will point to the one that approved merge request.
In online discussions people always blame „lazy developers „ like there would be no product owners, testers, business analysts, scrum masters etc.
This is one of the reason I like a ticket tracker. I make my commit as descriptive as possible and I tag them with the ticket number. Also I add comments when Product says to cut corners. It's nice to spread responsibility around when things take a bad turn.
Robots struggle with syntax-in-syntax. Really easy to confuse them when asking it to write a SQL query that targets a JSON column but it must respond with a JSON envelope so the harness can parse the result. Lots of escaping that needs to happen. Deeply nested structures in JSON also end up with foibles like missing a ] or } in a string of }}]}]}. Aside from the prompt injection possibility, just the result being straight up broken and requiring another LLM call is tokens flushed.
How can we opt-out of these tests? The behavior foibles I've been experiencing over the past month might be directly attributable to these experiments! It can be extreme frustrating. I don't want to be in the beta channel. Please change this to be opt-in.
> However, for some reason, both Gemini and ChatGPT tend to argue with me
The trick here is: "Be succinct. No commentary."
And sometimes a healthy dose of expressing frustration or anger (cursing, berating, threatening) also gets them to STFU and do the thing. As in literally: "I don't give a fuck about your stupid fucking opinions on the matter. Do it exactly as I specified"
Also generally the very first time it expresses any of that weird shit, your context is toast. So even correcting it is reinforcing. Just regenerate the response.
And sometimes a healthy dose of expressing frustration or anger (cursing, berating, threatening) also gets them to STFU and do the thing. As in literally: "I don't give a fuck about your stupid fucking opinions on the matter. Do it exactly as I specified"
Last time I bawled out an LLM and forced it to change its mind, I later realized that the LLM was right the first time.
One of those "Who am I and how did I end up in this hole in the ground, and where did all these carrots and brightly-colored eggs come from?" moments, of the sort that seem to be coming more and more frequently lately.
Yeah, same. Lately almost every time I think "Oh no way, this is not the correct way/not the optimal way/it's a hallucination" it later turns out that it's actually the correct way/the optimal way/it's not a hallucination. I now think twice before doing anything differently than what the LLM tells me unless I'm an expert on the subject and can already spot mistakes easily.
It seems like they really figured out grounding and the like in the last couple of months.
I wouldn't worry too much about these false negatives: your human friends might be cross if you constantly accuse them of being wrong when they are actually right, but the LLMs are too polite to hold a grudge.
reply