A few months ago, a client asked me to shave about €200 a month off our agreement.

[averlice@fedi.blindsoft.net]

hahahahahahahaha tipical.
ouch!

that's just stupid. sometimes a human sysadmin is better than chatGPT. +, there are certain things chatGPT (and even claude) just cannot fix in minutes that a human sysadmin can in about 5minutes.

[stux@mstdn.social]

@stefano Holy shit..

That required some digital balls right there

I do not have the patience you have sir, my respect!

[rozenglass@fedi.dreamscape.link]

@stefano@bsd.cafe Slowly, more and more of my critical response calls end up tracing their root cause back to some random LLM slop changes that someone working at the client thought they can just shove in important places without them knowing what they're doing. I'm also having to condition my support contracts on not having the client use LLMs in the parts they need my help with, because I really have better things to do with my life than try to understand what a statistical word generator threw up at random hours of the night. I was starting to doubt the market utility of my skills as the LLM hype was growing, but as the it keeps going I find that the fear mongering does not reflect reality at all on the ground, in my consulting experience so far. The popular sentiment is starting to feel more like fanatical faith than a technological reality. Perhaps this the effect of a massive number of people going all-in on a fundamentally shaky bet, with their finances and their reputations, so the more the tech shows signs of falling short of the expectation, the louder those invested in the movement preach of doom and missing out and falling behind and various other woes. As if we all believed hard enough, and prayed hard enough, the machine gods will finally bring forth the new age, in which their risky bets makes them the magical winners. Humans are quite interesting.

[stefano@mastodon.bsd.cafe]

@oxy in bash? I'm (A)in

[stefano@mastodon.bsd.cafe]

@gnaddrig @GrundeinkommenAberRichtig exactly. This is not an anti AI post. This has always happened: people think they can save money using [$TECH] and getting rid of experienced people. Then something goes wrong. Then they try to contact you to clean up the mess they did.

If you're doing it only for money, you can get a good profit. If you (also) do it for passion, it can be disappointing.

[fedihacker@masto.es]

@rozenglass @stefano

#LLM #slop #ChatGPT #claude #AI #consulting

[stefano@mastodon.bsd.cafe]

@rooktallon @FreakyFwoof it depends - not a "x per hour" rate. We have an agreement for server management, backups, monitoring, etc.

[stefano@mastodon.bsd.cafe]

@stux thanks. Yes, I'm quite patient, but sometimes my patience is not enough

[nick@treffen.geekyschmidt.com]

@stefano I think the price goes up, no?

[justine@snac.smithies.me.uk]

An unbelievable thing for folk to do and certainly would feel like a slap ( slop ) in the face.

[jeff@bluenoser.me]

@stefano I learned, way before LLM and spicy auto-complete were "things", to never shave my price without a shaving in scope too.

When I do fee estimates they are very detailed, most in my industry don't share that, but I do.

So if they ask me to "find" $500, I will ask where they want to cut back.

Most times they don't, and hire me anyway. When they don't, its not personal, they just have different priorities.

[timwardcam@c.im]

@jaz @stefano Yes, exactly that.

One should now have *three* tiers of rates:

Cheapest: I built the whole system from scratch.

Expensive: I "just finish off" this system that the under-bidder has "nearly completed".

Enough to retire on: I have a look at ANY system which has had ANY AI input (but without any promise that I, or indeed anyone else, can EVER fix it).

[crabbyit@infosec.exchange]

@stefano Not surprised. Every throw money at AI instead of knowledgeable people decision has backfired.

[crazyeddie@mastodon.social]

@stefano I think maybe you should take this as a learning experience. When someone asks you to reduce your pay to increase their own profit, but put it in mealy mouthed bullshit framing to manipulate your empathy, you'll recognize it for what it is and nope the fuck out BEFORE getting screwed.

Silver lining.

I probably would have ate that bullshit too. Thanks for sharing your lesson.

[stefanhirsch@mastodon.social]

@stefano

that's pretty cheeky

[jztusk@mastodon.social]

@jan @stefano @mat3

I mean, knowing that someone is using ChatGPT to go in and monkey with their infrastructure makes it much more likely that they're going to screw it up in obscure, senseless, and harder to fix ways.

That's a good argument for charging then a higher rate. "Working along side GenAI" is a higher-tier service - I'll provide it, but it costs.

[3dcandy@mastodon.social]

@stefano I have ditched a ton of clients for exactly these reasons this year

[mr_gharice@mastodon.social]

@stefano
AI doesn't have any understanding. Depending on the context, it pieces things together in the way most likely to provide a solution to the problem. And normally you can work with it. But there is no understanding of the problem. Also many critics don't get this.
Don't get me wrong—I use it, and it makes my work easier in many ways. But there are still those moments where the result is total BS and where you need to understand it to know that it is BS.
@zleap

[educoder@mastodon.online]

@stefano @jan @mat3 that's a good idea, because it's likely you'll end up arguing with them about some nonsense the token generator spewed out which they then make you responsible for. Get out of there before they try to pin the responsibility for their own foolishness on you.

[michaels@mas.to]

@stefano unfortunately, all clients are using some sort of AI system - and it's worse when they use the free plan without logging in.