We recently had an AI workshop type of thing as our company leadership is also falling for the AI-first BS, and it really felt like an Emperor has no clothes situation to me.
Gist of it was that the VP of Eng and the CTO would sit down with the engineers and showcase AI tooling and techniques on using them etc, specifically Cursor which the VP surely has a vested interest in somehow with how much he salivates about it.
It started off fine with some nice tips and context on how LLMs work for the unaware, but the absurdity truly started when they tried to fix some minor UI bugs that were sitting in the bug board as a showcase. One of the UI guys looks at the ticket for maybe 30 seconds before saying "oh it's probably X".
The VP and CTO spent the next hour and a half trying to prompt their way into solving the bug. Failure, after failure, after failure. They tried a million different prompts, giving it full code context, limited code context, access to CLI tools. Some of the prompts they wrote were easily 1000+ words full of excruciating detail for what to try. Ultimately none of it worked, and yes they were using the latest and greatest SOTA models like Gemini 2.5 flash and Claude 3.7 and o3 and...
The final fix, from that same UI guy as before, was a literal single character change in exactly where he said it was that took him all of 2 minutes to write up, test and push, a minute of which was just him finding the project in his messy folder.
And after the absolute waste of time that hour and a half was, they, without a hint of irony, say "See how useful it was in our debugging?". I had to stifle a laugh because I felt like I was in some nightmare 1984-esque doublespeak world where the blind AI hype supercedes every ounce of real life evidence we all just witnessed before our very eyes.
Coincidentally, 7 very strong senior and staff engineers quit shortly after that, and I'm myself also looking for a change. Basically the entire company is in a bit of a "what about our actual product that every customer is complaining about?" Kind of phase, but leadership is rolling full steam ahead with tacking on AI crap on top of everything that literally nobody except for the C-level cares for.
*Addendum: not to say I don't view LLM tooling as useful, it undoubtedly is. My main issue is there's 0 nuance involved anymore, we all have to pretend as if it's universally useful literally always and for every single usecase, even when evidence points to the opposite. It can't just be "oh nice, this will save some time for these specific tasks!", no it's "you will be evaluated on how much you use AI and specifically Cursor, whether you're more productive with it or not"
We recently had an AI workshop type of thing as our company leadership is also falling for the AI-first BS, and it really felt like an Emperor has no clothes situation to me.
Gist of it was that the VP of Eng and the CTO would sit down with the engineers and showcase AI tooling and techniques on using them etc, specifically Cursor which the VP surely has a vested interest in somehow with how much he salivates about it.
It started off fine with some nice tips and context on how LLMs work for the unaware, but the absurdity truly started when they tried to fix some minor UI bugs that were sitting in the bug board as a showcase. One of the UI guys looks at the ticket for maybe 30 seconds before saying "oh it's probably X".
The VP and CTO spent the next hour and a half trying to prompt their way into solving the bug. Failure, after failure, after failure. They tried a million different prompts, giving it full code context, limited code context, access to CLI tools. Some of the prompts they wrote were easily 1000+ words full of excruciating detail for what to try. Ultimately none of it worked, and yes they were using the latest and greatest SOTA models like Gemini 2.5 flash and Claude 3.7 and o3 and...
The final fix, from that same UI guy as before, was a literal single character change in exactly where he said it was that took him all of 2 minutes to write up, test and push, a minute of which was just him finding the project in his messy folder.
And after the absolute waste of time that hour and a half was, they, without a hint of irony, say "See how useful it was in our debugging?". I had to stifle a laugh because I felt like I was in some nightmare 1984-esque doublespeak world where the blind AI hype supercedes every ounce of real life evidence we all just witnessed before our very eyes.
Coincidentally, 7 very strong senior and staff engineers quit shortly after that, and I'm myself also looking for a change. Basically the entire company is in a bit of a "what about our actual product that every customer is complaining about?" Kind of phase, but leadership is rolling full steam ahead with tacking on AI crap on top of everything that literally nobody except for the C-level cares for.
*Addendum: not to say I don't view LLM tooling as useful, it undoubtedly is. My main issue is there's 0 nuance involved anymore, we all have to pretend as if it's universally useful literally always and for every single usecase, even when evidence points to the opposite. It can't just be "oh nice, this will save some time for these specific tasks!", no it's "you will be evaluated on how much you use AI and specifically Cursor, whether you're more productive with it or not"