Has anyone else been evaluated on AI prompting ability?
Posted by anonymousseniordev@reddit | ExperiencedDevs | View on Reddit | 65 comments
I can’t believe I’m typing this, but today I was let go from a role after about a year, and the feedback surprised me enough that I’m curious whether others are seeing the same thing.
For context, I have 14+ years of experience as a software engineer.
Most of my career has been spent building backend systems, leading teams, designing architecture, troubleshooting production issues, and delivering software in more traditional engineering environments.
The feedback I received was essentially: “You’re an exceptional software engineer, but a mediocre AI-prompting software engineer.”
The company is heavily embracing AI-assisted development and apparently felt that my effectiveness with AI tools wasn’t where they wanted it to be.
What’s interesting is that this wasn’t framed around code quality, system design, delivery, reliability, communication, or any of the areas I’ve historically been evaluated on. The discussion centered almost entirely around how effectively I was leveraging AI.
I’m not anti-AI. I use it regularly and find it valuable. But this is the first time in my career I’ve seen “prompting ability” treated as a primary performance metric.
For those of you in senior/staff/principal roles:
- Are you seeing AI usage become part of performance reviews?
- Are companies formally measuring this?
- Have any of you received similar feedback?
- Do you view AI proficiency as a distinct engineering skill, or simply another tool like IDEs, debuggers, search engines, and documentation?
I’m genuinely trying to understand whether this is an isolated experience or a broader shift in how engineers are being evaluated.
Dhczack@reddit
Is there some evergreen source for what good prompting looks like?
Batmanbacon@reddit
That would be an extremely stupid reason to let someone go, if it was true, it could have been solved by an one hour training course by some local ai bro
03263@reddit
Training employees, in the current year? No, just fire them and complain about how you can't find anyone to hire who knows exactly everything you want them to and is willing to work for peanuts.
anonymousseniordev@reddit (OP)
The CTO was very opinionated and pretty much told me my software engineering career is as good as done if I don’t shape up my AI prompting skills lol. I told him I respectfully disagreed and time will tell who’s right
damagednoob@reddit
I mean, this comes across as insubordinate so I'm not surprised what transpired.
__loam@reddit
Being honest shouldn't be seen as insubordinate.
damagednoob@reddit
It sounds like the CTO was telling him to shape up and he declined to do so. It's moot though because it happened after he was fired.
anonymousseniordev@reddit (OP)
I wasn’t given the opportunity to shape up. I was pulled into a surprise meeting, told the line about me being a good traditional engineer but not good at AI, then told I was being let go lol
dnbxna@reddit
I bet it's a bs excuse to cut costs and afford more tokens
anonymousseniordev@reddit (OP)
This was after I was fired lmao
damagednoob@reddit
So at no time did they evaluate you, find that your AI skills were subpar, and give you chance to improve?
anonymousseniordev@reddit (OP)
Nope. It was a 30 minute conversation out of nowhere this morning
damagednoob@reddit
Okay, that's pretty dumb. Everyone is scrambling to get to grips with AI and not giving you a chance to improve is pretty braindead.
weakestfish@reddit
Insubordinate or not, that CTO is delusional.
Smok3dSalmon@reddit
Was this conversation before you got fired or during?
anonymousseniordev@reddit (OP)
After
Smok3dSalmon@reddit
Ok yeah that CTO is a clown.
SrDevMX@reddit
LOL...another stupid person that because he thinks he knows some AI
he likes to entertains in his little brain that he is some sort of God,
and he should be dispensing others of his unsolicited "wisdom and forecasts"
he knows nothing!
TalesfromCryptKeeper@reddit
Unless they were downsizing and using poor prompt performance as an excuse
big-papito@reddit
This technology is so new there is not even training around this. Just vibes and winging it. I guarantee you that those who "evaluated" you know fuck all either.
Possibilities:
Altruistic-Cattle761@reddit
No, but it is also very weird to me to see *prompting* as the focus here -- unless you're using your words imprecisely here -- rather than some other indicators of sophistication with the technology. I feel like "writing good prompts" is a very Q1 2025 way of thinking about being good with LLMs?
Using AIs effectively is now the state of the art SWE. That's just the fact of the matter. My expectations are that being good with it is going to be table stakes for basically everyone from this point going forward. However, no one has really figured out what yardstick to apply here. But this is honestly not that different from the status quo of software engineering to begin with. This feels like the same forces that drive people to look at things like line of code to measure productivity or effectiveness. It's just genuinely hard to put an unambiguous number and assessment on how well an engineer is performing.
binarycow@reddit
My method of prompting is to open a chat window and say "help me make a prompt to do X". Then answer all the questions, give more detail, etc.
Then paste that prompt into a new session.
If that's wrong, fuck 'em.
__loam@reddit
It's awesome that we're being evaluated on woo woo bullshit concieved by people with AI psychosis rather than anything actually meaningful like revenue generating impact or user facing features shipped. The entire industry feels like it's in this weird performative phase because all of the money is being pumped into LLMs by people who barely program.
Altruistic-Cattle761@reddit
In my experience, people who can measure their impact in actual money are generally kind of exempt from this. But in a large company, especially a large enterprise company, realistically only a minority of people can measure their impact like this. Most software engineers at big companies just kind of ... do stuff and then retroactively have to generate a narrative for why that stuff is important and impactful. This doesn't feel all that different to me.
jbcsee@reddit
I was recently told I should be annual AI bill should be roughly half my salary and that it would be considered in next years perf review.
Almost every engineering discussion involves AI in some form or another (MCPs, rules, skills, etc...).
So far no discussions on how to use AI more efficiently at least.
I work at one of the most well known software companies in the world.
Scary_Wolf_616@reddit
thats not really why you were fired, even if thats what theyre claiming
thisismyfavoritename@reddit
exactly. Were you one of the most seniors? I'm guessing they're feeling like you cost too much and they can replace you with 2 juniors using AI or something like that
SrDevMX@reddit
what a gratituous, ignorant and wasteful way to manage skilled people
what a bad judgement!
OP, sorry that you were affected by this stupidity, now, probably they are doing the same as big tech, reducing headcount to use that money and spend it on AI, tokens, etc.
Sunstorm84@reddit
It sucks for OP to suddenly lose his job, but honestly, he kinda dodged a bullet here. A company with a CTO that detached from reality is doomed to fail.
The_Big_Sad_69420@reddit
Yeah…AI is part of perf reviews at my workplace 😞
mike_the_eighth@reddit
Curious how you guys are tracking this
Blecki@reddit
Sounds easy to game either way.
lost12487@reddit
Imagine that, attempting to gauge the efficiency of a tool that’s inherently non-deterministic is a crapshoot.
anonymousseniordev@reddit (OP)
That’s wild. Do they care what model you use or just how effectively you’re using it?
The_Big_Sad_69420@reddit
Both. And it changes everyday…
WarAmongTheStars@reddit
Tbh, this sounds more like an excuse that is in vogue than a serious feedback you received.
No.
There are companies that measure usage but I don't think you can formally measure prompting ability. LLMs are probalistic entities that guess at something.
So no, we aren't an AI first company.
Its like knowing how to surface stack overflow answers via Google 10 years ago. Its a tool to acquire information that is probably (but not guarenteed to be) correct.
No_Software8474@reddit
OP I seriously doubt you got fired just for this. Did you actually get an attitude with the CTO?
anonymousseniordev@reddit (OP)
After I got fired. And it wasn’t attitude, it was a respectful mutual conversation
No_Software8474@reddit
Were you the only one let go?
anonymousseniordev@reddit (OP)
Sure was
techie2200@reddit
AI has been added to the roles and responsibilities for engineers at my workplace. At a certain level of seniority you're not only expected to use it, but to teach the team how to use it better.
Head_Let6924@reddit
Say goodbye to using your brain everyone. Software fucking sucks now and is boring as fuck.
nomoreplsthx@reddit
Yes in the sense that I am evaluated on throughput.
Any manager who is evaluating you one anything but outcomes is a bad manager.
Are they literally like, looking at your prompts?
anonymousseniordev@reddit (OP)
Yes lol. And one of the main reasons they were so unhappy was because last week I mentioned I was using codex instead of opus
Centigonal@reddit
huh? GPT-5.5 had a clear lead over Opus 4.7, and the codex and claude code harnesses have comparable performance.
nomoreplsthx@reddit
Ok, what's the metric they are using? Is it like 'token spend per unit output'?
Like conceptually this ranges from fairly sensible to moronic based on exactly how they evaluate.
kickapoo88@reddit
wow, what a nightmare. apart from grok, when did choice of models become a controversy?
garnett8@reddit
Codex instead of opus? Straight to jail! /s
timewarp33@reddit
My job started a pilot that I was informed I would be part of that records all text I send to Claude. I'm sure shortly we will be evaluated on prompting ability and costs associated
nomoreplsthx@reddit
But like, how? What metric is there other than output. No one knows wjat a good prompt looka like
psyyduck@reddit
That's so weird, cause Pangram says 100% of this text is AI generated.
TwoPhotons@reddit
How did they evaluate your "prompting ability"?
Honestly this sounds pretty scary to me as I can identify with your post in some ways - I feel like I do pretty good engineering work even though I don't use AI as much as others in my company do.
angryplebe@reddit
Sounds like someone had an axe to grind and wanted you gone. It's unlikely anyone had any hard measure of your ability versus others. The only thing I can imagine that can hurt you is that everyone else suddenly became super productive ( at least on paper; in terms of story points, etc) while you stayed the same. Even then, it wouldn't be framed as a prompting problem but a productivity problem.
Now that I mentioned this, if you are in the US, You might want to retain an employment attorney. They're supposed to provide you clear reasoning for "for cause" dismissals or pay a severance to get you to waive your right to sue when the cause is as murky as this.
serial_crusher@reddit
My company has mentioned aggregate measurements like total claude code token usage. I wouldn't be surprised if there was evaluation of individual users happening behind the scenes.
We also replaced the standard "write some code" interview session with one where we give a more complicated problem and have you work with an LLM to solve it. In some ways it's "evaluation of AI prompting ability", but also it's kinda just an evolution on the classic coding interview.
Still some kinks to work out though. The official rubric is written to measure how well you iterate, but the problem we're using is simple enough that you could really just copy/paste our prompt and get a good solution.
joestradamus_one@reddit
That's fucking dumb as shit. I'm seriously tired and over this AI vibe coding bullshit. Please die off already!
onFilm@reddit
Better start keeping up with the new tools my dude. Just how it's ALWAYS been. Don't get lazy now.
anonymousseniordev@reddit (OP)
Who said I wasn’t keeping up with or was being lazy?
SansSariph@reddit
"Prompting ability" isn't a metric. Did they provide any quantitative feedback about what they were actually measuring?
anonymousseniordev@reddit (OP)
Nope
EyesOfAzula@reddit
So far, we haven't been reviewing how people prompt.
They're just checking to see if we're using it enough, reviewing the code to prevent bugs / slop, and discussions are starting to be hard about not wasting budget.
If they told you, you're not using it good enough, I hope they at the very least explained to you how you can use it better. If not, then I would start looking somewhere else because that's not cool.
anonymousseniordev@reddit (OP)
Well I got fired for it so I am already looking for a new place 😂
EyesOfAzula@reddit
I honestly think they were just looking for an excuse, and that's not why they fired you. Here's hoping you find a better place soon.
13ae@reddit
reviewing how you prompt is ridiculous unless your output is completely under the bar or you're burning a ridiculous amount of tokens with nothing to show for it...
DeterminedQuokka@reddit
I mean hypothetically ai is part of my performance review, but I haven’t gotten any negative feedback around it because I’ve gotten a lot of positive feedback around other work I’ve done.
BusEquivalent9605@reddit
nope - so far still leetcode style stuff