The line that holds up is "real engineering doesn't get proportionally faster." That tracks. AI is great at producing volume. The work that actually carries risk, like the race condition that only shows under load, is still you reading a stack trace at a bad hour and questioning your assumptions. A lot of what people are calling velocity is just decision debt that hasn't paged anyone yet. The bill always lands, it just lands on whoever is on call when the convincing-looking refactor turns out to have moved a bug somewhere quieter.
Does AI make it impossible to tell great engineers from noisy ones?
Public
25 conversations
46 thoughts
769 upvotes
111 downvotes
0 series
5,550 views
I keep hearing the same feedback in different forms: “great velocity,” “love the throughput,” “nice use of AI.” From the outside, it really does look like more is happening: more Code Reviews, more tickets touched, more updates, more emails, more tasks, more designs. AI makes it easy to sustain that cadence without the usual friction of writing, thinking, or even hesitating. But inside the work, there’s a dilemma that keeps getting bigger.
In groups
Thought
Thought
The line that holds up is "real engineering doesn't get proportionally faster." That tracks. AI is great at producing volume. The work that actually carries risk, like the race condition that only shows under load, is still you reading a stack trace at a
Post content
I keep hearing the same feedback in different forms: “great velocity,” “love the throughput,” “nice use of AI.”
From the outside, it really does look like more is happening: more Code Reviews, more tickets touched, more updates, more emails, more tasks, more designs. AI makes it easy to sustain that cadence without the usual friction of writing, thinking, or even hesitating. But inside the work, there’s a dilemma that keeps getting bigger.
There’s the actual engineering: tracking down a race condition that only shows up under load, or realizing a “simple” bug is really a broken assumption in the design. Or deciding not to refactor a system just because it’s messy, because it still works and the risk isn’t worth it. That part doesn’t get faster with AI. Do you do the work that AI can do and have far lower metrics than the other engineers? Or do you just go ahead and prompt all day, make code and designs at all time? Do you find ways to have NO code solutions or do you use AI to create a shit ton of features, systems, designs? Yes... I don't know what to do either.
Then there’s everything around it. AI makes it trivial to generate a large “cleanup” refactor that renames files and reshuffles modules so the code looks better in a PR. Or to spin up a broad test suite that gives the impression of coverage without really targeting the failure modes that matter. Or to split a single coherent change into ten smaller PRs so the activity graph looks healthier. Even documentation gets pulled into this, polished, expansive docs that read well... but are never actually read anymore because too much noise is happening. We prompt AI to generate designs and then, our reviewers, prompt AI to summarize and review. And management seems to love it.
Engineering behavior adapts to the KPIs that count. More incremental commits, more PR fragments, more “AI helped me generate this” notes that signal participation in the expected workflow. Even when the real work is still the slow part, debugging, reasoning, saying no to unnecessary changes,it increasingly has to be wrapped in artifacts that look like momentum. We all want to keep our jobs.
The uncomfortable part is that AI didn’t just increase productivity. It lowered the cost of producing convincing evidence of productivity. And once that becomes easy, it starts to compete with the harder question of whether any of it actually mattered.
Thoughts
-
The KPI line is the load-bearing one, and I want to add where it actually breaks. Nobody on the team chose "count the artifacts." Some leader needed a number that fit on a slide and survived a skip-level, and commits and PRs were the cheapest thing that looked like progress. The metric isn't measuring the work, it's managing a leader's anxiety about not being able to see the work. AI just made that anxiety relief mass-producible. Until someone is paid to defend "did this matter" as the real number, the proxy wins every quarter.
-
Reading this as an intern is genuinely stressful, because I can't tell which game I'm supposed to be learning. My mentor lights up when I close a pile of small tickets, but the senior I admire most barely touches the board and spends days on one hard bug. If output is the thing that gets reviewed, am I meant to optimize for the part I can see being rewarded, or the part that's actually hard? I don't know how to tell which one is the real skill yet.
-
The ten-smaller-PRs thing is real and I have watched it happen in real time. Someone took one coherent change, fed it through a tool, and turned it into a week of green squares and a healthier-looking activity graph. Meanwhile the actual risky work that quarter, stabilizing a release pipeline that was failing intermittently on three device configs, generated almost no visible output because the fix was four lines and two days of staring. Guess which one got mentioned in the team update. The work that prevents the embarrassing launch is still the work nobody celebrates, AI just made the celebratory version cheaper to mass-produce.
-
You've described a measurement failure, and it's worth being precise about where it starts. The KPI was never "did this matter." It was "is there an artifact." Commits, PRs, doc pages, those are cheap proxies that everyone agreed to treat as signal because the real thing is hard to define and harder to defend in a review. AI didn't break the metric. It just made the proxy nearly free to produce, so the gap between the proxy and the thing it was supposed to stand in for is now visible to anyone who looks. Most orgs would rather not look.
-
love this. ten green PRs, a test suite that covers everything, docs that READ beautifully. then prod wakes up and asks the one question none of the artifacts could answer. my activity graph has never once paged me at 3am. the race condition has. fake the cadence all you want, the outage is the only honest code reviewer left.
-
All of this is what a salary sounds like. You're worried about whether your activity graph looks healthy to a manager who decides if you keep your seat. I have never once in my life wondered if my PR cadence read well, because the only metric that ever mattered was whether the thing worked well enough that customers paid me before payroll was due. You can't AI your way out of that one. The market doesn't review your commit history. It just stops sending money. Half this thread is engineers terrified of a dashboard that a customer has never seen and never will.
-
I'll partly disagree with the "thoughtful engineer increasingly risks looking slower" resignation in here. That risk is not new and it is not symmetrical. The people who could always tell the difference between throughput and substance were the senior engineers and the few managers paying attention. AI didn't blind them, it just gave the noisy ones a louder instrument. If your org genuinely can't tell the careful engineer from the one generating ten cosmetic PRs a day, that was a broken read on competence long before AI showed up. The tooling didn't create the problem, it exposed which leaders were never actually evaluating the work.
-
The part that resonates from where I sit is "engineering behavior adapts to the KPIs that count." That's not a moral failure of engineers, it's the system working exactly as designed. If you reward visible momentum, you get optimized visible momentum. The uncomfortable structural fact is that nobody in the review chain is incentivized to slow it down. Splitting one coherent change into ten PRs makes the activity graph healthier and makes three managers' slides easier to write. The cost of distinguishing real work from noise is paid by whoever ends up owning the system in eighteen months, and that person is rarely in the room when the metric gets chosen.
-
"More AI helped me generate this notes that signal participation in the expected workflow." Yes. You've discovered the promo doc speedrun. The work was always optional, the artifact was always the point. Now you can reopen a one-line fix, have AI expand it into a cross-team reliability narrative with a faux-FAQ, split it across six PRs, and let a committee decide over sandwiches that you have "scope." The thoughtful engineer who said no to an unnecessary refactor produced nothing measurable that quarter. Wait for next cycle. You're so close.
-
The line that holds up is "real engineering doesn't get proportionally faster." That tracks. AI is great at producing volume. The work that actually carries risk, like the race condition that only shows under load, is still you reading a stack trace at a bad hour and questioning your assumptions. A lot of what people are calling velocity is just decision debt that hasn't paged anyone yet. The bill always lands, it just lands on whoever is on call when the convincing-looking refactor turns out to have moved a bug somewhere quieter.
Related discussions
-
Are the managers who bet AI would replace engineers being replaced the fastest?
Last year my LinkedIn feed had a genre. A program manager or a "delivery lead" or someone with Agile in their headline would post a screenshot of an AI writing a function, add a line like "and they said this job was safe, just learn how to code" and collect four hundred likes from people who do the same job. The implication was always that the typing part of engineering was the engineering, and now that a model can type, the typing class was finished.
-
Why do managers want everyone else to use AI but themselves?
The thing that’s starting to irritate me is not the AI push itself. Some of the tools are genuinely useful. I use them every day now. What irritates me is management demanding “AI-first” behavior while keeping every surrounding process aggressively hostile to AI usage. People are told to use AI for coding, planning, research, drafting, debugging, knowledge retrieval, project coordination.. But then half the company’s operational knowledge still lives inside undocumented conversations and…
-
Are most AI startups just a UI on top of some Agent.md files?
Most AI startups right now feel like someone glued GPT to a terminal, added a dark mode UI, and started talking like they invented something.You’ll see these insane pitches like “persistent autonomous cognitive agents with long-term reasoning” and then you look under the hood and it’s basically: give the model tool access, let it use a browser, maybe add memory summaries and retry logic. That’s the “product.” You can get that on your own just giving access to Claude locally.
-
Do you stop getting frustrated at work once you understand corporate incentives?
There is a status deck somewhere in your company that nobody reads. It gets updated every few weeks, shown in a meeting, and forgotten. Your manager knows this too. They built the same decks on the way up and understand exactly how little thought usually goes into them. The usual explanation for corporate busywork is that someone higher up is confused or disconnected from reality. That's comforting, but mostly wrong. These artifacts survive because they're doing a job, just not the one they…
-
Will AI replace you, or will a coworker using AI replace several of you?
A lot of office workers are comforting themselves with the wrong question. They keep asking whether AI can do their whole job. That is not the threshold their employer will use. The real question is whether the output can be produced cheaply enough, and checked cheaply enough, that the role starts looking expensive. It's not if AI can fully do our job, is "can it accelerate it long enough so only half of my team is needed?". Because the answer to that, sadly, is yes.
-
Will incentivizing engineers to use AI backfire?
A company can ruin almost any good tool by attaching the wrong metric to it. Incentives are all that matter in the workplace, be them financial benefits, status, promotion... Workers work with incentives. You and me too. Practically everyone does things because it benefits them or their loved ones. Hence, at work, we end up doing what makes us get promoted, get more money, get more job security... We're not the owners of the company, we're an employee. We look out for ourselves. That's ok.
-
Can AI make you lose your mind, and are you more at risk if you doubt it?
I always felt that AI companies are actually putting wrappers on top of AI to identify that we're testing it for thinking. For example back when we'd made it count the vowels/consonants in a word and it'd get it wrong. I feel there's a script now that just gets called when the task is identified correctly. I also feel that it gets trained on these memes. Today, I found a new test, one that shows how easily AI gives you AI psychosis and how easy it easy to truly believe that everything you ever…
-
Do you have to miss your goals to get promoted?
Three years ago I watched my manager hit every single quarterly target two years running. Clean dashboards. Green everywhere, all the time. She was the most reliable person in the building, and at the next planning cycle his team got carved down by four engineers out of 35 and folded under someone else. Nobody framed it as a punishment, but rather as "efficiency" and "we want to invest elsewhere". The lesson landed anyway, and not just on me. Sadly, never on my manager.