(Editorial note: I originally wrote this post over on the Hit Subscribe blog. I’ll be cross-posting anything I think this audience might find inte...
For further actions, you may consider blocking this person and/or reporting abuse
This article gave me pause because I advocate using AI, and I often speak of it as a companion for deeper learning. However, I understand the nuances that you discuss. It's primarily why I'm looking forward to carving out a space as a debugger, especially with 'vibe-coded' projects. It is funny. A while ago I listened to an 'HTML All The Things' podcast episode that discussed the 2025 job market and said the people who will survive in this new market will be the ones with deep knowledge, so I remain hopeful.
Carving out a niche fixing vibe coded projects makes a lot of sense to me. IME, every time there's a happy path cost reduction push (e.g. AP "offshoring" back in the 2000s when I was earning my living as a software engineer), there's a lucrative, if less shiny, living to be made helping people when they happy path isn't quite as happy as they anticipated. In the 2000s it was "hey, turns out things go wrong when you outsource collaborative labor to folks offset from you by 12 hours and frequent language barriers."
Love your take on technician vs strategist. I found that situation in my career too, I coined a name for it "the not-a-VP effect." Often we give our solution/opinion/take and it passes unnoticed. Then a VP/director/executive comes and says the exact same thing. And guess who people hear? Not the one who writes code.
Yep, exact same dynamic. Rightly or wrongly, the world segments projects into strategy and execution and tends to prefer to have separate people doing them. "Consultant" vs "staff" is an easy way to segment that, but organization leadership also tends to default to the "strategy" group.
Lovely
That completely nails what’s been bothering me about all the 'get good at prompting' takes - feels like a race to the bottom.
What’s one example where you’ve seen someone really carve out a unique niche alongside AI, not just compete with it?
Not really, but I wouldn't necessarily know if I had. Generally if I hire freelancers or vendors I'm not super interested in the details of how they execute their trade, provided I get results. Almost without exception I'm only aware of other people using Gen AI techs because they tell me they're doing so, rather than because they're doing something dramatically better or more impressive.
That's not to say people aren't doing it -- I just don't have any observed experience.
Wonderfully put blog entry it reads as a poem written from the deepest crevasses of my mind, it addresses all the important issues I think about in the current day to day as well as what I have mentored in the past
Well, Done Good Sir!
Much needed. Thanks!
Great post
If you're good in whatever you do now, you'll find ways how to utilize AI tools to become even better. And you very much should do that. It is a tool. The way of using is up to you. And of course, solely being proficient in putting paint on canvas with a brush doesn't make you a famous painter itself. So just trying to be "good at promting" surely isn't a good goal to follow.
I've used AI in a few projects so far and routinely use it while writing code, but I really don't understand why people are so excited about it... It can't solve simple typescript typing issues, nor is it capable of debugging more complex issues with package dependencies. Maybe it works better in more standardized languages, but in the JavaScript ecosystem, it will fail.
On the other hand large companies like Microsoft apparently don't care about quality, so maybe AI will be a good fit for companies like that.
After a dialog with GPL about this text, here are some snippets:
The author is correct that LLMs (like me) dramatically accelerate the commoditization of knowledge work. What used to require specialized technical skill is now accessible to far more people, with lower barriers (language-based prompting instead of formal training).
The article rightly points out that "prompt engineering" is not a deep or defensible career moat. LLMs are explicitly being optimized to require less prompting expertise over time.
Developers who use AI as a force multiplier while moving upstream strategically.
Many developers today are using me (ChatGPT) to boost velocity while still making architectural decisions, solving novel problems, and engaging in high-context human decision-making.
Developers focusing only on tactical, repeatable outputs (especially without business context or human judgment) do risk being replaced by cheaper, faster AI-supported workflows.
The article does not fully recognize that AI literacy will increasingly become table stakes for even strategic or leadership roles in tech.
Ignoring AI completely risks irrelevance.
Yes, superficial prompt engineering is shallow.
But deep LLM tool integration, fine-tuning, RAG systems, AI workflow orchestration, and AI safety/compliance roles require sustained technical depth.
The article dismisses all AI expertise as "weak sauce", which is too simplistic.
The article highlights real LLM weaknesses (hallucination, imprecision), but implies they will remain fatal flaws. In reality, these limitations are narrowing with each new AI iteration. For developers ignoring AI on the basis of today’s flaws, this is a dangerous bet.
No, developers should not reject AI tools out of fear of commoditization.
The correct response is to integrate AI tools selectively and strategically while moving towards higher-value roles.
AI is here to augment your craft, not replace your identity as a problem solver and thinker. The article offers an important warning—but the smart developer’s path is synthesis, not rejection.