I think it won’t replace people for quite some time.

But it will replace people that don’t want to use AI! It still requires someone to closely monitor, tweak a prompt, change direction, or manually fix some code.

That said, a product manager or QA specialist who is motivated to learn how to prompt, “vibe code”, and interact with an LLM is going to be much better off than a software engineer that just wants to hand write code.

Where things also really get interesting is in the micro SaaS space. I reflected on this recently at https://weaver.so/articles/vibe-coding-and-saas.

I think anyone working on small tools that improve a specific process (but doesn’t fit a business need precisely) will find themselves losing customers. Motivated folks will just start building their own tools to meet their needs instead.

Agree with your stance on micro SasS space, I've just built two internal tools for my marketing co-founder, one was a Tiktok dashboard to track impressions for all our accounts. Another was dynamically rendered pages for the Tikok posters, that rendered posting info and schedule from notion. Both took 1 - 2 days each. They are very basic but do the job and i don't have to pay $20 - $50 for another service.

Engineering is being good at using engines. Knowing all the systems that this engine is composed of, optimizing it, making it do what it's meant to do with minimal error, coming up with tests to verify that it does, facilitating all technical communication.

There's plenty of that work coming up. People don't like working with a completely different engine but it's inevitable; before this everyone was complaining about JS frameworks. Once the AI dust settles, it's back to spending 20 hours of the weekend learning the next framework.

> especially from VCs and CEOs

That is the key phrase. People who do not do the work want to replace people who do the work with cheap alternatives. This is not new, they just see a new way to accomplish it.

So it will go the same way as all the other ways they try to reduce personnel costs. They will evaluate your cost and the value you bring and compare it to the value they can get using someone/something else that has a cheaper cost. If you deliver equal or better value for the same cost, you keep your job. If you do not, you lose your job.

Same old story, new AI flavor.

Trying to predict things is not my preferred thing.

I try to make myself easily replaceable by others. Non-special, in the context of a job. It's a hallmark of teamwork, for decades.

To face that prospect with AI as a replacement seems unexpected and the wrong kind of question. I can't transfer knowledge to it like I would to a team mate.

I don't understand your use of 'artistry' here. Do you mean your 'craftsmanship'? I will answer like you meant that.

Do something that is rare, valuable and honest. Share it with others so it doesn't fade away. There are many of those kinds of things. It does not guarantee you will be valued, and that's ok.

I think AI can and will replace many stakeholder-facing SWE eventually. It's not there yet but I don't see why it can't reach there. On the other hand, it needs very good integration and a complete revamp of human workflow to adapt to AI (hint: a lot easier than making AIs that adapt humans). Even if AI never replaces some roles, the increased productivity means that less humans are needed. The next consulting wave is FAANG pushes internal teams to integrate with AI and then these teams leave to consult other companies in AI integration to get big $$$.

I work as a data engineer, and I'm pretty sure that the current AI is good enough to do around at least 50% of my work. I look around and believe the same applies to my data analyst colleagues as well.

I believe stakeholder-facing engineers are the first to get hit, like data engineers and frontend developers. The reasons are:

- In general their work isn't very technical (it can be, but not common) and often repetitive. There are tons of examples and documentation online.

- They face business stakeholders directly. The communication is not always smooth, so the stakeholders have the incentive to train themselves talking to AI trying to get things done ASAP.

What's the overlap between AI and low-code tools as far as your workload goes?

We’re at a stage, where we have to come up with ever more complex benchmarks, as the old ones are getting saturated, yet, anywhere I look, I don’t see any significant productivity boost. I can’t point to any product on the market or a company and say that they visibly increased their velocity.

Yes, it useful, but anyone saying that almost all software workers are getting replaced incredibly soon is either misinformed, or is selling something.

There is going to be a workforce reduction for sure with a shift towards more AI tools, however its going to be driven by the economic policy rather than usefulness of AI, as US economy slowly dwindles away. Of course service quality is going to drop as a result, but thats just how the cookie crumbles.

Even if traditional "digital world" SE mostly goes away, the capabilities causing that will make focus shift to automating virtually everything else very quickly, so we'd be in good company...

Pivot into robotics?

Programmers who know to use ai will replace thosw who dont.

That seems rather dubious. Some developers are far more replaceable than others. Developers reliant upon AI to do the work are only an administrative step away from replacement.

eh... first company that fires all of its devs for AI slaves, and then doubles down on the decision when it becomes obvious that was a poor choice will become an interesting case study for MBA's on what not to do.