[xyst]

Right. Instead of many people in a business (ie, accountants, lawyers, product owners, software engineers). The “workforce” can theoretically be reduced to specialists within a field:

- “prompt engineer - accounting” who has vast history of accounting practices

- “prompt engineer - lawyer”. No more farming out writing up complicated EULAs to Skadden. Have your general counsel do all of this for you with the help of “AI”. Or maybe have an IPO prepared without the help of a middleman such as “JPM” or “Goldman Sachs”. Billable hours drop significantly.

- “prompt engineer - swe”. Instead of multiple teams of engineers. Have “Devin” scaffold out the basic application while you focus on architecting an end to end solution.

All of those “middle class jobs” once held by mid tier specialists have evaporated.

[sublinear]

I don't know enough about the other disciplines, but the SWE one would be the other way around and serially dependent. The architecture comes first and then the implementation details can be handed off to the prompt engineer.

Then the code review will still require a human senior developer and maybe the architect. The testing after that can be fuzzed by AI but will still need to be confirmed by a human senior QA specialist.

In other words, AI can lower the bar for entry level work (not by much), but it does not eliminate or create any new jobs. It's very similar to what search did for entry level work. Search results "powered by AI" are also over a decade old.

[dghlsakjg]

This has been my impression overall with AI so far as someone with Senior in their title.

AI is very good at doing things that can be unit tested, or doing small implementation details.

It falls flat on its face when you ask it to understand architectural patterns, or even reference other classes/modules in a broader codebase.

The hard part about coding is rarely the code, more often it is understanding the human side of things, and how you can fulfill real world needs of users.

[robjan]

I think it actually replaces an army of junior engineers or your offshore development house. The real engineering continues while the grunt work is farmed out to AI. It also kills the junior engineer since there will be years where they are less valuable than something with a marginal cost close to zero.

[asdff]

It will be interesting to see what of two possibilities happen: mass unemployment or simply more bullshit jobs and steady state employment. I imagine when computers hit the scene people had the same feelings about losing their accountants and human calculators and secretaries and what not, yet there is not a huge population of out of work people, people are doing different roles. You might even argue that while a total headcount per business might go down, it might be balanced by more businesses emerging. Considering the Fed's dual mandate they might opt to incentivize bullshit job growth over some great society reset to either mass unemployment or a basic income utopia.