How to Automate Software Engineering

Another thing to keep in mind is that it is also possible that many startups attempting to fully automate software engineering roles will die / bankrupt or shut down trying before it 'eventually' happens; whenever that is.

builder.ai was the first casualty claiming to use AI to replace software engineering teams to build products. Now bankrupt. [0]

It could cost a lot more money than initially estimated than other industries if (by their own admission) it is the last profession to be automated.

Let's see in the next 10 years if some of these 'startups' are still around on this 'mission' and whether if software engineering will be fully automated and dead for humans in the next decade which that is the narrative that they are all hyping for more VC money.

[0] https://finance.yahoo.com/news/builder-ais-shocking-450m-fal...

This "problem" is pathological and those selling solutions are misleading those who buy into it.

Software engineering often isn't just answering "how" to accomplish something, but also all the five other questions starting with 'w'. Software ate the world a very long time ago, so people being allergic to code while trying to run a business is only sounding more and more absurd as time marches on.

What are we really trying to solve for here? All that could be automated reasonably well has been already. In most cases you do not want a stochiastic result, but exact code to be reused. We use libraries in our code and we have reproducibility of results. The code that needs to be written for most applications is minimal and only grows large as the business refines what they want. This code is unique and virtually worthless to any other business. The code mirrors the organization. We already know all this for decades. It's very confusing to me to keep hearing about this insistence that we need to automate software engineering.

> Once this occurs, many software engineers could perhaps transition into adjacent positions that rely on similar expertise but are significantly harder to automate, such as software engineering management, product management, or executive leadership within software companies.

> In these roles, their responsibilities would shift from writing code and debugging to higher-level oversight, decision-making, and strategic planning—until these responsibilities can be automated too.

When they spoke on Dwarkesh's podcast they seemed to think this would take 30 years. Not sure why we coders are so quick to be automated but the rest aren't.

https://www.dwarkesh.com/p/ege-tamay

> Software ate the world a very long time ago, so people being allergic to code while trying to run a business is only sounding more and more absurd as time marches on.

I don't agree with this at all. Part of the pitch of code is packing really nasty semantics into a single interface. This inherently reflects in the services we provide to clients. This interface should naturally correspond to UI, if that's what you offer. If you can only express the interface for your service in code, you've failed.

Unless of course you primarily operate a code-centric product, like an sdk.

We really have to STOP listening to these people and see that they ALL have a vested interest.

In the third paragraph of the podcast Dwarkesh already told you that he already invested.

>>> (disclosure - I’m an angel investor),

At that point the co-founders and Dwarkesh himself will agree on everything and they will say anything to get more VC money - Even if the timelines are unrealistic. (Because that is the scam).