The End of Software
evanjconrad.comMy Alexa cannot even find the right episode of paw patrol no matter how hard I try to say episode 84...
Also I've never seen a stakeholder being able to expressing clearly what they want. To get a somewhat usefully specification it's an uphill battle. Programming is almost trivial after that
I guess one could say that actually complicated code makes up for less than 5-10% of any non trivial project. Beyond that I agree with the underlying point, getting stakeholders to clearly state their needs in a way that’s actionable is the hardest part.
Yup. And beyond that, once you get your stakeholders to express their needs clearly enough to cover all corner cases, what you basically have done is programming. Except the source code is in English than in C or Java or what have you, but it's just as complex to understand.
I wrote a long comment detailing why I think this won’t work at all, certainly not in the next five years. But I realized that the article isn’t that far off, it just got one detail wrong. If the imagined future comes to be, Apps don’t make sense as a concept anymore. If we have AI that can generate an app for any use case that I might have, then we also have AI that can just do the thing. No need for bespoke interfaces.
That is a future I find much more reasonable. But looking at the state of AI Assistants today I still have a very hard time believing that this will come to be in the next 5 years. But in 20 years? Sure, why not.
Yeah, no. This might work for 80% of software (via the 80/20 rule), but the other 20%? You really think that an AI will be able to generate a stock trading system, or a high-performance web browser? I don't think this is possible in 10 years (barring AGI), let alone 5.
Some smartphone apps barely count as software. Equating them to all software that exists, is almost insulting to me.
So far my career has been secured by the small details, edge cases and the field of scientific computing. I think these tools will not work well for the latter.
But I’m still planning my career change. Not because of AI.
If the level of complexity in todays development is any indicator I probably can’t do this past 40.
I've switched from Java coding to DevOps, and I must say that the complexity level dropped like a rock for me. I won't lie, it still gets complicated, but I'll take troubleshooting Docker or Apache over Spring annotations, any day.