For as good as AI is at various things, it can be something that can be used to create something that is very insecure and error prone very quickly should the user start asking for conveniences that come with security considerations.
We will see though.
Given all the AI discovered vulnerabilities that's been cropping up, were handcrafted programs really that much better?
No hand crafted programs where not 'better' and I am not saying that we can ignore and not use AI, I am saying that knowing what you are actually producing and how it will run on a system and knowing what are the right things to want are better then someone who producing a program that does everything they want but what they wanted was bad, insecure (by design), poorly thought out and scales badly (in terms of adding new features).
AI can produce good code, good enough code and build excellent scripts and wrangle tools like socat, but leaving it on it's own and just asking make me a web app that does X. Further how many bright ideas will come to the users mind when they get excited for example "actually I don't like the way this database works each user should have a table, I think that would be better and easier to work with and look better as I can just inspect a user like a report!", time will tell but users of AI to vibe code apps who will be liberated from the details of how will not question their why's.
We don't need a fault-free method of producing software to realize that any levers used in the process can only amplify errors and their consequences. Sometimes the answer is to go slower and make it correct before pouring wattage on the system. Sometimes the answer is go go fast and break things.
The only certainty should be that there is no universally correct answer to how to seek answers. Sometimes an ad hoc app built quickly without looking at code is the best solution. It certainly shouldn't be a surprise that any gains therein are not without cost. Or that any such machinery can only ever make knowledgeable people doing slow work on fundamentals more necessary, not less.
That means we're going to be leaning on a 3rd party solution such as Cloudflare's security suite. That was at least part of the story for the recent PHP issue.
> That was at least part of the story for the recent PHP issue.
For a widely known exploit effecting all versions of PHP, they can do this but I don't think they can do this for unique or just less wide reaching security vulnerabilities. While some inspection could be done and rules could exist to flag up potential attacks (I have seen this suggested), I think this might be part of a solution but this was still just conditional rules filtering these requests, I don't think we could produce programmatic rules to filter for exploits in a crappy web app or even an expensive AI model (I would imagine WAF will never be able to run a lot of the traffic though even fairly small AI models as they will cost too much for most applications per a packet).
I think it’s not so much about being able, but more about wanting to
Think about what drew you to programming in the first place. Most people in the world are not drawn to those things, even if AI can do them for them
Organizations might need fewer programmers per application, but they still want someone (call them something else if you want), that will own and take accountability for the systems they implement. Usually non-programmers (or non-technical people), not only have very little interest, they also might not have be abilities to do the type of work dynamic required (long quite hours, lots of screen time, not that many meetings or social time as other roles)
I'd guess more than a quarter of the visitors of this site would answer with "money" at this point. Most people do seem drawn to that.