> technical professionals without deep software development skills, such as IT project managers, data engineers, and enterprise architects,
Perhaps IT project managers get a pass, but imagine being an "enterprise architect" without deep software engineering skills. My first nightmare is to have to work with an enterprise architect that can't code. My second nightmare is maintaining software they built with AI.
Non technical architects are a thing, mostly at nontechnical companies.
I use to work for a health insurance company that had these people. It went about as well as you expected.
Devs were given very basic flimsy architecture diagrams that would only cover the happy path and require constant back and forth to clarify what actually needs to be.
The only thing I came away thinking was that health insurance companies should not exist. Or at the very least, they need real competitive markets. If the majority of your income comes from payments by the government maybe you should just be nationalized and save the public some money at this point.
But I digress.
This is going to become very popular in those industries. It’s similar to joining a company that dumps low-code solutions on you while refusing to allocate any resources to build in-house tooling. Maybe at least the AI outputs will provide solutions using common libraries and tooling. That would be a win over low-code tooling imo, at least from the perspective of a dev that would be maintaining it. At least that’s my inclination.
> the pricing model seems weird: it's free to create the app, but you pay by the user-hour.
This seems incredible for proof-of-concept prototyping, to quickly validate ideas. And if one of these does turn out to be a good idea, it then shouldn't be too expensive in terms of user-hour to just use it as a reference while building the production-grade implementation.
It looks like it enables people who don't understand tech to use natural language to deploy resources on AWS that are billable, so I bet the engineering resources Amazon put into this won't be wasted at all.
AWS has for years spun resources for you via various wizards and "service of services" and it's not uncommon that the resources are wildly cost-inefficient vs spinning up your own.
Huh. That’s like saying they can’t provide their “well-architected” guides unless they can prove it’s the least chargeable (and obviously there are other ways to judge a system beyond price).
> Build secure, scalable applications in minutes instead of days—no professional software development skills required
What is the big push for companies to NOT have professionals involved. These things are not trivial, at least to get right. Not having a professional involved is not the benefit they think it is.
I would offer a counterpoint: most software in existence was written by not-software-professionals in Excel (most likely poorly).
Within reason I think there is a rational basis for not having to involve software engineers for every project - especially if the SMEs with understanding of their requirements are the ones building it.
This will probably fall over in the same space as Excel spreadsheets do though, when the domain complexity outgrows it, way before anyone is able to recognise that.
The only reason the world hasn't collapsed into a giant puddle of N-squared recursive CPU execution is because Excel's limitations place natural boundaries on the blast-radius of any single SME's poorly built software before they are forced to engage someone who has at least SOME idea WTF they're doing.
I say this as someone who is much closer to the SME end of the spectrum than the Software Engineering Professional end.
"THIS SOFTWARE IS PROVIDED AS IS..." you can't say the same for an employee, and can haul them/the company to court for negligence.
Also who knows if they've trained the model on specific AWS architectures, so the App Studio may simply not create an app that crashes a datacenter (where a third-party dev could, somehow).
I don't think this will ever see great use for building production apps, but I think this could be very cool for requirements gathering.
Give project managers/sales/support desk/etc access to this. Let them build their thing, push it into limited use, evolve the design based on feedback, and then hand it over to developers to reverse engineer into a maintainable app based on the app's behavior and the chat logs that built it.
It's a WYSIWYG editor for the 21st century; we all know the code will be crap, but it's a decent enough starting point to build something real from.
I agree with everything you said except for that it may never see great use for building production apps. I don't think that App Studio is sufficient for production apps on its own, but I do envision a future where AI-based engineers (e.g. future descendants of SWE-Agent[0]) would use something like App Studio as a basis for doing the production work that you outlined there, with some of their automated testing involving replicating the behaviour of the App Studio app as the black-box "legacy system".
> technical professionals without deep software development skills, such as IT project managers, data engineers, and enterprise architects,
Perhaps IT project managers get a pass, but imagine being an "enterprise architect" without deep software engineering skills. My first nightmare is to have to work with an enterprise architect that can't code. My second nightmare is maintaining software they built with AI.