>Software has been the force multiplier behind nearly every business transformation of the last two decades. The constraint was the cost and time of producing and managing it. That constraint is collapsing. As the cost of producing software collapses, demand for it will expand. Last year, the developer platform market used to be measured in tens of dollars per user per month, this year it is hundreds/user/month and headed to thousands. Not only is the value of software for builders increasing, but we believe there will be more software and builders than ever, and we will serve an increasing volume of both.
almost like a copy of my post :) https://news.ycombinator.com/item?id=47982975
We've seen these tech waves several times - C and COBOL instead os ASM, CAD/4GL, template generation, Visual Basic and the likes (good old Delphi), Java (which allowed to a lot of mid-inept people to write compilable non-immediately-crashing programs), spread of python, and now AI. Every time we have an expansion of the industry, and every time glorious promises which get delivered on modestly. The point here is that they get delivered on.
And with AI i suppose it will be similar, though much better than before. In those previous waves human brain was the limit. This time we throw that limit away from the start - nobody will be able to comprehend the sheer amount of AI-generated code. Yes, that approach will hit some limit down the road of course too...
> The point here is that they get delivered on.
... so where's the delivery?
I have no doubt that AI is making some programmers quite a bit more productive. But if it is even 10% as good as all the marketing claims, we should be seeing an explosion of new tech startups, and a huge increase in feature shipping rate and number of bugs closed. Why isn't this obviously happening? Where's the next Dotcom Boom or Cloud SaaS Explosion?
What I am seeing instead is million-line AI slop pet projects whose sole "user" is its developer, and large companies falling over each other to enshittify their products. If there's no genuine user value being delivered, who's going to pay for those thousand-dollar-per-month developer tools?
>Where's the next Dotcom Boom or Cloud SaaS Explosion?
i see it isn't your first rodeo :) So, in Dotcom the companies needed huge financing for hardware and those money were the main limiter, in Cloud SaaS era small teams with relatively small financing mostly for salaries were able to deliver large - AirBnb, Uber, WhasApp, ... - and the employees, their brain abilities and their ability to work together were the main limiter. Now with AI we don't have these limiters. I'd say the slopped up Claude Code and OpenClaw are the examples of the new wave which is just starting.
>large companies falling over each other to enshittify their products.
Oh, yes, each wave the software is even more sh.tty than before, and this time i think we're really in for a shock to our imagination of how sh.tty it can get. All these datacenters here and later in space would need some slop to churn through :)
My bet is that we'd not have a software as a static set of bits existing for more than one execution. I think we'll have Just-In-Time software. An ephemeral one. It will be generated on the fly for specific task and discarded after. That will keep those datacenters busy at least for some time.
Another storyline i, with some horror, expect is merging of the coming boom of actual physical robots with the boom of AI-slopped software - that should be fascinating :)
I feel like "just in time" software is something we already had-- things like VBA and AppleScript showed there has always been an audience for scratch-your-own-itch tooling for work scenarios that aren't programming-centric.
It would be irresponsible to treat it as completely ephemeral though; clever tooling would make it easy when you remember "I already solved this issue 3 months ago, let me pull that back and reuse it."
What terrifies me is doing it with the current slopbox user experience. From a UI perspective, it's clumsy system that discourages developing mastery in favour of guesswork and gacha. (When you said the wrong thing in a classic command line, it at least told you so rather than trying to stagger along with it) And as an executing tool, it's simply sluggish-- once you've expressed what you want, Claude takes minutes to do what a regex does in milliseconds.
I wonder if the latter is fixable-- pre-configure the bot to generate answers as reusable code instead of slowly pumping the changes themselves.
> I think we'll have Just-In-Time software. An ephemeral one. It will be generated on the fly for specific task and discarded after.
For years I've been telling people that every office worker should be able to do at least some programming, just to avoid ever having them spend several days manually repeating the same handful of steps on a large set of data.
I can 100% see AI taking over this market. Teaching office workers to write half-decent prompts is probably easier than teaching office workers Python. But you don't need a $1000/month subscription to write barely-good-enough-to-run-once one-off scripts, and you can't build a business solely on ad-hoc scripts.
> the employees, their brain abilities and their ability to work together were the main limiter. Now with AI we don't have these limiters
Was it? Don't we?
There has never been a shortage of college kids willing to throw together MVPs. Sure, hacking together the bare minimum of business logic with auto-generated Rails code and a $20 Bootstrap template during a hackathon is being replaced by an afternoon talking an AI into generating a Tailwind-styled SPA in whatever Javascript framework is fashionable this week, but what does it really change? Writing MVP-level code was never the hard part.
The hard part is the engineering behind making it scalable, extendable, and durable. That's still staying the same: you're now just giving the prompt to an AI rather than a junior dev. If anything, having to deal with inept managers now sending full-blown AI slop proposals rather than blabbering a handful of buzzwords and leaving the professionals to fill in the rest is going to slow down our ability to work together.