I just attended a tech conference for a major tech company where many of the sessions were focused on their low-code GUI app builder. Overall, I enjoyed the conference and all the speakers. Truly, some good work going into their product and a different perspective than I’m used to.
Something on my mind, though, is that I don’t think low coders (in general) realize yet that agentic coding exists, or at least what it means for them.
Picture this. If you’ve spent the past decade+ riding the low-code wave, you’re probably amazed at how far you’ve come. Dragging and dropping can get you really far and allows for building many different types of applications and websites.
I remember when I first started getting into things beyond basic websites, I tried a tool called Dropsource that exported native Java and Swift mobile apps from their GUI builder. Great team, great product, didn’t end up getting popular. Using that opened my eyes to what was possible with code and gave me a stepping stone for getting from zero to, let’s be honest, somewhere beyond zero but not quite one.
Even though I haven’t regularly relied on using drag-and-drop low-code tools for some time, I’m consistently impressed by the progress. Still, I can’t help but think that the number of amazing new features a low-code tool can have is now pretty limited. It takes a lot to be impressive. Sure, graphics tools like Figma adding low/no code features help close the gap between UI and using a design, but ultimately, that’s not very notable outside of those who use Figma. It seems that we’re very far along on low-code’s S-curve. And that’s okay, maybe even natural.
Innovation S-Curve Made with Claude
If you’re thinking ahead or are familiar with technology strategy, you’re probably thinking that a technology maturing and living happily ever after isn’t the whole story. Typically, a given technology will reach maturity and slowly keep getting slightly better, but then get disrupted. Think Horses and Automobiles or Video Killed the Radio Star.
▶ Spotify Embed: Video Killed The Radio Star - Single Version
The problem with low-code is that, as expected, there’s a new S-curve that just so happens to converge on similar use cases to low-code. There’s major hesitation among those who have spent years only doing low-code to, in what I imagine to be their perspective, regress back to using code. That’s understandable…but if trends continue, the idea that dragging and dropping together UIs and logic will still be a thing in 2035 seems ludicrous. Maybe even by 2030 or sooner.
For many years, it was always preached that low/no code would take the developer’s job, but if low/no code doesn’t watch out, it will be first on AI’s job-taking list. And here’s why.
AI is probably the most influential technology of this decade (fingers crossed I’m proven wrong). It is pretty amazing what new opportunities it unlocks. Like I mentioned earlier, sometimes innovations in one technology revitalize the improvement of another. The transistor did this for pretty much every technology. AI might be similar, and in many technologies and industries, it already is.
The problem with low-code is that it is at risk of becoming obsolete in favor of a more efficient alternative that just so happens to be what it was intended to replace.
The idea that AI somehow needs a user interface or other abstraction on top of an already high-level language is seemingly misguided. Templates and UI libraries are great. They’re hugely beneficial in fully human-developed projects where you want visual and functional standards. But hiding languages behind buttons and switches is irrational.
A large language model is literally designed for language.
It is as if we, as humans, have developed our most valuable technology, written language, and decided, “Wait, guys, what if we went back to hieroglyphs?” It doesn’t make much sense.
Photo by AXP Photography / Unsplash
Button clicking is fine for those who don’t fluently speak the language, as babies do, pointing at a toy or bottle, but with AI, everyone can speak any language and express their ideas using terminology and descriptions that make sense to them (even through voice!).
In a world where AI is faster than any human at reading and writing, it doesn’t make sense to trivially click buttons. Or write code for trivial tasks.
Sure, there’s an argument for keeping low-code similar to how low-code might make sense to have the ability to see the code or add custom code. But is this really worth the technical investment to build the buttons and interfaces vs just reading code created by AI directly, or better yet, having AI read it? That seems like a stretch.
AI doesn’t need low-code abstractions because natural language is now the abstraction. Low-code replaced many trivial development tasks. Now AI is replacing triviality.
Liked this? Get the next one.
No algorithms or news feeds. Add the RSS feed to your reader and new posts show up as soon as they're live.