Cropped shot of computer programmers working on new code
Job displacement is a serious issue everywhere, but computer science professionals should prepare for some belt-tightening in their field.
A Semafor article published this month, written by Reed Albergotti, shows how Amjad Masad, CEO of Replit, is enthusiastic about cutting the firm’s workforce in half while increasing revenue by about 500% on the back of agenticAI .
Replit’s new tool is said to be able to “write a software application that runs with nothing but a natural language prompt,” and it will usher in a new renaissance in computing while costing some careerists their jobs.
These are the kinds of insights that lead us to realize that AGI, to some extent, is already here. Much of the rest is semantics – will we have “weak” or “strong” AGI? Or, likely, something in between, say, until 2030?
Accelerated timelines
Covering the rapid innovation, Albergotti points out that Masad didn’t think this kind of progress would be possible until this year.
“If you’ve listened to Masad in the last few years, ‘Agent’ (Replit’sengine) shouldn’t be possible yet,” Albergotti writes. “He said at one point that it might not be possible this decade. Even when he created an ‘agent task force’ to develop the product last year, he wasn’t sure if it would work.”
But Anthropic’s Sonnet Claude 3.5 demonstrated high marks on the SWE bench late last year, and rapid progress in reasoning models has changed the game.
Essentially, Replit had compiled a lot of internal data that represented its knowledge based on the coding process. But now, Masad realizes, that strategy will be outdated as codeless technologies proliferate.
“Just the fact that we’re able to get here without using our own data raises a lot of questions for the industry,” Masad was quoted as saying. “As long as we maintain the rate of innovation and the rate of progress, and continue to deepen it, I think we can continue to be ahead. But the business question is, ‘what is the sustainable gap?’
What, indeed, is the gap for a modern company, as AI levels all playing fields? Will it consist primarily of company branding?
Codeless Services: How Useful Are They?
The rest of the article reveals that Replit’s business model is now a kind of template for Claude’s use of AI coding, leading to that label, quoting Masad: “we don’t care about professional coders anymore.” There is also a reference to something that Albergotti points out was attributed to Masadi himself: “Amjad’s Law,” which suggests that every six months, the parties receive a certain return for knowledge of the code base.
The Personal Computing Revolution
Here’s another part of the article that I particularly liked – it’s where Albergotti suggests that we can take cues for AI and codeless development from the ways visual operating system interfaces developed during the late 20th century. He mentions the “stealth” process of using PC-DOS—that you had to issue commands at the command-line interface to get things done. Then came Windows – a bold new visual design where all you had to do was click a mouse. The menu commands are written for you. They were nested and easily visible with a few points and clicks. And this led to many more people being able to use personal computers effectively.
It’s the same with the codeless process. All of a sudden, you don’t have to know how to code to create software.
Practical Applications of No-Code: Hurdles Remain
Despite all the enthusiasm here and elsewhere, I’d argue that accessible design without code is still some way off.
After playing with some of the prominent no-code systems, what I realized is that if the user tools are not easily intuitive, the non-programmer faces the same problems they would if they were trying to learn a computer programming language.
In other words, how do you make these visual objects do things? How do you build variables into interfaces even if you’re not typing lines of code?
I even asked ChatGPT about the best no-code tools, and it returned options like Wix and Squarespace. But website builders are not new. Software builders are much younger. And I would argue that they are still new enough that most people can’t figure out how to use them.
Maybe we’ll need an app similar to the original Windows desktop GUI to bring all these non-technical people into the fold and let them all be programmers.
At the same time, it’s not too early to think about the victims of this kind of change – the professional coders themselves. It’s easy to forget that these are people, with families and bills to pay. It is high time we think about how to fund human life when AI is able to do most of our jobs.
