How AI will replace programmers (and why that's brilliant)
The quiet revolution happening in code while the world debates whether AI can think
A recent conversation between Y Combinator partners Tom Blomfield and David Lieb over a post made by the former on X.com. Tom's analogy comparing software engineers to "highly paid organic farmers" awaiting their "combine harvester moment" has touched a nerve—and revealed a truth many aren't ready to face.
But the data tells a story that's impossible to ignore.
🧠 The combine harvester moment for code
Tom's farming analogy cuts deep because it's historically accurate. The combine harvester didn't just make farming more efficient—it eliminated 90% of farming jobs whilst increasing food production by orders of magnitude.
"The first computer programmers that existed didn't do a job that was anywhere like today's software engineers," Tom noted. "They were writing machine code. They were making punch cards."
The progression is clear: each generation of programming tools has abstracted away the previous layer of complexity. Now, AI coding agents represent the next—and perhaps most dramatic—abstraction yet.
The Y Combinator data reveals everything
The numbers from Y Combinator cohorts tell the story of transformation in real-time:
→ Two cohorts ago: Approximately 0% of founders used AI as their primary coding method
→ Last cohort: 25% of companies used AI tools for most of their code
→ Current cohort: 33-50% primarily write code using AI agents
This isn't gradual adoption—it's exponential transformation happening at the cutting edge of entrepreneurship.
💡 What Tom's experiment revealed
Tom Blomfield, co-founder of Monzo Bank, decided to test these tools himself. What he discovered shocked even him.
Starting with simple games built using no-code tools like Lovable and Replit, he quickly moved to more sophisticated platforms like Cursor, Windsurf, and Claude Code. His breakthrough project: rebuilding his 20-year-old Tumblr blog.
"In 90 minutes, I set up hosting, I wrote new blogging software, and I migrated 15 years of blog posts over to the new platform," Tom revealed.
But the real test came with recipes.ai—a serious application with 35,000 lines of code, thousands of users, and a full interactive voice agent.
The kicker? Tom wrote zero lines of those 35,000.
"After about the first 5,000 lines, I stopped even reading the code," he admitted. "I just prompt, I'd auto accept, I'd go and make a coffee and I'd come back and a new feature was built."
This from someone who hadn't written code professionally for 10 years—yet found himself more productive than when he was actively coding a decade ago.
📊 The Jevons Paradox argument (and why it misses the point)
Critics quickly deployed the Jevons Paradox defence: as the cost of software development plummets, demand will increase exponentially, maintaining or increasing employment for programmers.
Tom acknowledges this partially: "I basically agree with that as well. But my counterargument is that it won't be humans meeting that demand."
The mathematics are stark. Even if software demand increases 10x or 100x, when productivity per person approaches infinity (dividing by zero, as Tom puts it), the human workforce required shrinks dramatically.
The future of on-demand software
We're moving toward what Tom calls "ephemeral programs"—custom software that spawns when you have a problem, solves it, then dissolves back into the digital ether. ChatGPT already hints at this future, spinning up mini-applications to solve specific user queries.
🌏 Beyond programming: the knowledge work revolution
The transformation extends far beyond software engineering. Tom and David discussed examples across industries:
Legal: Swedish company Lora proved that even lawyers—historically resistant to efficiency tools—will adopt AI when competitive pressure mounts. The old wisdom "lawyers never buy software" crumbles when AI becomes a competitive necessity.
Marketing: Companies like scriptbee.ai represent an entirely new category—marketing agencies run by AI agents, built using similar coding tools Tom demonstrated. These autonomous systems handle campaigns, content creation, and client management with minimal human oversight.
Medicine and Finance: Y Combinator has seen a surge in companies automating knowledge work across these sectors. What started as "fringe" ideas two years ago are now thriving businesses.
"We see plenty of examples of these companies succeeding and actually being used in these industries," David noted.
The physical work advantage
Some professions enjoy natural protection from this wave:
→ Surgeons, plumbers, electricians: Physical presence and dexterity remain human domains
→ Trade unions and regulatory bodies: May create artificial barriers to protect jobs
→ Niche specialisations: Complex, human-judgment-heavy roles will persist longer
But even these face eventual disruption as robotics advances.
🤔 The agency and taste question
A crucial question emerged in Tom and David's discussion: what remains uniquely human?
David argues it's problem identification and taste: "If you look at the best software that you use in your life, very likely there is a single human being behind that team that built that and that person obsesses over making that product excellent for the user."
The challenge isn't technical capability—it's programming AI to have that obsession, that taste, that drives exceptional user experiences.
➡️ The founder's unprecedented opportunity
Both Y Combinator partners agree: there's never been a better time to start a company.
"These tools give high agency individuals superpowers," Tom emphasised. "If you are a potential founder thinking about starting something, I don't think there's been any point in history that's been better than today."
Why now is exceptional
Smaller teams, bigger impact: Two-person teams build what previously required 40 engineers
Faster profitability: Companies reach sustainability quicker, often without Series A funding
Lower barriers: Technical complexity no longer blocks great ideas
Industry transformation: Law, education, medicine—formerly software-resistant sectors—are suddenly open to disruption
Tom's advice for aspiring founders
Tom's recommendations are practical and urgent:
"The first thing I would do, and I'm encouraging all of my friends to do this, is just to stay up to date with the latest tools. They might not be perfect for your industry yet, but I'm betting a lot of money that at some point they will cross that tipping point."
David adds the human element: "Get good at identifying human problems to go solve... if you overindex and get really good at just understanding people and seeing problems... that skill relative to all the other skills needed to be a good founder is going to be the one that is more important."
The transition challenge
Both acknowledge the darker implications. David highlighted the potential for "hundreds of millions of people displaced" during what could be an extremely rapid transition.
"The idea that they're going to retrain as in a different job I think is going to be extremely painful," Tom admitted. "I can see the societal impact and turmoil being very very grave for 10 or 20 years as that transition happens."
Yet they maintain optimism. Tom argues: "If you just look at each step along this progression and you had to choose as an individual, do I want to live in this world or the world one click further into the future? I think in all cases the world and your experience and what you can do with your life is going to be better in the next click forward."
You might have these questions
Are these AI coding tools actually reliable for production systems?
Tom addressed this directly: while not perfect today, the rate of improvement makes reliability questions irrelevant. "The argument that these tools are never going to be good enough I think is just a losing proposition."
How can current software engineers prepare?
Both partners recommend immediate experimentation with AI tools, even if they're not perfect for your specific use case yet. The competitive advantage belongs to early adopters.
What about code quality and maintainability?
Tom's experience suggests this concern may be overblown. He stopped reading the AI-generated code after 5,000 lines and still successfully built a complex application. The focus shifts from code craftsmanship to system architecture and user experience.
Will there be any programming jobs left?
"There will be demand for smart people who know how to wrangle these AI coding machines," Tom predicts. "If we want to call those people software engineers, so be it. But I think the job is dramatically, dramatically different."
How fast is this transition happening?
The Y Combinator data suggests very rapid adoption among startups. Tom believes software engineering jobs as they exist today "will not exist in five or 10 years."
➡️ A future worth building
The conversation between Tom and David illuminates both the promise and peril of our current moment. We're witnessing the kind of technological shift that happens perhaps once in a generation—the mechanisation of human intellect itself.
Tom's final insight resonates: "I genuinely do believe right now and probably the next 5 years is the best time in the history of humanity to build something from scratch."
As someone building tech companies across continents—from Unschool in India to caisy.io in Germany—I see this transformation happening in real-time. The founders who embrace these tools aren't just building faster; they're reimagining what's possible when human creativity meets unlimited execution capability.
The question isn't whether this revolution will happen. Tom's Twitter thread may have sparked controversy, but the Y Combinator cohort data provides undeniable evidence: it's already here.
The question is whether you'll be part of shaping it, or watching it happen to you.
This analysis draws from the Y Combinator "Breakdown" episode featuring Tom Blomfield and David Lieb.