Will the job title software engineer go away?

It may change. Boris Cherny, the creator and head of Claude Code at Anthropic, has said publicly that he expects the software engineer title to start fading. The replacement labels people float are things like builder or product manager, as coding becomes something anyone on a team can do with an agent. A title going away is not the same as a job going away. Cherny still runs an engineering team that ships more than ever; the people just do more specifying, reviewing, and deciding, and less hand-typing. The label follows the work, and the work is moving up a level.

Are software engineers still in demand?

Yes, and the demand may rise. When software gets dramatically cheaper to build, companies usually build a lot more of it, so more projects become viable, not fewer. Someone still has to decide what the agents build, judge whether it is correct, and own the outcome when it ships. Founders building AI tooling report that demand for engineers to oversee and direct autonomous code generation is increasing. The most exposed role is the narrow, entry-level, ticket-implementing one; the aggregate demand for people who can turn intent into trustworthy software is being redistributed toward judgment, not shrinking.

What should software engineers learn to stay relevant?

Three skills move up in value. First, specifying: describing precisely what should be built in plain language an agent can act on. Second, judging: choosing which of many plausible implementations fits your product, users, and constraints, which is where taste lives. Third, verifying: reviewing and testing agent output, because when code is cheap to generate, checking it becomes the bottleneck. None of these is typing faster. The highest-impact move is to get fluent at directing a fleet of coding agents and checking their work, so you are the person using AI rather than the one out-produced by someone who does.

Will AI Replace Software Engineers?

Q: Will AI replace software engineers?

No, but it is automating the most routine part of the job. AI coding agents are good at the well-documented, repeatable work that used to fill a junior engineer's day: scaffolding endpoints, writing basic tests, fixing common bugs. That part is getting cheap. What grows more valuable is the work a model cannot do on its own: deciding what to build, designing the system, and verifying that the output is correct. The engineer whose value was mostly typing is exposed. The engineer who moves up to judgment and learns to direct agents is more valuable than before.

If you write software for a living, or you are a founder deciding whether to hire engineers, "will AI replace software engineers" is the question sitting under everything right now. The honest answer is no, not in the way the headlines imply, but the job is changing faster than almost any other knowledge job. The part of the work that looks like typing code is being automated. The part that looks like judgment is becoming the whole job. A Y Combinator conversation titled "Coding Will Be Solved For Everybody" lays out where this goes, and it matches what the people building the coding tools say about their own teams. Here is the founder-to-founder version.

The short answer

AI is not going to make software engineers disappear. It is going to automate the most routine, well-documented part of coding, which was never the hard part of the job anyway. What remains, deciding what to build, designing the system, and checking that the result is correct, grows more valuable, not less. The engineers who struggle will be the ones whose value was mostly typing. The ones who thrive will be the ones who move up to judgment and learn to direct the machines that do the typing. If anything, demand for people who can build the right software is rising.

This is a sharper version of a broader question we answered in our honest take on whether AI will take your job: the work gets repriced, not deleted. This post is about what that repricing does to software engineering specifically.

What "coding will be solved" actually means

The phrase sounds alarming until you look at what it means in practice. In the Y Combinator conversation "Coding Will Be Solved For Everybody," the argument is that writing code is becoming something anyone on a team can do with an agent, not a specialized skill guarded by the people with "engineer" in their title. On the teams already working this way, everyone codes: product managers, designers, engineering managers, even finance. Coding stops being a bottleneck and becomes a shared tool.

"Solved" does not mean "no humans needed." It means the mechanical act of translating a clear idea into working code is getting cheap and reliable. That is a big deal, because for decades that translation was most of what a junior engineer did all day. When a model can scaffold an endpoint, write the tests, and fix a routine bug in minutes, the value of being the fastest typist on the team drops toward zero. The value of knowing what to build and whether the result is right goes up.

The job title may change. The work does not disappear.

One prediction from this world is blunt: the title "software engineer" itself may start to fade, replaced by something like "builder" or "product manager." Boris Cherny, the creator and head of Claude Code at Anthropic, has said publicly that he expects the "software engineer" title to start going away, and he is not speaking in the abstract. He has said he has not written a line of code by hand since late 2025, and instead spends his day directing large fleets of coding agents. On his team, he has described everyone carrying the same title and everyone building, regardless of background.

Read that carefully, because it is easy to misread as "engineers are obsolete." Cherny still runs an engineering team, and it produces more than ever. What changed is what the people do: less hand-typing, more specifying, reviewing, and deciding. The label on the role is downstream of the work, and the work is moving up a level. A title going away is not the same as a job going away. Blacksmiths became machinists; the metal did not stop mattering.

This is the discipline shift we covered in the move from vibe coding to agentic engineering: the engineer stops being the person who types the implementation and becomes the person who specifies, directs, and verifies it.

Why demand can rise even as coding gets automated

Here is the part the "AI is coming for developers" headlines miss. When something gets dramatically cheaper, people usually buy a lot more of it, not less. Cheaper software means more software gets built. On a16z, the analyst Benedict Evans makes exactly this point: as building software gets faster and cheaper, the likely result is far more software in the world, not a fixed amount built by fewer people. Every company that could not previously afford custom software now can. Every internal tool that was not worth building now is.

That expansion needs people to steer it. In a Y Combinator Root Access conversation with founders building tools for AI agents, the read was that demand for engineers is increasing, precisely because more autonomous code generation raises the need for human oversight, review, and creative direction. Someone has to decide what these agents build, judge whether it is correct, and own the outcome when it ships. That someone is an engineer, whatever the business card says.

None of this guarantees any specific job is safe. The entry-level, narrow, ticket-implementing role is the most exposed, the same way it is across knowledge work. But the aggregate demand for people who can turn intent into working, trustworthy software is not shrinking. It is being redistributed toward judgment.

What the job actually becomes

If the typing is automated, what fills an engineer's day? Three things grow to take its place.

Specifying. The scarce skill becomes describing precisely what should be built, in plain language an agent can act on. A vague spec produces vague software fast, which is worse than slow. Writing a clear specification is now core engineering work, not paperwork you rush through to get to the "real" coding.

Judging. Andrej Karpathy, a founding member of OpenAI and former director of AI at Tesla, makes the point that taste and judgment stay essential exactly as agents handle more of the technical details. The model can produce a hundred plausible implementations. Deciding which one fits your product, your users, and your constraints is human work, and it is getting more valuable, not less.

Verifying. When code is cheap to generate, checking it becomes the bottleneck. Founders who run coding agents describe spending their time reviewing output, doing the QA that used to be someone else's job, because that is where the risk now lives. As a CISSP, I would draw a hard line here: anything touching money, customer data, or security is exactly where a human has to stay in the loop, no matter how good the model gets. Verification is not a chore at the end of engineering. It is becoming the center of it.

Notice that none of these three is "type faster." That is the whole point.

How not to be the engineer who gets replaced

The dividing line is not human versus AI. It is the engineer who uses AI versus the one who does not. The founder or engineer who ships with a fleet of agents out-produces the one doing everything by hand, and the gap widens every quarter. This is also why this question is different from the team-structure version of it: that post is about how the shape of the whole engineering team changes as the pyramid flattens. This one is about whether you, personally, stay valuable. The answer to the second is yes, if you climb toward judgment and get fluent at directing agents.

The engineers who get left behind will not be the ones who lacked talent. They will be the ones who kept measuring their worth by lines of code typed by hand, long after that stopped being the scarce thing. The scarce thing is knowing what to build and being able to tell whether what came back is right.

What to do this week

Automate your own most repetitive coding task. Whatever you do by rote, hand it to an agent this week. Being the person who automated it beats being the person whose value was doing it by hand.
Practice writing specs. Take one feature and write the specification so cleanly that an agent could build it without you in the room. That skill is the new core of the job.
Make review your strength. Budget real time to read and test what agents produce, and get good at spotting where they are subtly wrong. That is where the work is heading.
Draw your no-fly zones. Decide which systems, especially money, data, and security, always keep a human in the loop, and hold that line.
Ship one thing end to end with agents. Own an outcome from idea to production using AI for the implementation. That is what the surviving role looks like.

Will AI replace software engineers? No, but it will replace the engineer who refused to change with the one who did. Learning to run your work, and your company, as one person plus a fleet of agents is exactly what we teach in AI Operating System for Startups.

Sources

Coding Will Be Solved For Everybody (Y Combinator), the conversation this article distills: coding becomes a shared tool, everyone on the team builds, and the "software engineer" title starts to shift.
On Boris Cherny's view that the title will fade and his agent-directed workflow: Business Insider, Fortune, and his talk Why Coding Is Solved, and What Comes Next (Sequoia Capital).
On more software getting built as it gets cheaper: a16z with Benedict Evans, "The Economics of AI Usage".
On rising demand for engineers to oversee agents: Meet The DevTool Founders Building For AI Agents (Y Combinator Root Access).
On taste and judgment staying essential: Andrej Karpathy, "From Vibe Coding to Agentic Engineering" (Sequoia Capital).