What is the current situation for older programmers, and what paths are open to them?

I think opportunities for older programmers are gradually starting to open up. In my view, programmers who are still employed this year and next year are unlikely to lose their jobs.

This view may be at odds with the prevailing narrative that AI will replace programmers. AI is clearly reducing the number of programming jobs, so how could older programmers still have opportunities?

In the past, companies didn’t usually reject older programmers because they couldn’t code. The main reasons were that younger people could stay up late, work overtime for long stretches, pick up new frameworks quickly, and accept lower pay. If a 35-year-old was still doing APIs, pages, and basic CRUD work, a company would surely choose someone cheaper and more willing to push themselves to do the same job.

But AI is now dismantling that way of comparing people.

Given the same task, a 25-year-old programmer can ask an agent to write code, and so can a 40-year-old. Completion speed, typing speed, and the number of APIs someone can memorize will soon become nearly equal. In the past, younger people could stretch the workday to twelve hours through sheer energy, while older programmers had family responsibilities in the evening. Now, much of the mechanical work can run in the background with an agent. People can spend most of their time clarifying the problem, checking the results, and fixing issues.

The two areas where older programmers are most disadvantaged—stamina and working hours—have been greatly weakened by AI. At the same time, AI amplifies the parts of their experience that are truly valuable.

Someone who has spent ten years working on payment systems sees the word “retry” and thinks about duplicate charges, idempotency keys, channel timeouts, and reconciliation. A newcomer sees a request to add a loop to an API. Someone who has maintained a large system knows why a field that looks redundant can’t be removed, which jobs must not run concurrently during month-end settlement, and which business stakeholder to consult before a release. You can’t learn these things from framework documentation.

An agent can generate countless possible solutions, but it doesn’t know which one will hit a landmine left behind three years ago. Experienced people don’t have to design every option themselves. If they can quickly identify the right one, their value becomes clear.

My view is that a large number of programmers will be pushed out over the next two years.

AI is shrinking a particular kind of work: the requirements are already clear, you write code to fit an existing architecture, and a senior colleague steps in when something goes wrong. These are the kinds of roles that grew out of companies scaling up their teams. Many companies will find that three senior engineers who understand the business, working with a group of agents, can produce more than a team of a dozen people did before.

Of course, I’m not saying that age alone will keep you in a programming career. People who don’t know how to use AI, can only maintain an old framework, or take “I have experience” to mean “I’ve done it this way before” will be pushed out even faster than younger programmers. If you’re expensive, slow to learn, and have no advantage in stamina, companies have no reason to keep you.

The people who will last are those who are willing to let AI write code and can still understand it; who don’t insist on typing every line themselves but know what they must check personally; and who can turn the pitfalls they encountered a decade ago into tests, rules, and review checklists so an agent won’t fall into them again. Only these people will see their careers grow longer.

In the past, older programmers often felt forced to move into management to prove themselves. No matter how deep their technical expertise, they worried that companies only cared about how many hours they worked. With AI, a senior engineer can use tools to deliver projects that once required a group of people. They don’t necessarily have to manage dozens of people to take on greater business responsibility. There will be more independent developers, small teams, remote consultants, and technical leads in specialized industries than before.

Another thing will gradually become clear: there will be a gap in how new programmers are trained.

Once AI handles all the simple tasks, newcomers will lose the stepping stones they need to practice. They’ll be able to generate polished code very early in their careers, but they won’t have experienced an overloaded database, a cache stampede, a rollback of corrupted data, or an all-night investigation of a production incident. A few years from now, the market won’t lack people who know how to use agents. It will lack people who have seen how systems fail and know when not to trust an agent.

For older programmers who have already paid their dues, the scars they carry will become a scarce asset.

So, older programmers shouldn’t compete with younger people over who can type faster, and they shouldn’t focus only on moving into management. They should turn their experience into things AI can use as soon as possible: project rules, test cases, incident manuals, business constraints, and code review standards. You make the calls; let the agent do the legwork.

AI will first make many jobs disappear, then widen the gap in value between those who remain. As long as we can make it through the culling over the next two years, I believe things will get better for all of us.