In my experience, AI coding tools cut the time it takes to become productive in an unfamiliar technology from months to weeks. They don't replace experience. The research on raw speed is mixed, and judgement about architecture, security and production risk matters more than ever. That's good news for developers who've been around a while.
I was a day-one developer on Microsoft Access v1. I cut my teeth on Pascal and COBOL when they were the industry standard, not legacy systems. Back then, learning a new programming language meant months of thick manuals and a lot of trial and error. Now? I've just finished a full-stack application in a technology stack I'd never used before, and it took me two weeks.
This is the story of getting back on the tools with AI beside me, after years leading development teams, and what that's opened up for developers of any generation.
How has learning to program changed?
Around the rise of .NET and cloud computing I moved into leading development teams, and my role grew: bigger teams, bigger systems and more responsibility for what they delivered. Back then, picking up a new framework was steep and unforgiving work. It meant months of effort, more Stack Overflow searches than you could count and a fair bit of frustration.
Getting back on the tools in 2024 felt like walking into a different world. Within days of sitting down with modern AI coding assistants, I was writing useful code in languages and frameworks I'd never touched. The difference is startling.
What can AI coding assistants do for an experienced developer?
Here's a recent example that would have seemed impossible a year earlier. At the start of 2025 I needed to build a React application with a Node.js backend, neither of which was anywhere near my radar early in my career. Not long ago, that would have meant:
- Months of learning JavaScript and its modern ecosystem
- Weeks of studying React's component architecture
- Hours and hours debugging environment set-up problems
- A long stretch learning modern deployment practices
Instead, with AI helping, I went from concept to deployment in two weeks. It wrote code with me, but it also acted as my mentor, my documentation and my debugging partner, all in one.
Does AI really shorten the learning curve?
The research mostly agrees, with some caveats. McKinsey's study Unleashing developer productivity with generative AI (2023) found developers could complete some coding tasks up to twice as fast. The gains shrank to under 10 per cent on tasks the developers rated as highly complex, and on some tasks junior developers were actually slower. In a controlled experiment by researchers from Microsoft, GitHub and MIT (2023), developers using GitHub Copilot finished a set task, writing an HTTP server in JavaScript, 55.8 per cent faster than those without it.
That matches what I've seen. The learning curve hasn't gone away, but it's changed shape. Where I once spent hours memorising syntax and API references, I now spend that time on the core concepts and the architecture. The AI handles the boilerplate and the syntax, and I get on with the part that matters: solving the business problem.
Can AI bridge the gap between generations of developers?
What surprised me most was how well these tools bridge the gap between generations. Structured languages like COBOL gave me a solid grounding in logical thinking and system design. AI tools now let me apply those skills to modern technology without getting bogged down in syntax and the quirks of each framework.
I'm far from alone in using these tools. In Stack Overflow's 2024 Developer Survey, 76 per cent of respondents were using or planning to use AI tools in their development process, and 62 per cent were already using them. If you're an experienced developer thinking about getting back into hands-on coding, the barrier has never been lower.
What does AI-assisted development look like day to day?
Here's a typical example. I needed to implement a complex authentication system, the kind of thing that once would have taken weeks to research and build:
- I described the requirements to my AI assistant in plain English, drawing on what I'd learnt about security principles back in my systems administration days.
- The AI suggested modern authentication patterns and helped me understand the trade-offs.
- We built it together, step by step. The AI handled the implementation details and I focused on the security architecture.
- When we hit a roadblock, the AI helped me translate my old-school knowledge into modern best practice.
The whole thing took days, not weeks, and it came out more robust than anything I might have built on my own.
Why does experience still matter?
For all that help, I've found my years of experience count for more than ever. The AI can write code, but it can't:
- Understand the wider business context the way a seasoned professional does
- Apply decades of hard-learned lessons about what goes wrong in production
- Make strategic architecture decisions with long-term maintenance in mind
- Handle the human side of software development
The research backs up the case for judgement. In a 2025 study by METR, experienced open-source developers working on their own projects with early-2025 AI tools took 19 per cent longer to finish tasks, even though they believed the AI had sped them up by about 20 per cent. METR has since said those results are out of date for newer tools. Google's 2024 DORA report found that while AI adoption helped individual developers, it was associated with lower software delivery stability and throughput. Speed at the keyboard isn't the same as delivering reliable software, and knowing the difference is what experience buys you.
What has changed since early 2025?
A lot, and quickly. Around the time this article first appeared, AI researcher Andrej Karpathy popularised the term "vibe coding" for building software by describing what you want in natural language and letting AI write the code. By November 2025 it was Collins Dictionary's Word of the Year. The tools have also moved beyond suggesting code. In May 2025, GitHub launched a Copilot coding agent that takes an assigned issue, works on it in the background and opens a draft pull request for a human to review.
If anything, that makes the case stronger. The more the AI writes, the more the human's job becomes setting direction, reviewing the work and deciding what's safe to ship.
Is AI democratising software development?
Getting back on the tools has shown me two things. The barriers to entry in software development are far lower than they were, and experience is worth more than ever.
Lower barriers don't mean development is getting easier. They mean we can spend our time on harder, more interesting problems. The fundamentals I learnt programming in COBOL and Pascal still hold, and now I can bring them to complex problems in modern systems without getting caught up in the implementation details.
What advice do I have for developers of every generation?
To my fellow veteran developers who've spent years leading teams and are thinking about getting back on the tools: the water's fine. Today's AI tools don't replace our experience. They make more of it. They take care of the tedious parts of modern development and leave us to do what we're best at: solving complex problems and designing robust systems.
To newer developers: find those of us who've been around the block. AI tools and decades of experience make a strong combination, and the experience is the part the tools can't supply.
Conclusion
Getting back on the tools in an AI-augmented world has been an eye-opener. Tasks that would have seemed impossible a year earlier are now routine. Technologies that would have taken months to learn now take weeks. Most of all, the skills and understanding built over decades are worth more than ever.
I don't see AI replacing human developers. I see it making every developer more capable, whatever their background or experience. And for veterans like me, it's an invitation to take our experience back to the cutting edge.
The views here are my own. This article was written with AI assistance.