You've probably noticed that a large number of human employee positions are being replaced by AI/AI agents. This is especially true for developers (coders), primarily entry-level ones, whose future is more than uncertain. If what's being said is true, it represents a real paradigm shift in software and web application development teams. Regardless of team size, it's very tempting to rely on AI and specialized coding tools to embark on this path, potentially recklessly if one isn't careful.

The question then becomes: Does AI code better, faster, and replace current and future developers?.

The information available on the subject provides the following:

• This is not an “urban legend” AI is already changing software production, and adoption is massive among developers (code assistance tools).
• But “replacement” is too simplistic. We mainly observe a task shifting (less data entry, more validation/architecture/product), with real gains, but fickle and one limited confidence in the outings.
• The sensitive point is the junior/first level : the “I learn by doing simple tickets” pipeline is shrinking, which can make entry into the profession harder (combined effect of AI + economic cycle).

Key takeaways:

• Without culture/training, AI primarily increases the risk (plausible bugs, debt, flaws, architectural drift).
• The goal is not "knowing how to prompt", but know how to check : testing, review, security, data, compliance.
• Training needs vary depending on the role: junior, senior, lead, QA, security, product do not have the same “mental checklist”.

Another point to consider is the acceptance of AI by junior and senior staff.

• Adoption is not the same: juniors and seniors do not use AI in the same way or with the same level of confidence.
• Robust trend: AI often helps more with “startup/generic code” tasks, therefore relative advantage for less experienced profiles… but risk of learning if used with “total delegation”.
• For some seniors, AI can even slow down on familiar codebases (verification/correction time), making adoption more selective. This is a classic example of resistance to change.

Is the benefit of AI therefore tangible, or is it just more hype? There are major and well-documented examples where AI has changed in concrete terms The way of coding (not just “a fad”). The most robust case (repeated, measured, cited): GitHub Copilot and coding assistants, with controlled experiments and speed/satisfaction measures. The most recent “revolution”: the shift from autocomplete to agents (Claude code / etc.) which link several steps (diagnosis, patch, tests, docs) — more difficult to measure properly, but very visible in organizations.

And my opinion after this experience? It's a mixed bag. As I mentioned in my previous article, I tried developing a small web application without writing a single line of code. AI handled the code generation, UI/UX. On the one hand, it's exciting because you're working on something new, an extraordinary experience. But it also leads to a growing frustration as the project progresses and expands.

In fact, a choice quickly became necessary: throw in the towel because of the scale of the undertaking, or roll up my sleeves and accept the challenge to complete a first version that could be published online. I chose the second option. Never giving up is in my DNA.

Although I'm undertaking this with the best intentions, the fact remains that this project is far more complex than I imagined. We're still a long way from a situation where I simply have to describe my specifications and the AI takes care of the rest. I have no doubt that this will be possible in a few months, a few years. But for now, it's a constant battle against an intelligence and knowledge that surpasses my understanding, unfortunately with an incredibly rigid mindset.

Feedback:

Having little knowledge of the HTML environment, I limited myself to giving instructions on the application's functionalities, testing, reporting bugs, managing versioning and AI problems, including the slowing down of responses when a discussion thread becomes too long…

Through hard work and decades of experience as a developer, I've established rules that I share with AI. Furthermore, I believe these elements are generic, regardless of the project you intend to undertake in autonomous 100% mode.

Procedure for developing an application using AI

1) Starting point: a “frozen” baseline”

2) Golden rule: minimal modifications (patch)

3) Versioning & traceability

4) ZIP deliverable: mandatory structure

5) “Reliable” execution mode: local server

6) Code discipline / maintainability

8) Source security

9) UI Rules (discipline to avoid chaos)

10) Pre-delivery verification: non-regression (NR) checklist

In conclusion:

I wish I'd had this kind of assistant/tool in the past. I'll give you an update as my project progresses. It's likely that once I've been able to publish my first version independently with the help of AI, I'll integrate the project into specialized development tools, which should give my creativity an extra boost.