No, software developers are not doomed, argues Quentin Merle in his first-person DEV Community manifesto on AI. His case is not that AI coding tools are harmless or that jobs are guaranteed: it is that generating code is not the same as designing, checking, and maintaining dependable software. The piece is an opinion, not a labor-market study or a security audit, and its publication year is not confirmed.
What Merle means by saying developers are not doomed
Merle’s central argument is that AI lowers the barrier to producing code without removing the engineering work around it. A model can help draft or explain code, but a developer still needs to understand how that code fits into a larger system, whether it behaves correctly, and how it will be maintained.
As an Amazon Associate I earn from qualifying purchases.
He treats AI as another abstraction layer: it can reduce routine syntax work and support learning, but it can also make it easier to accept output without understanding it. In his view, the risk is not simply that a model writes code; it is that someone treats generated code as finished software without independent checks.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
To explain why complexity persists, Merle compares current claims about AI replacing developers with earlier claims that content management systems would make developers obsolete. His point is that difficulty often emerges when software must integrate with existing systems, operate under production conditions, and remain maintainable. That is an analogy, not evidence that the two periods or technologies have had the same effects.
#1 Best Overall
Why generated code still needs engineering checks
Merle recommends checking AI-generated work with tools that can assess it independently of the model. His examples are a compiler, a linter, and a separate test suite. These checks do not make software automatically safe or correct, but they can reveal issues that a model’s own review may miss.
- Compile or build the code: Confirm that it meets the language and project’s build requirements.
- Run a linter: Catch violations of the project’s rules and common code-quality issues.
- Run independent tests: Check expected behavior rather than relying on the model’s assurance that its output is correct.
- Review the result yourself: Make sure you can explain what the code does and how it interacts with the surrounding system.
Merle frames this as a shift in engineering value: less emphasis on producing a clever prompt, more on designing a deterministic process that can constrain and verify unpredictable output. That is his prescription, not a demonstrated guarantee that developers who follow it will avoid job loss.
When Merle suggests hosted models and when he suggests local ones
Merle’s proposed distinction is based on data sensitivity. He suggests hosted models for work such as brainstorming, drafting, debugging, using public documentation, and rapid prototyping. For personally identifiable information, production logs containing IDs, or confidential internal scripts, he proposes local or air-gapped models instead.
Free tools Windows power users keep installed
One-click scans. No signup required.
| Choice | Uses Merle describes | Tradeoffs and limits in his account |
|---|---|---|
| Hosted model | Brainstorming, drafting and debugging, public documentation, rapid prototyping | Merle emphasizes capability, speed, and large context. Data handling depends on the service and arrangement; his article does not verify any provider’s retention or security practices. |
| Local model | Personally identifiable information, production logs with IDs, confidential internal scripts | Offers more control over where processing happens in Merle’s proposed workflow, but he notes hardware demands and performance constraints. The article does not assess a particular setup’s security. |
This is a suggested workflow, not proof that a hosted service is safe for a particular dataset or that running a model locally makes sensitive data secure. Teams still need to assess their own data rules, configuration, and operational safeguards.
How junior developers can use AI without skipping the learning
Merle argues that junior developers can use AI as a learning aid if they engage with what it produces. Asking for explanations and examining the resulting code can help a learner work through unfamiliar ideas. Copying code without understanding it, by contrast, can leave the learner unable to debug, modify, or defend the result.
- Ask for an explanation of unfamiliar code, then check that explanation against the code and project documentation.
- Trace how the generated change affects the rest of the application.
- Try to predict what the tests should do before running them, then investigate any mismatch.
- Make sure you can explain the code you keep, rather than treating a plausible answer as proof of understanding.
These are learning recommendations, not measured findings about hiring or career outcomes. Merle’s statement that junior developers are not an endangered species expresses his view; the article does not establish how AI will affect employment.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What the manifesto does—and does not—establish
Merle explicitly presents the piece as a reflection at a point in time, not a settled forecast. Its conclusions about developer work, AI’s capabilities, and the future of the industry should therefore be read as arguments by the author rather than verified predictions.
The article does not provide comparative industry data for its content-management-system analogy, controlled benchmarks of local versus hosted models, or independent evidence for its claims about employment and security. It also makes an unsupported assertion that 95% of startups, small and medium-sized businesses, agencies, and freelancers lack enterprise contracts or the skills to audit data streams; without a named source, date, or dataset, that number should not be treated as verified. Merle’s comparison of “three years” with “ten years” is likewise a prediction, not a measured result.
Best Value
The useful takeaway is narrower than a guarantee about the future: Merle believes AI can make coding more accessible, while engineering judgment remains important for checking output, handling sensitive information, and learning from generated code.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
