From JavaScript Frameworks to AI Tools: The Same Debate, Different Wrapping (lassala.net)

The piece argues that the perennial tool wars in software (Visual Basic vs. Delphi; Angular vs. React) have simply migrated to AI-assisted developer tools, with fast-rising players like GitHub Copilot, Cursor, Windsurf and Claude Code. A recent LinkedIn claim that “Claude Code is 10x better” than Cursor prompted a closer look: suggested practices (create a project goals file like CLAUDE.MD, ask the tool to explore before coding, use /clear to reduce hallucinations, automate repetitive tasks with custom commands, paste screenshots for richer context) are sensible but not novel. Many of these capabilities — e.g., context auto-summarization, screenshot inputs, custom workflows — already exist in other tools, or are user practices that predate current UIs. The core critique is that tool superiority claims often lack context: better at what, for whom, and by what metric? For the AI/ML community the takeaway is pragmatic: focus on enduring software practices (BDD, TDD where appropriate, pair programming and specification-driven development) rather than hype about individual tools. Technical implications include prioritizing workflow fit, reproducible benchmarks, and explicit evaluation criteria (accuracy, latency, hallucination rates, developer velocity) when choosing tools. Pairing with an AI requires skillful collaboration, not just issuing commands; tools enable workflows, but practices and measurement determine long-term effectiveness.