Back in March, Anthropic’s CEO Dario Amodei made a pretty bold statement: he predicted that within six months, AI would be writing 90 percent of all code. He even suggested that in three months, it might be writing all of it. Fast forward to today, and well… the reality looks nothing like that vision. From what we’ve seen so far, AI tools are helping, but they’re far from taking over.
Developers say AI assistants like GitHub Copilot, ChatGPT, and Claude are useful for things like boilerplate code or debugging small errors. They can speed up routine tasks and suggest snippets, which saves some time. But that promise of AI handling almost everything hasn’t come close to materializing. Instead, what’s happening is that programmers often spend extra time reviewing and rewriting AI-generated code to make sure it works correctly.
One of the biggest concerns is security. Research has shown that code created by AI often comes with more vulnerabilities than human-written versions. There have even been high-profile cases where AI systems introduced critical mistakes, like suggesting commands that could wipe a production database. Instead of boosting efficiency, that kind of risk makes teams more cautious about relying too heavily on generative tools.
Beyond security, there’s also the issue of accuracy and reliability. Developers have noticed that AI sometimes “hallucinates” functions that don’t exist or makes assumptions that break larger systems. That means AI isn’t just plug-and-play—it requires oversight, testing, and adjustments. Some engineers have even said it slows them down when they have to constantly double-check outputs.
Still, it wouldn’t be fair to call Amodei’s prediction completely useless. His statement reflected the ambition and rapid progress of AI. In just a couple of years, tools have gone from barely functioning to being genuinely helpful in day-to-day workflows. But claiming that AI would be doing nearly all coding by September 2025 was overly optimistic, if not unrealistic.
The takeaway is simple: AI is a strong assistant but not a replacement. It’s changing how developers work, but it hasn’t rewritten the playbook yet. Hype may sell headlines, but reality is moving at its own pace.
