Code IS dead
Code IS dead
Like, dead-dead.
I remember sitting down to debate this exact issue with my friend and master Nabil Belakbir, CTO at Nemo Technology, three years ago. At the time, we were playing around with a Visual Studio plugin called “AI Code” or something like that. It hooked 2022-era ChatGPT straight into the editor, saving us from the relentless copy-paste loop. That was the first real wake-up slap to my pride as a software craftsman.
I tried to resist as long as I could, arming myself with all the usual talking points:
- Code still has to be reviewed by a human
- Good design patterns are too hard to prompt into an AI
- AI will never grasp the full context of the codebase, adjacent systems, and actual business needs
- Software architecture requires a high-level, big-picture vision unique to human wisdom (lol)
- You can’t hand production over to an AI
- Critical paths will always need a human pass for optimization (concurrency, performance, databases, latency, and especially security)
- Letting an AI craft a public API is straight-up suicide
Then, one by one, those bulletproof convictions crumbled. First in my own head (if LLM benchmark curves keep climbing like this, we’re cooked), then across social media (tech giants quietly deploying autonomous AI at scale), and finally among tech influencers who suddenly couldn’t talk about anything else—including major industry figures who had preview access to LLM builds 6 to 12 months ahead of the rest of us.
I even threw myself into the trenches of LinkedIn posts and comment sections, arguing against AI moderates with their cozy, sanitized rhetoric. AI will just be a co-pilot for humans, AI is never going to replace coders, AI can’t write usable code, blah blah blah…
Deep down, I was desperately hoping they were right. Truly.
Then a contact in the US tech scene whom I deeply respect kept dropping the same line every two or three months: “Issam, get ready. You haven’t seen anything yet…” The writing was on the wall—far removed from wishful thinking and naive optimism.
Today, consensus is finally forming that gpt6-Astra is true AGI. I personally see AGI as a slow creep—a gradual realization among AI thinkers rather than a single model descending from the clouds. But this time, folks, we’re actually here.
I wrote my final lines of hand-crafted code back in October 2025, and I meticulously reviewed my last PR a few weeks ago. Sure, I’ll probably keep scanning sensitive components—security hooks, public APIs, the high-stakes stuff—but raw code has officially become modern-day assembly language. The developer has morphed into a supervisory artist, guiding the AI toward real business value, product experience, and actual utility.
Code is well and truly dead. The traditional software engineer job is done, finished, terminado, baraka.
Long live the age of AI!