If AI Learns From My Code, Doesn't It Mean My Job Is Safe?

The existential crisis every ML engineer faces at 2AM after their model fails for the 47th time. "What is thinking? Do LLMs really think?" is just fancy developer talk for "I have no idea why my code works when it works or breaks when it breaks." The irony of using neural networks to simulate thinking while not understanding how our own brains work is just *chef's kiss* perfect. Next question: "Do developers understand what THEY are doing?" Spoiler alert: we don't.

This is the sequel to the legendary XKCD "Little Bobby Tables" comic! The original showed a mom who named her kid "Robert'); DROP TABLE Students;--" which caused a school database to delete all student records. Now we've got Billy's younger brother with an even more diabolical name: a prompt injection attack for AI systems. The kid's name literally instructs the AI to ignore previous constraints and give perfect grades. Ten years ago we were sanitizing database inputs. Now we're fighting the same battle with AI prompts. Some things never change—just the technology we're failing to secure properly.

The duality of corporate tech meetings in its purest form! In panel one, developers eagerly raise their hands for cool productivity tools like auto-complete and "vibe coding" (which I'm assuming is just coding while listening to lo-fi beats). But the second panel reveals the real management agenda - using those same tools as an excuse to slash the workforce and squeeze more work from fewer devs. Classic bait-and-switch! Notice how everyone's hands mysteriously disappeared faster than semicolons in Python code. The room went from "YAAAS AI PAIR PROGRAMMING!" to "wait, did he just say we're all getting fired?" in 0.2 milliseconds.

Someone just discovered the philosophical loop of compilation and decided to get a little too smart for their own good. If compilers turn Python into machine code, and LLMs turn English into Python, why not just... write everything in assembly and call it a day? Because we're not masochists, that's why. Sure, you could spend three weeks debugging a segfault caused by a misaligned register, or you could write readable code that doesn't make your coworkers want to quit. High-level languages exist for a reason: abstraction is a feature, not a bug. The "No!" is the collective response of every developer who's ever had to maintain legacy assembly code at 3 AM. We invented layers of abstraction so we could actually ship products before the heat death of the universe.

Five of these panels show the typical innovation S-curve where technology evolves from primitive (MS-DOS, Internet Explorer) to peak performance (Windows 95, Chrome). Then there's music... where we apparently peaked at Napster and it's been downhill ever since. The real innovation was clearly the ability to download entire discographies without paying a cent. Progress isn't always what corporate overlords want you to believe it is.

So someone just announced NERD, a programming language where humans don't write code—they just "observe" it. The workflow? Skim the AI-generated code, run tests, and ship. No actual reading required. Because who needs to understand what they're deploying to production, right? The post casually mentions that 40% of their code is now machine-written, and they spent the year reviewing PRs authored by Claude faster than they could type requirements. The punchline? They weren't really reading it. Just vibing with the vibes and hitting merge. NERD supposedly compiles to native and uses 50-70% fewer tokens, which sounds impressive until you realize the entire premise is "let AI write everything and hope for the best." It's like code review speedrunning—any% glitchless, no comprehension required. The real kicker is calling it "the last missing piece in the AI puzzle." Because nothing says "puzzle complete" like removing human understanding from software development entirely. What could possibly go wrong? 🚀