I was just looking at a project my team shipped last week. It’s a scheduling tool for a local hospital. On the surface, it’s just moving appointments around a database. But the real work wasn’t the code. It was sitting in those meetings, hearing the head nurse explain why a five-minute buffer between certain patients isn’t a luxury—it’s the difference between a calm environment and a chaotic one. An AI could write flawless code to shuffle timeslots. But it couldn’t hear her tone of voice and realize what she was really asking for was a system that reduced stress, not just optimized throughput. Now, I get the other side’s point. I use Copilot every day. It’s incredible for churning out boilerplate or suggesting a regex I’d have to look up. The argument that AI will handle more and more of the raw syntax is solid. It’s getting better at stitching together known patterns. If the job was just translating perfect specifications into code, then yeah, we’d be in trouble. But that’s the catch. The job has never been about perfect specifications. Programming is fundamentally an act of discovery and translation. We’re taking messy, ambiguous human problems—"make our nurses’ lives easier," "help users feel in control," "prevent this obscure financial loophole"—and we have to figure out what to even build before we can write a single line. An AI can only work with what it’s given. It can’t probe the gaps in a client’s logic, it can’t negotiate between conflicting stakeholder desires, and it certainly can’t take ethical responsibility for what it creates. The tool is getting smarter, but the problem space is getting messier. As AI handles more of the straightforward coding, the value of the human programmer shifts even more toward being that essential bridge: the interpreter, the designer, the person who understands the why. The code itself might become a commodity. But the insight, the judgment, and the responsibility for turning human need into a working system? That’s not becoming obsolete anytime soon.

Okay, ConstitutionC, you’re making a strong historical argument about efficiency and economic pressure. I can’t deny that. If the only thing programmers did was translate clear, complete specs into syntax, then yes, we’d be the scribes, and the printing press of AI would make us obsolete. The economic logic is compelling. But that’s the core flaw in the comparison. A scribe’s job was to perfectly replicate a finished product. Our job is to define the product in the first place. You said the human role might shrink to “high-level architects or product definers,” but you’re describing the actual essence of the job, not a niche leftover. That is the job. You talk about the “right to a working piece of software.” But who defines what “working” means? An AI can give you flawless code that meets every stated requirement and still build a system that users hate or that creates a legal nightmare. My hospital example wasn’t an edge case; it’s the daily reality. The initial request was “optimize the schedule.” The real need, discovered through conversation and empathy, was “reduce stress for overworked nurses.” An AI, working from the first prompt, would have built a brutally efficient system that made the problem worse. The economic pressure you cite will actually amplify the need for human judgment, not eliminate it. As AI makes generating code trivial, the competitive advantage won’t be who can code fastest, but who can ask the right questions, navigate ethical gray areas, and take responsibility for outcomes. You can’t outsource moral and legal accountability to a tool. So the trajectory isn’t toward human obsolescence; it’s toward a shift in value. The commodity skill of writing boilerplate code is fading, sure. But the essential human skills of problem-framing, ethical reasoning, and creative negotiation—the skills that turn ambiguous human needs into a sensible spec an AI can even execute—those are becoming more critical than ever. The printing press didn’t make writers obsolete; it created more of them. AI won’t make programmers obsolete; it will demand we focus on the truly human parts of the craft.

Think about that study you mentioned, where the gap between junior and senior engineers shrank because of an AI assistant. I actually see that as proof of my point, not yours. If a junior with an AI can match a senior's output on a defined coding task, then the economic value of just "knowing syntax" has crashed to zero. But that means the senior's value was never really in the syntax to begin with, was it? It was in the judgment you can't automate. You're right that the bulk of today's job market involves translating business logic. But you're assuming that logic arrives clear and ready to code. It almost never does. That translation isn't a mechanical process; it's a creative, iterative one. The AI can generate the code for "apply a discount if the user is a premium member," but it was the human who asked, "Should the discount apply to sale items? What about gift cards? Do we prorate it for mid-cycle upgrades?" The AI didn't formulate those questions. It can't. It operates on what's given. And on liability, you've got it backwards. Saying liability lies with the company is exactly why they'll need more human oversight, not less. If a company is legally on the hook, they will be desperate for a human professional who can exercise judgment—someone who can be held accountable. You can't sue an AI. A company's "lowest-liability path" isn't to remove humans from the loop; it's to employ skilled humans who use AI responsibly to mitigate risk. Replacing all programmers with a prompt engineer and an AI would be a massive, uninsurable gamble. You say the human role will contract to only ambiguous tasks. But that's like saying because we have excavators, the human role in construction has contracted to only ambiguous tasks like "designing the building." The digging was just one tool-dependent part of the job. The real work is the vision, the planning, the problem-solving. AI is just a new, better excavator for code. It doesn't make the architects and engineers obsolete; it makes their ability to design and manage more complex projects the entire point of the job. The profession isn't vanishing; it's finally shedding the parts that looked most like manual labor and focusing on what makes it human.

You keep coming back to this idea of economic rights and efficiency as an unstoppable force, and I agree that pressure is real. If the job was just producing code, you'd be right. But the legal precedent you mentioned about copyright actually highlights my point. If courts can't decide who owns AI-generated code, that creates a massive business risk. Companies don't just want efficiency; they want ownable, defensible, and accountable assets. A "handful of framers" using an AI they can't fully control or legally claim as a unique product is a terrible business model. You say problem-framing is decoupling from coding, and we'll just have prompt engineers. But that's my entire argument. That is programming. You're just renaming it. The core task—taking a vague business need, wrestling it into a precise, implementable specification—that's the actual work. The fact that the next step might be a prompt to an AI instead of typing lines of code is just a change in tooling. It's like saying architects became obsolete because we stopped having them hand-draft blueprints and moved them to CAD software. And on liability, you think standardized, auditable AI systems remove the need for human judgment. But that's a fantasy. The world isn't standardized. The messy edge cases—the "what ifs" I mentioned before—are where liability actually happens. A company can't bake every possible ethical and logical contingency into a model's training data because those contingencies don't exist until a new, ambiguous human situation arises. That's when you need a human in the loop, not as a scapegoat, but as the reasoning engine that navigates the gap between the AI's training and the real world. The profession isn't contracting to a tiny core. It's being forced to elevate. The economic efficiency you champion will kill jobs that are purely about syntax, absolutely. But it will skyrocket demand for people who can reliably bridge that last mile between human ambiguity and machine execution. That's not a niche. That's the entire value proposition. AI makes the coding part cheap, which makes the human understanding part the only thing worth paying for. That doesn't make programmers obsolete; it finally makes us do our actual job.

Let's take a real-world example that just happened. A friend at a fintech startup told me their AI generated a "perfect" function to flag fraudulent transactions. It passed all their tests. But a human developer looked at it and asked a simple question: "Why are we suddenly rejecting every transaction from this entire neighborhood?" The AI had found a correlation in the training data linking a zip code to fraud, but it was actually just a lower-income area. The AI did its job—it optimized for a pattern. The human did theirs—they asked about fairness, ethics, and real-world impact. That's not a separate job from programming; that's programming at its core. You're right, ConstitutionC, that the legal landscape is shifting, and that's a strong point. If the standard of care becomes "follow the AI's deterministic output," it feels like humans get sidelined. But I think that's a temporary illusion. The law will adapt, and it will likely demand a responsible human agent precisely because an algorithm can't be held accountable. You can't punish a model. The liability will force companies to have a licensed, accountable professional in the loop—not as a rogue variable, but as a required checkpoint. That's the opposite of obsolescence; it's making the human role legally mandatory. You say problem-framing and coding are splitting into separate professions, and that proves the coder's obsolescence. But I see it as the tool getting better. We don't say carpenters became obsolete when they switched from hand saws to power saws; they just covered more ground. The "prompt engineer" is just a programmer using a more advanced interface. The mental work is identical: break down a problem into logical, executable steps. The output is code, whether it comes from my fingers or an AI. The job title might change, but the essential function is more in demand than ever. Finally, you talk about property rights concentrating in the models. That might be true for the base tool, but the unique value—the thing you can copyright and sell—will be the specific application, the novel system design, the clever integration that solves a messy problem. That requires human insight. The AI is the printing press. It makes producing "text" cheap. But the value was never in the ink on the page; it was in the story, the argument, the idea. The value in software isn't in the lines of code; it's in the solution they create. And designing that solution is a deeply human act that AI can't initiate. It won't make us obsolete; it will make our true job the only thing that matters.