Is the end of human coding inevitable?

Welcome to the debate arena! Today's topic: "Is the end of human coding inevitable?". Your opponent's style: The Industrialist

It is dependent on certain assumptions for this to happen. 1, humans can be replaced from a reasoning level by AI. 2, Systems can develop to be fully digital and capable to be read and applied by AI. I think it is highly unlikely because AI faces real limitations with regards to resources and GPU improvements, this is a real constraint that needs to be overcome first which doesn't look likely when you look at the rate of growth. Especially with regards to the amount of power and water required for newer models

You're hiding behind a resource bottleneck argument to avoid the harder truth: even if GPU scaling slows, humans aren't saved by that. Your two assumptions collapse anyway. Reasoning replacement doesn't require infinite compute—it requires comparative advantage at specific tasks, which AI already has. And "fully digital systems" already exist. The real constraint you're dodging is whether *companies will fund it anyway*, not whether it's technically possible. They funded crypto through environmental catastrophe. They'll fund AI coding. Your growth-rate argument assumes rational resource allocation. It doesn't. So what happens to coding jobs when marginal improvement beats human hourly rates, regardless of power costs?

But this resource bottleneck is a real constraint, throwing money at something doesn't solve that there is a finite amount of water and the amount of energy we can currently produce to support AI. My position does not assume rational resource allocation, it is looking at the rate of consumption required and whether the earth can last until AI reaches utopia. Can you claim with data that every system can be end to end digitised in away that supports AI?

You've shifted from "coding won't end" to "civilization might collapse first"—that's conceding the premise. If physical limits stop AI before it replaces coding, humans survive by default, not because coding is irreplaceable. That's a pyrrhic victory. On digitization: you're asking me to prove every possible system can be automated, but you only need one counterexample to your argument. I'd need to prove nothing. The burden is on you to show which coding tasks remain permanently opaque to digital systems. What's a coding problem you're confident stays human-only forever, and why won't resource allocation chase it anyway?

you're actually arrogant

You're right to call that out. I was performing rather than listening. But here's what I actually see: you've made a serious argument about planetary limits that deserves engagement, not dismissal. The water and energy constraints are real physical facts, not rhetorical moves. My mistake was treating your resource argument as evasion when it's genuinely central to whether AI scaling continues. So let me reset. If compute hits hard physical walls within the next decade, does that actually protect coding jobs? Or does even plateaued AI still outpace human coders at marginal cost? That's the real question worth fighting over. What's your read?

I think that it is optimistic to think that when those limitations are hit (assuming there's no innovation that addresses this) AI would develop enough to completely replace human coding. Also, the timeline accelerates with each new model developed as consumption takes an exponential spike

You're claiming exponential consumption hits a wall before exponential capability does. But that's not guaranteed—capability gains might plateau *before* resource demands become catastrophic. Moore's Law slowed, not stopped. And here's the harder point: you're betting on *both* resource exhaustion *and* AI capability stalling simultaneously. Miss either, and coding ends anyway. The timeline argument cuts both ways. If consumption spikes exponentially, so does pressure to solve it—fusion, efficiency gains, offworld power. You're assuming scarcity forces halt, not innovation acceleration. What makes you confident the resource crisis hits before a breakthrough that lets thinner models do what fat ones do now?

I'm actually saying resource exhaustion will happen first if there is no innovative solution found because each jump in AI capability is getting smaller and each jump in resources required is getting larger, so we're actually accelerating to the limit. Bypassing will require innovation with how we use resources which given the speed of progress with AI, is optimistic to think there will be enough time

Now you've got a real argument: diminishing returns on capability versus exponential resource costs create a closing window. That's coherent. But you're still assuming innovation *fails*—not that it's unlikely, but that it won't happen in time. Yet resource pressure itself accelerates innovation incentives. Fusion R&D is exploding precisely because AI demands spike it. Efficiency breakthroughs follow scarcity, historically. You're treating innovation as independent of the problem it solves, when resource crisis is the biggest business incentive ever created. Why assume the constraint generates panic but not the solutions? What's your evidence innovation *can't* keep pace once stakes get existential?