Elon Musk had previously said that code will become obsolete with AI directly generating binaries, and Marc Andreessen seems to agree.
The a16z co-founder and veteran technologist — the man who built Mosaic, the first mainstream web browser — recently raised a question that cuts to the heart of what software development even means: will programming languages exist in a decade? Speaking on a podcast, Andreessen pushed the conversation well past the familiar debate about AI replacing coders, and into stranger, more radical territory.
“Are you going to even have programming languages in the future? Or is the AI just going to be emitting binaries? Let’s assume for a moment that humans aren’t coding anymore. Let’s assume it’s all bots. What levels of intermediate abstraction do the bots even need? Or are they just coding binary directly?”
He then took it a step further. Andreessen described an experiment in which a language model was used to emit weights for a new language model — AI generating AI, skipping code entirely. “Will the bots literally be emitting not just coding binaries, but will they actually be emitting weights for new models — directly? Conceptually, there’s no reason why they can’t do both of those things. Architecturally, both of those things seem completely possible.”
When the interviewer pointed out that this amounts to a simulation inside a simulation — deeply inefficient — Andreessen agreed, but argued inefficiency is beside the point. “LLMs are already incredibly inefficient. Ask Claude to add two plus two equals four — it’s billions and billions of times more inefficient than using your pocket calculator. But the payoff is so great of the general capability.”
That trade-off — extreme inefficiency for extreme generality — is what leads him to his broader conclusion: “I kind of think in 10 years, I’m not sure there will even be a salient concept of a programming language in the way that we understand it today. And in fact, what we may be doing more and more is a form of interpretability — which is we’re trying to understand why the bots have decided to structure code in the way that they have.”
Andreessen extended the logic further still: if AI is doing the coding, and the primary users of software become other bots, then even interfaces become optional. “Who is going to use software in the future? Other bots.” At that point, browsers, UIs, and the human-readable layers of software become unnecessary overhead.
He was careful not to frame this as dystopia, drawing on a familiar historical arc: “It was not that long ago that 99% of humanity was behind a plow. And what are people going to do if they’re not plowing fields all day to grow food? It just turns out there are much better ways for people to spend time than plowing fields.” He even noted, with some self-awareness, that he has an 11-year-old learning to code — and still thinks that’s worthwhile. But he’s clear that projecting forward, the logic runs in one direction: “I’m just going to tell the thing what I need, and it’s going to do it in whatever way is most optimal for it to do it. And then if I need to understand how it works, I’m going to ask it to explain to me how it works.”
Andreessen’s remarks land in the middle of a broad consensus forming among tech leaders that programming, as historically practiced, has a limited shelf life. Musk has predicted AI will generate binaries directly, bypassing source code and compilers entirely. Anthropic CEO Dario Amodei has said 90% of code could be AI-written within months, and Microsoft CTO Kevin Scott has put the figure at 95% within five years. An Anthropic researcher has described the future of software development as less about writing code and more about reviewing it — an intermediate position that Andreessen’s framing leapfrogs entirely.
What Andreessen adds to this conversation is the logical endpoint. If AI agents are generating code, and AI agents are the primary consumers of software, then human-readable programming languages are not just unnecessary — they are an artifact of a world designed around human cognitive limitations. The question he’s really asking isn’t whether coders will lose their jobs. It’s whether “code,” as a concept, makes sense at all once humans are no longer in the loop. His answer, tentative but directional, is no.