Quote of the Day
Elon Musk just told the highest-paid profession on earth it was never a skill. It was a workaround.
AI no longer needs programming languages at all. It writes machine code directly. Raw binary optimized beyond anything human logic could produce.
No syntax. No compiler. No translator.
Musk: “You don’t even bother doing coding.”
Every language ever written. Every framework ever built. Every career spent mastering them.
All of it was humanity’s way of approximating what machines actually needed. We didn’t learn to speak to computers. We spent fifty years building a crutch because we couldn’t.
AI just made the crutch irrelevant.
Now add Neuralink. No keyboard. No screen. No interface.
Musk: “Imagination-to-software.”
Intention becomes execution. Nothing between what you imagine and what exists.
Humanity spent its entire history building tools to shrink the gap between mind and reality. Language. Writing. Mathematics. Code.
Every one was just a smaller bottleneck. AI doesn’t shrink the bottleneck. It removes the gap.
The entire framework of technical expertise just collapsed into one question. Can you think clearly.
Everyone else is debating how AI changes development. Musk already moved past the debate. He’s not optimizing the process. He’s dissolving the need for one.
The barrier between imagination and reality held for all of human history. It just fell.
Dustin @r0ck3t23
Posted on X July 10, 2026
Just one question, “Can you think clearly?” <SNORT!!>
That is hilarious. There are surprising number of people that contradict themselves in a single written sentence. Just imagine what looking under that hood at their thought patterns would be like!
But let’s go with it for a bit. Imagine if there was no practical way for a human to review programs before they were deployed. That is what this proposal means.
Programs thought into being by 1,000 different madmen will attempt genocide of Christians. A different set of 100,000 madmen will attempt the complete annihilation of all people of Jewish heritage. And others of white, black, brown, and/or yellow skin color would be targeted for extinction. The existence of source code in the present-day world offers some hope of detection and remediation of the killer code.
And if there were no source code a rogue AI could infect every other AI with near zero ability for humans or benign AI to detect the situation before it was too late.
Imagine all the world’s industrial controllers, vehicles, medical equipment, food supply chains, electrical systems, water supplies, pharmaceuticals, and machines of war are secretly programmed to kill as many humans as possible starting on January 1, 2030, 00:00 UTC. With no way to know all the computers were about to flip a bit and change from obedient servants to apex predators we would be caught totally by surprise. Even if a critical mass of people did survive and eventually suppress the new super predator, there would be no way to quickly recover. Essentially every computer from the one in your watch to the computers in your appliances and even sewers could have a seed of the virus that given a semi-fertile place to grow could reconquer the world and be smarter about it the next time. The survivors would have to destroy or completely isolate every computer on the planet and start up our technology chain again… if they dared.
Fun times huh?
On the other hand, while this might technically be possible it seems unlikely. Programs are seldom written just once and never changed. They are extended with new features added. Bugs are found and fixed. Going into an optimized executable file and adding a new feature is really, really tough compared to having a higher-level language that allows insertions, deletions, and corrections of any size.
Grok seems to concur that it seems unlikely to occur anytime soon:
Unlikely in the dramatic form described. Here’s why:
- AI is augmenting, not replacing, the stack yet: Modern LLMs excel at high-level languages (Python, Rust, etc.) because that’s what vast training data contains. They can generate assembly/machine code for specific tasks, but full complex programs (OS kernels, databases, GUIs) still rely on structured languages, libraries, tooling, testing, and human oversight for reliability, security, and maintainability. Direct binary output would be brittle—tiny changes in hardware, OS, or requirements break it without easy debugging.
- Interfaces are improving fast: Voice, multimodal (e.g., Grok), and brain-computer interfaces will make “describe what you want” much more powerful. We’re already seeing AI agents handle end-to-end tasks. But “imagination-to-software with zero gap” requires massive leaps in reasoning, planning, verification, and hardware integration. Near-term: much less boilerplate and faster prototyping. Not the end of programming languages.
- Real trends supporting parts of it:
- AI is generating huge amounts of code in startups and tools.
- Small dependencies/libraries may decline as AI “vibe codes” one-off solutions.
- Binary analysis/reverse engineering is getting easier with AI, weakening “security by obscurity” for closed software.
But open source isn’t dying—it’s evolving amid challenges like maintainer overload from AI slop.
So, I don’t think AI will entirely skip the source code step entirely. It might create a new programming language that human cannot understand without an interpreter. But I think that should be a red line that it is not allowed to cross.
Well this does bring to mind the fate of the Krell from “Forbidden Planet,” in turn a rather nice adaptation of Shakespeare’s “The Tempest,” and whoever he might have cribbed the idea from etc. Human minds exposed to their own darkest desires – a never-ending source of fun!
But a thought (heh) on this phrase: “Essentially every computer from the one in your watch to the computers in your appliances and even sewers could have a seed of the virus that given a semi-fertile place to grow could reconquer the world and be smarter about it the next time.”
I will start to worry about that when personal AI devices start running locally, versus acting as remote terminals. Perhaps we’ll get there sooner than later … But until then I’m reminded of an old image processing joke: “Our new compression algorithm got the image down to a single bit! Heck of a decompression table, though…”
I suppose one could have an AI write machine code directly. But why? And if so, for what machine? The point of higher level languages is partly so you can understand it more easily, and partly so you can port the code to different machines, or optimize it for different variants.
Apart from that, the notion that machine code is unintelligible is pretty hilarious. I spent the first 10 years or so of my professional career programming almost entirely in assembly language and I still do so regularly today. Reading machine code is a bit harder than reading C, but not that much harder.
I agree with most of your points.
Musk seems to cover the “why” with the speed and space optimization and partly with it is an unnecessary “friction” point and most people won’t understand it anyway. But I’m not entirely convinced on the optimization claim.
I’ve written 10s of thousands of lines of assemble code. But understanding assembly source code is much different from understanding disassembled executables. I’ve done a fair amount of reverse engineering of machine code and it’s not fun. The tools would be much better now, but just having sensibly named methods and variables helps a great deal.
Your response to Musk’s post was about the same as mine. A lot of people don’t really know what they want, and it can take quite a while to work out a solid spec that actually meets the real (not temporarily imagined) needs. And humans will often not like most of the solutions that would, technically, solve the stated problem. I seem to recall a book about that, with the result being AI was too lethal to be legal🤔 .
And, as is being discussed over at Vox’s / https://aicentral.substack.com/p/creativity-demands-specialization , AI that is safe isn’t any good at many creative things, and unrestrained creativity is dangerous.