Worse, IMO, is the never taken branch taking up space in branch prediction buffers. Which will cause worse predictions elsewhere (when this branch ip collides with a legitimate ip). Unless I missed a subtlety and never taken branches don’t get assigned any resources until they are taken (which would be pretty smart actually).
Well, I'm sure they were when "cloud" was the latest buzzword in public company reporting. Now that its AI I'm sure the next quarterly will show massive (fabricated by reshuffling) growth in their AI initiative.
I agree. (And believe me, I don’t enjoy fooling myself into believing a better outcome than can be expected.) so far my experience is that AI assistants are a tremendous help, but it still feels like programming and still needs a programmer to drive it.
The thing to look at is Sora. I have, at my fingertips, the power to generate all this short form content. I could make any Instagram Reel I can think up! Except that's not for me. I'd rather sit here with Claude code making shit. That's never going away. Even if there's no money in it, it's how I'm wired.
I share the instinct. but I think we might be wrong directionally.
Looking at history, every time a tool automated part of our work (I’m thinking of calculators, high-level languages, libraries, frameworks) people warned that skipping fundamentals would be fatal. But
None of those transitions made understanding obsolete.
If AI coding agents are different, and not knowing how it all works becomes an unrecoverable error, as you imply, that would be a first. So I am inclined to side with history and guess that a best of both worlds exists. In which, sadly, hand coding is practically gone (I don’t like it either).
We shouldn’t use agents as blindly as the prompt might invite of course. A new engineering discipline will likely emerge: one focused on supervising, validating, and shaping what the agents produce.
Believe me, i do have mixed feelings about all this. I love writing code with my hands. But we shouldn’t fight the tide; we should build a different boat.
Just to add to this, people say the same about eg citizen Kane being such a classic but without the context of it having genre defining firsts, the film doesn’t stand out as much to a modern viewer.
For those unaware, this is a very interesting guy, because he stumbled on (creating, through his business Medal) a valuable AI dataset that - by offering to buy his company - reportedly OpenAI offered him 500M for. The dataset, I understand, is first person game video plus controller actions.
He then realized the value, which is in short a way to teach models real world and gui operation common sense. He can train a model to predict, from video, what a controller would have to do.
This is expected to lead in breakthroughs in robotics, gui controlling, self driving, and more.
He responded by learning deep learning, and starting a new company, general intuition.
I respect this guy a lot for teaching js this.
Absolutely fascinating and I take his opinion seriously.
He's building Eureka Labs[1], an AI-first education company (can't wait to use it). He's both a strong researcher[2] and an unusually gifted technical communicator. His recent videos[3] are excellent educational material.
More broadly though: someone with his track record sharing firsthand observations about agentic coding shouldn't need to justify it by listing current projects. The observations either hold up or they don't.
reply