> One wonders whether a generation that demands instant satisfaction of all its needs and instant solution of the world's problems will produce anything of lasting value. Such a generation, even when equipped with the most modern technology, will be essentially primitive — it will stand in awe of nature, and submit to the tutelage of medicine men.
I still use plan mode, but during plan mode I will spin out sub agents to implement the current draft in a tmp dir and bring back lessons. I feel this keeps me grounded in my original starting point, rather than ending up with a implemention that meanders through the agents own discovery process. It also lets me ask concrete questions about (possible) implementation
I thought of that too and looked into it. (Or rather had an AI look into it! ;-)) Fortunately, there are extreme physical constraints on how quickly manual labor could be replaced, especially anything requiring dexterous work. Mainly, supply chain constraints on rare earth elements, which are required to make these motors feasible.
By some estimates, today's global supply is like 2% (or was it 0.2%?) of the total amount required to enable such widespread deployment of robots. Until we can scale that up (or figure out how to bypass that dependency) there is a very hard constraint on how fast they can be deployed.
I have a hard time believing that won't be overcome or surpassed when you compare it to the cost of producing an adult human capable of skilled trades.
The incentive is definitely there, but right now we physically just don’t have enough rare earth supply to make the number of motors required. So we simply cannot scale that production up to the level where it would be cost effective to displace humans. This is assuming fine motor control is “solved” which it is not currently, but I think LLMs will take care of that soon enough.
I think the company that cracks the mass market robotics puzzle will be the next trillion (quadrillion?) dollar company. I suspect that would require bypassing the dependency on rare earths, but there seem to be no serious contenders out there? All the big shots are focusing on using current technologies because they basically “solve” motors. Maybe the company that does crack this puzzle will come out of left field with an unconventional technology that scales with whatever raw material is abundant, like OpenAI did with ChatGPT and compute.
Yeah, that's a great question. I should probably write about it properly, if only to structure my own thoughts.
The main thing I gained was the wisdom that comes from the pain of doing it wrong.
The most painful lesson was finding out I was more attached to the idea that I had a validated product, than to actually having a validated product. (And treating marketing as an afterthought... lmao)
Also if I had realized how long it was going to take, I would have taken a very different approach. I accidentally took my R&D mindset into production, which... might have worked, if I were immortal and retired!
In short, aim to ship 10x sooner, because everything takes 10x longer than you think. (Reality has 100x more detail than it ought to.)
On the bright side, I also learned to put one damn foot in front of the other, persist despite the horror, and finish the damn thing.
> reducing the skill required to operate AI agents
For accomplishing the same task, yes.
But given these tools straight out of science fiction, why on earth would you be stuck doing the same things? There's no point spending human thought over something an agent has just automated yesterday.
Think bigger, take on more ambitious projects that are perpetually at the limits of what you and AI can accomplish.
> one where software engineers can focus on data structures, software architecture and algorithms.
I see this a lot and I'm not sure why people don't think AI will be able to do this too. The self-play training that got them writing code can be used for this too.
Given enough context for a business problem, sure.
But LLMs are not in a condition to judge how you should pick the technical solution to a business problem with several stakeholders, risks, and so on.
With the speed AI moves, a lot of technical decisions become reversible. And while engineering often makes decisions that could lie elsewhere in the business, outside of engineering, I could imagine those decisions moving elsewhere in a fully AI world.
Do you have examples of things that would be hard to train for? One that could be compensated for with changes elsewhere in the business process?
Yes, if Alex from BizDev is a scheming moron who consistently lies about the priority of features, it’s hard to keep an LLM on the loop about it when transcribing meetings and feature requests. If your boss is gonna be unavailable for a month and that means that a junior devs garbage PRs will be getting merged because the second in command is much laxer then you need to be aware for that and so on.
What I mean is that these things decant into technical decisions and even with all the AI in the world running a DB schema migration does not become any more trivial.
In the long term, I think you're correct. In the medium term, AI still won't know your business-specific workflows and data relationships, and humans are needed to define those things and let the AI build the scaffolding around it.
- Eric Hoffer
reply