I have recently been working with Google Jules and it has a similar approach. It spins up VMs and goes through tasks given.

It does not handle any infrastructure, so no hosting. It allows me to set multiple small tasks, come back and check, confirm and move forward to see a new branch on GitHub. I open a PR, do my checks (locally if I need to) and merge.

I understand that we are slowly taking away our own jobs but I do not find it depressing. I do find it concerning since most people do not talk about this openly. We are not sure how we are restructure so many jobs. If we cannot find jobs, what is the financial future for a large number of people across the world. This needs more thinking, honest acceptance of the situation. It will happen, we should take a positive approach to finding a new future.

Yes the prompts are clear enough but it depends on the capacity of the people involved. People have to internalize the math (or any other) concepts from language into some rules, syntax, etc.

This is what an agent can do with an LLM. LLMs can help take English and generate some sort of an algorithm. The agent stores algorithm not the prompt. I do not know what current commercially available agents do but this was always clear to me.

I am not sure I got your point about English. I thought Karpathy was talking about English being the language of prompts, not output. Outputs can be English but if the goal is to compute using the output, then we need structured output (JSON, snippets of code, etc.), not English.

This is something I have been attempting for quite a while now. One simple tool I started is a deterministic data extraction system where AI helps in finding out the data to be extracted but then the code would try and "template" it. When we have the template, the extraction on any similar string would happen deterministically.

Think of extracting parts of an email subject. LLM is great at going through unseen subject lines and telling us what can be extracted. We ask LLM what it found, where. For things like dates, times, city, country etc, we can then deterministically re-run on new strings to extract.

https://github.com/pixlie/determined