The resource being extracted is the application of knowledge to tasks.

Which might sound confusing or esoteric, but then again, white collar work exists strictly because there are abstractions over the economy that give that work leverage to run machines (companies) that then proceed to optimize this extraction and value creation process in other domains.

All this to say: the creation of digital <outputs> to be used and place in locations where they mean something to someone in the context of the abstractions built up inside corporations is the thing being extracted. How many of these units of output is the thing that changes and has a relation to Jevons paradox.

rofl

You are a search engine of sorts no?

Search does not necessarily meant "retrieval". It means search, over some space. That space can include ideas or options that are recombinations of other ideas and options. And the search heuristics can be learned I have no doubt, as those can be essentially pretrained.

They may not be shipping good enough products, but on the flip side it still feels like they have almost no competition outside of coding and image gen. The EOL termination of 4-o should be some evidence of this.

OTOH, why don't they ship good enough products? To me all of OpenAIs recent investments strongly suggest they hit a dead end with their current LLM approach. After all, if they knew the path ahead for GPT looks great, why don't they invest into training the next big thing instead of doing datacenters with the intention of renting them out?

This seems accurate, and plausibly the only way out. The biggest issue I see here is that in this case... the greed might topple a government

Something something true capitalism has never been tried before

Too big to fail. It's essentially 08 part 2 with the expected irr.

OpenAI deals will likely end up with government backing in the next 12 months. Then we’ll all be on the hook for it.

But this is not a bank, or an airline, or a real estate giant.

If OpenAI goes bankrupt, what happens? People won’t be able to write their precious slop oh no and serious professionals will just switch to any other LLM provider

Users might simply switch to Gemini or Claude but partnered companies might go bankrupt or see their stock tank.

You think so? What features are differentiated enough to warrant that?

Seems like they all have tools for reading files, editing, running shell commands.

Cursor has linting access, but this can be added to claude with hooks I think.

Subagent personalities, skills, structured output generation, customizable context cleanup, etc.

Most of those seem kind of useless... I probably do not know how to use them correctly, so, what would you recommend by way of subagents and skills?