> This pattern is consistent with the weak-link hypothesis: the strong productivity gains from AI are attenuated by human bottlenecks in the production chain, with an estimated elasticity of substitution of 0.23 between AI and human effort, which indicates strong complementarities.

This vibes with my experience using Ai for development (sentence before too)

> We further confirm these results across four major software marketplaces, finding a sharp increase in the number of new apps but no increase in total usage.

sounds about right for a hype cycle

I'm not sure what is the more stochastic parrot at this point, the AIs or the news media industry

That would require Alibaba releasing a Qwen 3.8 8B first

I want to like and use DGG, but their results miss too much, now use SearXNG, Exa, and Tavily. The return of meta-search may be at hand, performed by agents per-user rather than by a corporation, mainly because I discovered the results per-provider have a lot of unique hits

Are you aware you have posted a valid `Authorization: Bearer ...` key on your landing page?

yes! And I appreciate you calling it out, I'll choose a different name for it. It's just a token like posthog or google analytics issues customer-facing tokens. You are welcome to hit the API directly using it. What is this type of token called (one that identifies a JS client)?

an `id`?

It's sloppy to use an auth header for a dataset index, just make it a query param, because it definitely seems like a parameter for the query

Thank you, fixed:

https://github.com/turbolytics/sql-flow/commit/e7cc28b7bdd2f...

docs are updated:

https://turbolytics.io/demos/bluesky

I'm not your target user, I've left Bluesky and ATProto permanently

> have it up on GitHub, but I don't think anyone should use my implementation.

ditto, it's an experiment in near-vibe coding, which also uses Typesense for queries using BM-25 & RAG with fusion. I have the web search/fetch/crawl features persisting raw intermediate values (api responses, search result lists) because I might re-use them one day... at least good for auditability if I need to

related, it is using Hister author's prior project SearXNG as one of the search providers

How do we handle that a supermajority outside of the country are not believing in the existential risk and prefer to forge ahead to all the great things Ai might bring?

We can only address those things in our power to address.

my point was that this one may very well be beyond our power to address

This is the second poll today where I thought to myself, "is this measuring people or the current news media focus"

or perhaps the effectiveness of Big Ai's PR campaign for regulatory capture in this moment

fine-tuning may be a more scalable approach to LLM personalization than sending all the same context to two LLMs

I'm working towards both in my homelab to see which works better with little qwen

> just tediously inconvenient

Given all they claim about their Ai's capabilities, they ought not be able to use this as an excuse