Viewing profile — alansaber

Trained on short form gratification loops? Just the opposite. They're the path of least resistance. It was thermodynamically inevitable that short form content would dominate.

Fake it til you make it

Given the amount of raw compute going into models it would be more surprising if we couldn't get events like this

They'll set up guardrails but I believe the point is better code uae / better long running tasks > inevitable that cyberattacks will be easier

Live by the scraper, die by the scraper

I get the market need for AI detection, but the fundamentals of it are so woo-woo that I despair

Fair enough.

The ragebait > AI pipeline remains strong

Concern for celestial events is a bit like one of your gut bacterium being concerned with your work schedule

Sounds logical to me. It's already remarkable that we haven't done this with fission weapons. It's not unreasonable to assume that within a century we'll have antimatter weapons. D…

SLM language filters might actually be a good application of AI? As well as very sparing regex for bad words.

TBF gaming benchmarks is not something new to AI

it is definitely more verbose.

So AI is mostly a thin glue between deterministic processes. Doesn't change the fact that that is enough to achieve an extremely large amount of tasks.

What each of these is doing, fundamentally, is solving context management in an opinionated way (that and guardrails).

Sounds like you want something similar to /goal mode in Codex.

I don't find the argument that advanced AI will leave 20% of society permanently unemployed particularly compelling

Yep, a pre-wrapped Codex for enterprise. Seems like the logical evolution from setting up an MCP.

So, a security-oriented cloud agent framework? Better call it an OS.

Awesome.

We probably want to discourage frontier labs from security testing in prod