bishplsok

Actually Useful(tm) result out of the recent influx of AI hype -- "search engine" trained on, by and large, how people recommend links on sites like Reddit. It's a little hit or miss in my experience still, but when it hits it's legitimately brilliant, and mostly avoids the "Google search results are useless and dominated by SEO-farm garbage" issue.

https://metaphor.systems/

Said emails are in the link I posted above! It is, hilariously enough, literally someone trying to defend Scott by just reposting his emails in their entirety -- "Sunlight is the best disinfectant" indeed. If it's a hit piece, it's Scott writing a hit piece on himself.

"As little bias as you can," huh.

https://emilkirkegaard.dk/en/2021/02/backstabber-brennan-kni...

He's a neoreactionary apologist and self-admitted scientific racist. Not...exactly a bastion of unbiased opinion.

(And that's without getting into the deeply problematic nature of the capital-R Rationalists' horrible abuse of Bayesian reasoning as a replacement for actual expertise or scientific inquiry. Assigning arbitrary likelihoods to your opinions about the world does not make you unbiased!)

So, I just tested this; it's an access token that has no authorized scopes to access sensitive data. It literally only exposes public profile data, which basically just consists of name, picture info, and internal Facebook user ID -- the default public profile info listed in the developer docs.

Is it a security flaw? Absolutely. Is Facebook still terrible? Of course.

But this article is, frankly, egregiously inaccurate in its claims on the severity of the issue.