Much like the Internet, AI is an amplifier. Sensible usage makes you better, irresponsible usage makes you worse. With the Internet, the negative end of this spectrum is addiction: porn, gambling, social media. With AI, it’s a hyper-persuasive propaganda machine that can convince

It takes great strength and wisdom to pursue truth when we’re all drowning in lies, and when you speak some truth, it will often sound like you are losing your mind.

Symbolic of Big Tech conduct in general. Consuming a shared resource without playing fair.

There’s a reason they call him “Sam Snakeman.”

Eleven new research papers related to Internet censorship (from both NDSS Symposium and FOCI) are now on CensorBib, all thanks to gfw.report! censorbib.nymity.ch

Just signed up for an email delivery API service. The confirmation emails never arrived. Classic “you had one job!”

Taxes are to the government what food stamps are to its citizenry. Both create overreliance and learned helplessness.

Today I voted in the Primary Elections in Illinois, a state that does not require a photo ID to vote. I had to provide a name, address, and date of birth to get my ballot – all usually part of the endless data leaks found online. Not hard to cast a vote in someone else’s name.

The problem with data modeling is that if you get it 80% right you may end up in the same bad spot as when you get it only 20% right. Some things you really have to nail.

I’ve been experimenting with the Delegated Types pattern. Best approach I’ve seen for modeling hierarchical data (files, folders, comments), both in code and in SQL. It was originally built for Rails but you can make it work in Go too, albeit with some extra boilerplate.

Much better term for AI psychosis: Cyberpsycho, as coined by Cyberpunk 2077.

The trifecta of observability (logging, metrics, traces) is incredibly cumbersome to work with. It should really all just be one thing, like a single JSON object per HTTP request – very feasible. You can squeeze it all in there (logs, spans) and even derive metrics from it.

I wrote about data validation in Go. A simple interface and some tooling on top of it is all that’s needed for validating data in complex software projects: nymity.ch/writing/articl…