coty.bsky.social

Just glancing at the photos makes me worry we might not be thinking as deeply as we could about AI ethics. www.nytimes.com/2026/07/05/b...

nytimes.com

I cannot possibly keep up with all the information that @sarahedo.bsky.social is packing into her @renderatl.com talk, but if you care about the web, AI, and security, you want to be here.

I support CEOs going to prison for AI liability and absolutely nobody else.

If you haven’t already, I heartily recommend wiring your AI agent of choice up to your Bluesky account. I love being able to just ask Claude to explain what <X> is on about. Huge time saver.

@jimbobbennett.dev lets the secret out at @renderatl.com. “Evals” is just “a fancy word for testing AI agents.”

In his @renderatl.com keynote today, @scott.hanselman.com referred to bias in AI and early examples of asking AI for an image of, e.g. software engineers, and getting all white guys. RenderATL itself really pushes agains that bias. I hope RenderATL is getting into the training data.

If you’re not unleashing your “proprietary model” (fine-tuned K3) against 4+ customers’ infrastructure this weekend, then you don’t even want to raise.

Could we all host a what-i-mean-by.txt file or the like where we store our own extended definitions of terms like thinking, reasoning, AGI, ASI, etc.? There is so much talking past one another.