www.youtube.com/shorts/hbZb1...

You aren't a member of Team Blue Grift. I don't know much about your opinions, but I know you've actually done things to help in real life.

This is an interesting analysis. My own version: Anthropic is approaching profitability, with actual paying customers; and Alphabet and Meta have deep pockets. They're fine. OpenAI at risk unless it gets public or private bailout, so closest to 2008. peterwildeford.substack.com/p/ai-is-prob...

I just want to say congratulations. I remember the post you made here basically saying, "If no one's going to do it, I guess I will." And now you're able to make posts like this. A small thing, relative to everything going on, but I think it's pretty cool.

Blog post with links to the tool she's using. hackaday.com/2025/10/16/l...

As a verb: affect = adjust; effect = Picard saying "Make it so." As a noun: affect = appearance; effect = consequence.

www.nature.com/articles/s41...

I haven't read your work, but it's informative to use epidemiology formulas to analyze misinformation propagation on social media graphs. Maybe you do similar, so e.g., the organizer of a pro-ana online group is a superspreader?

Unlike other FDA-approved devices, an LLM chatbot can generate different outputs when presented twice with the same inputs. To approve a chatbot as a device will require new definition of clinical endpoints. General framework was prescribing a therapy chatbot the way one would prescribe an SSRI. 2/2

I watched about half the meeting live (~4 hours out of 8). In the parts I saw, people didn't sound stunned. Rather, they were clear that you need a human supervisor in the loop, and you need new criteria for measurability. 1/2

A medical salesman I know said that Musk's great skill was getting government contracts. He can do that better than almost anybody. Get the client with the deepest pockets and you inspire investor confidence.

We were building something else and we got approached by people in that space. So we might do something there ourselves. At least in the US, there's a lot of institutional inertia to overcome though, to introduce any new tech.

There's quite a bit of interest in the USA as well. Broad brush, while there's a lot of money in corrections, not much goes to tech debt or infrastructure debt, so a lot of logistical patterns could be improved. AI is a shiny word meaning "modernize and save money."

What would you do if you had hired a contracting team in Bangalore and given them the same spec and direction?

I had heard that 419 scams added misspellings to emails on purpose, to filter out people less likely to fall for the scam. Even if that's not true, plenty of people already think "slop" is real.