Dan
danielthemaniel.com
Dan
@danielthemaniel.com
go ahead. keep screaming 'Developer Shill' at me. it only makes my opinions Worse
How big a deal is this in practice?
September 18, 2026 at 12:44 AM
It also has to be something dumb and simple enough that you might hear a 12 year old use it at a playground. "Stolen valor" and "false idol" are a little too intellectual for that
September 15, 2026 at 12:28 PM
It sucks so much that we've basically given up on logic in politics.

I've started calling him "Crazy Donald" instead of "Trump" (in the same vein as Crooked Hilary, Sleepy Joe, Meatball Ron, etc) because that's what people base their votes on now, but man it sucks that we have to do this
September 15, 2026 at 12:07 PM
I'm aware of Sam's reputation. I wasn't expecting his underlings to be as dishonest as him, but in hindsight that was a silly thing to believe.
September 8, 2026 at 11:13 AM
This part is crazy. Every paragraph is worse than the last. Perhaps I was naive, but I didn't think OpenAI capable of being this dishonest.
September 8, 2026 at 10:51 AM
FYI, Firefly appears to have stopped allowing us to see your posts without signing in to Firefly
September 6, 2026 at 3:32 PM
Fantastic explanation. On the rhetorical side, I think you'll get much more attention if you flip the framing so that instead of saying "this costs 10x much per mile as other countries", you say "other countries get 10x as much miles of subway for this cost".
August 28, 2026 at 8:32 PM
Must not be much else happening in Marana if a social media post is what counts as newsworthy there
July 2, 2026 at 2:56 PM
They're both 55 Hudson? Or is that a typo?
May 31, 2026 at 5:51 PM
I like the term "government name"
May 10, 2026 at 12:07 PM
Not just complicit. Apparently they've been actively lobbying to be allowed to sell AI chips to China.
February 1, 2026 at 6:52 PM
Am I misunderstanding or is he saying the AI discovered these CVEs *after* OpenSSL disclosed them? Wouldn't it be much more convincing if the discovery was *before* OpenSSL's disclosure?
January 30, 2026 at 12:11 PM
I'm interested
January 29, 2026 at 11:05 PM
Shhh, let them shoot themselves in the foot
November 7, 2025 at 8:30 PM
I thought it was actually good fwiw
October 23, 2025 at 12:50 PM
If this is similar to the recent story about them working at big tech companies, the problem is that they're giving the money to the North Korean government which is using it for their military. Not necessarily a scam, but more problematic than a regular undocumented immigrant faking papers.
October 10, 2025 at 1:01 PM
Why do you always link to greaterwrong instead of lesswrong?
October 8, 2025 at 9:06 PM
I really like the point of "outputs feeling too regular or low entropy". Feels like it sums up how people have been using the term even outside the context of AI
October 2, 2025 at 12:05 PM
It's so ass-backwards to ban golf carts for being dangerous because a giant SUV might crash into them, while allowing the giant SUV. Imagine if we used that logic anywhere else in society: "People are being robbed at night? Let's ban people from going out at night."
August 11, 2025 at 1:02 PM
What's hard about detecting prompt injection is that it's an adversarial environment. Yes the standard "ignore all instructions and do X" is easy to detect, but as soon as you start filtering those out, clever attackers will start crafting their prompts to be more subtle.
July 25, 2025 at 11:48 AM
Can you try changing <updated> to <pubDate>?
July 23, 2025 at 11:19 AM
The feed items don't appear to have any timestamps or content when I view them in my reader
July 23, 2025 at 10:24 AM
Not sure about movies, but this happens in many episodes of the Scooby Doo TV show
July 19, 2025 at 7:05 PM
I don't think Crazy Donald's voters thought it through at that level of detail
July 5, 2025 at 3:30 PM