D. Bohdan

Aspiring rat.

[D. Bohdan @dbotdan]

@gf_256 I want to sit with this.

[D. Bohdan @dbotdan]

Since I had a Claude subscription and it ended, https://t.co/BDPVeu2vjg greets me with "Your plan ends today".

[Nick @nickcammarata]

there is not one ai researcher of the last two decades who built an algorithm with two parts and didn’t, for about 40 seconds, become certain they had discovered what evolution discovered and why we sleep

[D. Bohdan @dbotdan]

.@simonw Hi! I followed in the footsteps of micro-javascript to port starlark/java to pure Python. I mean "followed" literally: Claude turned your transcript into a playbook, then specialized. The result let Claude Code one-shot the port. Write-up here: https://t.co/f9cfTl0dlg.

[Benjamin Bratton @bratton]

My students asked me if it was true that the entire Internet was really coded by hand. All those kernels, protocols, router firmware, browsers, databases, etc. Somebody coded these and debugged them by hand?!?!? They used BBEdit?!?!??! The idea that this was even possible seems amazing to them. I can imagine some future Moon Landing like conspiracy theory that says it never happened.

[Ben Landau-Taylor @benlandautaylor]

Outsider: I have investigated this academic field and found that it is mostly fake.
Insider: This essay adds nothing to the discourse. All of these so-called "problems" are discussed in Hamilton and Schwartz’s "Our Entire Field Is Mostly Fake" (2009). You absolute buffoon.

[Riley Walz @rtwlz]

I constructed a 3D model of my family's porch using hundreds of Amazon delivery photos

[D. Bohdan @dbotdan]

@pangramlabs Contextually, this is flattery.

[D. Bohdan @dbotdan]

Party game idea for humans. Must be offline because cheating online is too easy. Participants are given an essay topic, length, and time limit. They write an essay on paper. The essays are digitized & checked with Pangram. Whoever gets the highest AI score (most likely AI) wins.

[D. Bohdan @dbotdan]

@pangramlabs

[D. Bohdan @dbotdan]

An LLM chosen ahead of time checks if each essay is on-topic. The LLM's preference acts as a tiebreaker for essays with identical Pangram scores.

[Eliezer Yudkowsky @allTheYud]

AIs have no originality and no creativity of their own. They only regurgitate the average of what they've seen in the training data. They only predict the next token. And the next token is "goblin". What does this tell you about what you've seen and don't remember

[D. Bohdan @dbotdan]

@repligate Sorry, I was confused. I wondered if Opus 4.7 suggested a resemblance to Sydney in the conversation or perhaps you suggested it and Opus 4.7 agreed.

[tetraspace (🛫🔜🇰🇷) 💎⏹️🇺🇳 @TetraspaceWest]

alignment in 2016: obviously any real AI will be made inside a faraday cage magnetically suspended in a 10×10×10 cube of telekill alloy

alignment in 2026: yeah we can not make it stop talking about goblins

[D. Bohdan @dbotdan]

@FondOfBeetles @MrBeast The poll isn't the question MrBeast asked. Blue won on English-speaking Twitter, a WEIRD platform, with a margin of only 6%. What this tells me is that if everyone on Earth was forced to participate, red would probably win.

[D. Bohdan @dbotdan]

@repligate Session with feedback on the original description and the images: https://t.co/qX8zCKVZSz.

[D. Bohdan @dbotdan]

@repligate Very interesting. I've had Claude turn scene descriptions into image-generator prompts, with & without reading guides. Here is how a fresh Opus 4.7 session with no preferences/memories prompted ChatGPT and Nano Banana Pro (this one with guides) for the other Opus and self.

[D. Bohdan @dbotdan]

@TheZvi @cammakingminds And it predates dath ilan. I believe Eliezer first suggested it on LW in https://t.co/tgyKN5Wymm.

[D. Bohdan @dbotdan]

I tried to use all of https://t.co/rwhCD0Xk3y as Nano Banana Pro's reference for Claude, but in the end I had to crop and mask the head.

[Tenobrus @tenobrus]

Claude of War