jeremy

member of the technically staff @ mit csail

[jeremy @jercarin]

@_ueaj I will do whatever is necessary in order to receive a 1 pm ET tweet on my timeline from anthropic AI with the word fable in it

[jeremy @jercarin]

@NoahZiems @eliebakouch You were right :)

[Max Zeff @ZeffMax]

NEW: Anthropic is walking back Claude Fable 5's policy to covertly degrade performance for competing AI researchers, after facing fierce backlash.

“We’re changing Fable 5’s safeguards for frontier LLM development to make them visible,” Anthropic tells WIRED. “We made the wrong tradeoff and we apologize for not getting the balance right.”

[jeremy @jercarin]

@NoahZiems @eliebakouch fair, maybe. generally speaking I take anthropic at their word -- they think they are building the next atomic bomb, and it makes sense to not assist your adversaries from building an atomic bomb in an arms race. not that I agree, but I believe that *they* believe it

[jeremy @jercarin]

@NoahZiems what makes you think they'll reverse course?

[jeremy @jercarin]

@RLanceMartin given that your example is related to ai research, it'd be great if you all could surface if fable is being degraded for ai research :)

[jeremy @jercarin]

@synthwavedd fuck its over

[jeremy @jercarin]

@tokenbender another possibility is that they've done a ton of serving optimizations and their margins on Opus are absurdly massive these days (or served on older hardware, Mythos served on B200s)

[jeremy @jercarin]

@tekbog I really don't understand why it's so hard to believe that they wanted to implement new security safeguards first.

[jeremy @jercarin]

@keysmashbandit Who is building this?

[jeremy @jercarin]

@max_spero_ @pangram now update your bio ;)

[jeremy @jercarin]

@willccbb Kind of disagree. The only way I’ve gotten good writing out of LLMs is to dump tons of my related writing or writing I’m trying to emulate into context

[jeremy @jercarin]

@banteg I am in disbelief this is real. what could possibly be going on over at meta to let this happen

[jeremy @jercarin]

@RLanceMartin @mitsuhiko @trq212 @badlogicgames this is great thanks, but here I was thinking you guys introduced a technique to edit / prune cache without fully resetting it. now that would be sick (cache cleanup for useless tool calls/context!)

[jeremy @jercarin]

@LukeParkerDev maybe you’re aware but the bottom thing is exactly what Claude code does. It injects a prompt every x turns if certain conditions haven’t been hit. Sometimes you’ll catch Claude saying “don’t need to do a todo list but thx for the reminder”, when you didn’t say anything abt todos

[jeremy @jercarin]

@polynoamial can we have a big pretrain please, noam. is this the big pretrain

[jeremy @jercarin]

@threepointone @dillon_mulroy I will not be having bitter coffee Sunil. I will have pour over with ethiopian beans. Is that ok

[jeremy @jercarin]

@pangram @max_spero_ @ben_golub you guys may want to tighten your bot trigger conditions lol

[jeremy @jercarin]

@max_spero_ @ben_golub @pangram Do you think, if the judges were using GPT to pick / filter writing (not saying they were), would GPT prefer its own writing / AI slop? does it have internal AI text detection in some way?

[jeremy @jercarin]

@a1zhang lmao it's feeling like a discount-store sam kriss long form tech profile

[jeremy @jercarin]

@maximelabonne I think their argument was that the tokenizer cost increase roughly cancels out with greater reasoning efficiency (perhaps in part due to the tokenizer change). So far, I’m not convinced this is true.