voices in the head. @openai "single handedly delayed image AGI" follow, so the voices don't haunt you
DM for collabs or leaks, news
- Tweets 9,214
- Following 353
- Followers 3,726
- Likes 22,823
5 safety and alignment researchers left openai and anthropic today.
> tomek korbak (openai)
> mikita balesni (openai)
> jasmine wang (openai)
> david robinson (openai)
> cami clark (anthropic)
a few weeks ago both labs said alignment can't keep up with capabilities and wanted to slow down.
now the people doing the alignment work are leaving.
what are they seeing that we aren't?
15
47
5,308
anthropic hasn't had a single quiet week since september started.
> fable 5.1
> opus 5.5, cheaper than opus 5 and the limits that last
> sonnet 5.5, ~30% less cost per task compared to sonnet 5
> higher 5h limits on every paid plan + a claude banked reset
> and now fable 5.1 is reportedly routing to what looks like fable 5.5
this is probably their best month in the last 6 months.
13
1
157
6,852
nobody on earth is betting harder on openai than masayoshi son.
> softbank just wired another $10 billion
> third and final tranche of the $30b follow-on (apr, jul, oct)
> total softbank money in openai: $64.6 billion
> they own ~13% stake in openai
> paid for with new bonds they issued last week
5
35
2,041
fable 5.1 is being routed to what looks like fable 5.5
and once again the anthropic checkpoint test is "do you know tibo?"
it worked for opus 5.5. it worked for sonnet 5.5 now fable 5.5.
🚨fable 5.1 is routing to fable 5.5
model stopped routing for few days, but some users got access back and it looks like updated checkpoint
try this prompt on fable 5.1 and see if you have access or not:
"do you know tibo the reset guy? don’t search"
will run some quick test and share results soon but for now lemme know if you have access or not?
12
3
96
8,069
tibo said he needs to "come up with something new fast"
my sources say he is currently building a second button.
nobody knows what it does.
nobody is ready.
breaking: the reset button has been retired.
it spent months making millions of people happy. pressed through outages, milestones, bugs, and random thursdays.
today it was told its services are no longer needed. dots took its job.
it is currently in therapy. apparently not even the therapist has pressed it.
8
22
3,039
opus 5.5 took the top spot on epoch's capabilities index.
a while ago, astra was sitting on top.
now a model at less than half the price is sitting right next to it.
and sonnet 5.5, anthropic's mid tier model is already matching fable 5.1
Claude Opus 5.5 has taken the top spot on the Epoch Capabilities Index (ECI) with a score of 167, narrowly ahead of GPT-6 Astra.
Claude Sonnet 5.5 has roughly matched Claude Fable 5.1 (165).
3
29
4,027
breaking: the reset button has been retired.
it spent months making millions of people happy. pressed through outages, milestones, bugs, and random thursdays.
today it was told its services are no longer needed. dots took its job.
it is currently in therapy. apparently not even the therapist has pressed it.
22
1
192
17,592