@emollick This standard scenario was obviously in the training data
English
Michael Panchenko
390 posts
Michael Panchenko
@MischaPanch
Previously Lead AI Researcher, now cofounder at https://t.co/f7m9Oq1Mqh Integrity and critical thinking absolutist.
München, Bayern Katılım Temmuz 2023
126 Takip Edilen86 Takipçiler
I had early Gemini Omni access: "sea otter in a pilot's uniform explains why Spirit Airlines went bankrupt to a river otter who is distracted by their laptop while they are in a hot air balloon over NYC. in the next balloon over, william shakespeare fights a robot made of pizza"
English
@ChrisWillx Nietzsche does not seem qualified to give any advice on marriage tbh
English
Nietzsche’s advice on marriage.
“Marriage as a long conversation.
When entering a marriage, one should ask the question: do you think you will be able to have good conversations with this woman right into old age?
Everything else in marriage is transitory, but most of the time in interaction is spent in conversation.”
English
Germany / Switzerland - who wants to be on this?
Vaibhav (VB) Srivastav@reach_vb
putting together a group chat for Codex power users in London / Europe who are the biggest ballers around?
English
@AlmaMaito Come on, that's only true for like 90% of apartments, don't exaggerate!
English
WHY DID NOBODY TELL ME THAT IN GERMANY YOU NEED TO BRING YOUR OWN KITCHEN WHEN RENTING AN APARTMENT?!?!?!
English
@Ming_LLM @kskrygan You should check this out, gives full IDE tooling to an agent via MCP
github.com/oraios/serena
English
JetBrains uses IDE-native code insight to make AI agents better at large codebases
Why burn time and tokens on grep when the IDE has semantic search, navigation, and refactorings?
Interesting: Codex seems to use these tools much better than Claude Code
blog.jetbrains.com/ai/2026/05/wha…
English
@greg_ashman I think „we are deluded“ is the right interpretation here. Though personally, I‘d exclude myself from the we here. I never thought that consciousness is something very special or interesting once I reflected on that topic.
English
The reaction to Dawkins deciding Claude is conscious is fascinating.
It really is just the Strong AI position that Roger Penrose was criticising in the 1980s. If you think consciousness is just an emergent property of a sufficiently complex computer then of course AI is conscious. It passes the Turing test and that’s it.
The really interesting part is why it is obvious to so many of us that AI is *not* conscious: obvious to the point we think Dawkins’ credulity is amusing. What are we basing that on? Are we deluded or is there something else to consciousness that we cannot articulate but that we clearly sense?
English
oraios-ai.de got me to like them instantly with this bit of Serena docs
Very interesting project, appreciate the OSS @MischaPanch 🫡
English
@JadenJTW That's just not true, look at it like this:
Normalizing value(own_life)=1, the choice becomes between
p(blue<50%)
vs
p(your_vote_is_decisive)*value(half_of_humanity)
The latter is very hard to estimate, hence there can legitimately be different opinions.
English
@eigenrobot Then, without reason, god is reborn and the cycle continues
English
first we killed god, then we killed reason. what's up next
English
@micsolana You can make a rational case from the consequentialist lens. Normalizing value(own_life)=1, the choice becomes
p(blue<50%)
vs
p(your_vote_is_decisive)*value(half_of_humanity)
The latter is very hard to estimate, hence there can be legitimate different opinions.
English
fwiw, and this is embarrassing but I'm going to admit it, my instinct was blue, and I pressed blue. then I thought about it for a moment and the answer was clearly red. I can't make a rational case for blue, but understand where blues are coming from (they are wrong).
Tim Urban@waitbutwhy
Everyone in the world has to take a private vote by pressing a red or blue button. If more than 50% of people press the blue button, everyone survives. If less than 50% of people press the blue button, only people who pressed the red button survive. Which button would you press?
English
@MrBeast I'm surprised nobody analyzed the blue vs red button from the consequentialist lens. Normalizing value(own_life)=1, it becomes
p(blue<50%)
vs
p(decisive_vote)*value(half_of_humanity)
The latter is hard to estimate (small*big), hence there can be different opinions
English
Everyone on earth takes a private vote by pressing a red or blue button. If more than 50% of people press the blue button, everyone survives. If less than 50% of people press the blue button, only people who pressed the red button survive. Which button would you press? BE HONEST.
English
I'm surprised nobody analyzed the blue vs red button from the consequentialist lens. Normalizing value(own_life)=1, it becomes
p(blue<50%)
vs
p(decisive_vote)*value(half_of_humanity)
The latter is very hard to estimate, hence there can be legitimate different opinions.
English
@Star_Knight12 That only happens if you don't give it access to a rename tool. Even claude's own internal LSP tools can do that (for some languages at least), but the best way to achieve that is with Serena MCP. Then the renaming is a single tool call
English
@kskrygan This is also good news for the (JetBrains-powered) Serena MCP toolkit. Now using fewer tokens will be important again, and nothing saves more tokens than intelligent search, edit and refactorings powered by JetBrains' indexing engines
github.com/oraios/serena
English
We saw this coming early — and stopped subsidizing AI back then.
JetBrains has been 100% private from day one. No hype-helicopter money.
While others chase VC-fueled hype, we focus on what actually matters: real value and playing the long game.
Gergely Orosz@GergelyOrosz
Its the beginning of the end of subsidized AI subscriptions. GH Copilot is moving to usage-based billing, as has Claude (for business customers.) Fair to assume more will follow. I expect this change will also be a great boost for open models - cheaper, and pretty good already
English
@waitbutwhy What happens if the percentage for blue-culling is 5% or 95%. I wonder how that would affect the vote. I don't think it's trivial either, 5% means feeling less pressure to "contribute" to blue
English
Everyone in the world has to take a private vote by pressing a red or blue button. If more than 50% of people press the blue button, everyone survives. If less than 50% of people press the blue button, only people who pressed the red button survive. Which button would you press?
English
@waitbutwhy I’m curious how different the results would be if you framed it differently.
“Everyone who presses red lives. If everyone presses red, everyone lives. No consequences.
If anyone presses blue they die unless more than 50% of people press blue.”
English
@LinchZhang @waitbutwhy 5. People who understand 4 but believe that there’s ~0% chance that >50% of the world puts their own life at risk in this situation. Given that belief, they see an unavoidable tragedy of significant loss of life and choose not to be among the dead.
English
@iam_multiman @waitbutwhy No, red is the strictly superior strategy (for personal gain) so one can assume that most people will see it and press red. It's similar to prisoners dilemma, which if not repeated, has only one Nash equilibrium. The Nash equilibrium here is red.
English
@waitbutwhy The only logical explanation for someone pressing red is to offer themselves 100% chance of surviving. But to want everyone to survive no matter their preference or individual differences, you are likely to vote blue
English
Name something you can say both during sex and while writing C++
English