Ryan Morris retweeted
Visitor: "Where's the fruit? Gorillas eat fruit."
Keeper: "Not here. We took it off the menu in 2014. Most big zoos have since."
Visitor: "Why would you take fruit off a gorilla?"
Keeper: "Because it isn't fruit any more. Everything in a supermarket's been bred for a century to be sweeter. A wild fig's about as sweet as a carrot. A supermarket banana's a pudding."
Visitor: "And the gorillas?"
Keeper: "Got fat. Teeth went. Two of the males were pre-diabetic. The vet said feeding them modern fruit was like feeding them chocolate bars with the wrapper left on."
Visitor: "What do they get now?"
Keeper: "Leaves, stems, a bit of veg. They sulked for a fortnight and lost the weight."
Visitor: "So the fruit was too sugary for a 180-kilo animal built to eat it."
Keeper: "That's the paperwork, yes."
Visitor: "I've got a bag of grapes in the buggy for my two-year-old."
Keeper: "Lovely. We'd not be allowed to give those to the chimps."
Ryan Morris retweeted
I grew up in a magical era when the leading man of the leading TV show at the time solved hard problems with intelligence, courage, honour and respect. I owe so much to his example, and continue to try to live by the example his character set.
Ryan Morris retweeted
Cavemen Issues Dire Warning That The Wheel Is Going To Eliminate Countless Jobs
Ryan Morris retweeted
This is the least surprising graph in the world if you understand how gradient descent works.
The upside of extremism
I just found the most brilliant bit by @JohnCleese
He explains Leftism better in 90 seconds, than you could learn from any academic in 20 years.
Market 'shape' from a market-neutral perspective: calculate residual returns (net of many risk-factors, mo, value, etc) - calc how much is explained by top 10 PCAs. Lower = more drivers, higher = fewer... other than March 2020, never been fewer drivers of idio returns
2
97
Ryan Morris retweeted
Trump Pledges That If Voters Defeat Socialism In November, They Will Each Get An Equal Redistribution Of Wealth From The Government buff.ly/kRXF5x9
Ryan Morris retweeted
OK China is officially the most cyberpunk country on Earth. Japan, America, Russia, and the rest have all been cyberpunkmogged.
Ryan Morris retweeted
The accounting mismatch in capex booms highlighted. It’s material in the tech space right now.
Ryan Morris retweeted
“In my youth I made it a rule not to drink a drop of alcohol before lunch. Now that I am no longer young, I keep to the rule of not drinking a drop before breakfast.”
Winston Churchill was the only British Prime Minister ever awarded the Nobel Prize in Literature. Reading him, you realize that to be a brilliant writer you don’t need to write long books — and to be a true philosopher, you don’t either.
Here are some of his quotes:
One. If you’re going through hell — keep going.
Two. You have enemies? Good. That means you’ve stood up for something in your life.
Three. Every crisis is a new opportunity.
Four. A smart person doesn’t make all the mistakes himself — he gives others a chance too.
Five. The best argument against democracy is a five-minute conversation with the average voter.
Six. Success is the ability to go from one failure to another without losing enthusiasm.
Seven. A kite rises highest against the wind, not with it.
Eight. A man who never changes his mind is a fool.
Nine. When eagles are silent, parrots begin to chatter.
Ten. Power is a drug. Anyone who has tried it even once is poisoned by it forever.
Eleven. Do not wish for health and wealth — wish for luck. On the Titanic, everyone was rich and healthy, but only a few were lucky.
Twelve. A lie travels halfway around the world before the truth has a chance to get its pants on.
Thirteen. Politics is almost as exciting as war, and quite as dangerous. In war, you can only be killed once — in politics, many times.
Fourteen. My tastes are simple. I am easily satisfied with the best.
Fifteen. If you want to have the last word in an argument, tell your opponent: “You may be right.”
Sixteen. The greatest advantage comes to those who make their mistakes early enough to learn from them.
Seventeen. People are very good at keeping secrets they do not know.
Eighteen. War is when innocent people die for the interests of others.
Nineteen. The greatest lesson in life is that even fools are sometimes right.
Twenty. It is far better to bribe a person than to kill him — and far better to be bribed than killed.
Twenty-one. It is easier to rule a nation than to raise four children.
Twenty-two. We live in an age of great events and small men.
Twenty-three. Nothing earns authority like calmness.
Twenty-four. The best way to ruin a relationship is to start trying to “sort it out.”
Twenty-five. When two people fight, the third one wins.
Twenty-six. If you kill a murderer, the number of murderers does not change.
Twenty-seven. A pessimist sees difficulty in every opportunity; an optimist sees opportunity in every difficulty.
Twenty-eight. You will never reach your destination if you stop to throw stones at every barking dog.
Twenty-nine. A nation that forgets its past loses its future.
Once, during a speech, Churchill was asked:
— Isn’t it pleasing to know that every time you give a speech, the hall is packed?
He replied:
— It is, very much so. But every time I see a full hall, I cannot help thinking that if I were not giving a speech but being led to the gallows, the crowd would be twice as large.
Ryan Morris retweeted
Singapore’s Foreign Minister on why he cannot accept negotiating with Iran for safe passage of ships. Definitely worth listening to:
Ryan Morris retweeted
And then the RL bubble burst, saving trillions of dollars.
Simply adding Gaussian noise to LLMs (one step—no iterations, no learning rate, no gradients) and ensembling them can achieve performance comparable to or even better than standard GRPO/PPO on math reasoning, coding, writing, and chemistry tasks. We call this algorithm RandOpt.
To verify that this is not limited to specific models, we tested it on Qwen, Llama, OLMo3, and VLMs.
What's behind this? We find that in the Gaussian search neighborhood around pretrained LLMs, diverse task experts are densely distributed — a regime we term Neural Thickets.
Paper: arxiv.org/pdf/2603.12228
Code: github.com/sunrainyg/RandOpt
Website: thickets.mit.edu
































