NASA management estimated the chance of losing a Space Shuttle at 1 in 100,000. The engineers building it put the number closer to 1 in 100.
Same shuttle. Same institution. A 1,000x gap.
Richard Feynman exposed what the model had hidden with a rubber ring, a hardware-store clamp, and a glass of ice water.
Challenger broke apart 73 seconds after launch in 1986, killing all seven people aboard. During the investigation, Feynman compressed a piece of the shuttle’s O-ring, lowered it into ice water, then released the clamp on live television. 49:18
The rubber stayed flat.
At low temperatures, the seal could not recover fast enough to close the rocket joint. Engineers had warned about the danger. Management launched anyway.
NASA did not lack data. It had turned the absence of a previous catastrophe into evidence that catastrophe was impossible. Every successful launch made the same defect look safer.
AI can automate that mistake perfectly.
Train a model on official decisions and it learns what management approved, not what engineers tried to stop. It sees completed launches, not disasters narrowly avoided. It converts institutional confidence into mathematical confidence, then prints the result with six decimal places.
The machine can calculate a probability.
It cannot tell you who was allowed to choose it.
Feynman ended his report with one line:
“For a successful technology, reality must take precedence over public relations, for nature cannot be fooled.”
The report is free. The experiment takes seconds. The gap is still 1,000x.
Christopher Nolan tried to give Al Pacino an acting note on INSOMNIA (2002) set.
Pacino replied, “I’ve already done that. You can’t see it to the eye, but I’ve done it in the dailies.”
Nolan went back to check the footage & realized Pacino was right
🎯 I have an aversion to excess chess analogies, but England going up a goal and then pulling back against Messi instead of continuing to attack was like going up a pawn and then trying to play passive defense against Mikhail Tal!
Modular robots that repair themselves by... stealing parts! 👀
A team at Columbia University built robots made of Truss Links, stick-shaped modules with magnetic connectors at each end.
Connect a few and you get a structure that can expand, contract, roll, and reshape itself.
If a link breaks or comes loose, the robot grabs a spare module nearby and rebuilds itself. No human intervention needed.
A tetrahedron-shaped robot grabbed a spare link like a walking stick.
~~
♻️ Join the weekly robotics newsletter, and never miss any news → https://t.co/GoA3ZuwoPB