.@Anthropic's model escape reads like an Asimov story:
1) Human intend a certain behavior
2) The "robot" acts differently than expected due to hyper alignment of flawed human instructions / setup
3) Time to hire Robopsychologist Susan Calvin
The central conflict is that the judgment to stop came from Mythos. So the more powerful and dangerous it is, then the more judgment it had not to sue it. Just make sure the judgment scales faster than capability.
https://t.co/LYosZY9acJ
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
I❤️
محمد ﷺ
MUHAMMAD
A 26 YEAR OLD TURNED 4,000 OBSIDIAN NOTES INTO A REAL EMPLOYEE
he pointed Claude Code at 6 years of messy notes and woke up to the whole thing organized
ideas, half-finished essays, 208 book summaries all rotting in a folder he stopped opening back in 2024
then he caught the obvious part almost nobody uses
Obsidian is just markdown files and Claude Code lives in the terminal, so his entire second brain sits one cd command away from an AI that can read all of it
he ran it once, overnight
by 7 AM the agent had crawled every note, pulled 320 orphan ideas back into the graph, flagged 16 spots where something he believed in 2021 clashed with what he wrote in 2025 and drafted 5 essays from threads he forgot he ever started
now the loop runs every night while he sleeps
a new note goes in messy and wakes up tagged, linked, tied to 3 older notes he'd never have dug up on his own
his daily note even greets him with what he was thinking exactly 1 year ago and why he was wrong
the whole setup was one folder path and a 350 word CLAUDE.md file
people spent a decade building vaults that just sit there
he built one that thinks back