@sabeer Pointing out the symptoms is endless and has been going on for centuries, just quote numbers from this Shadow Economy report and keep asking to 50x number of judges, everything will work out. Loading PDFโฆ https://t.co/y4JqurFaZC
HOW LONG VITAMINS STAY IN YOUR BODY:
1. Vitamin C โ 24 hours
2. Vitamin B12 โ Several months
3. Vitamin D โ Weeks to months
4. Vitamin A โ Stored in liver for months
5. Vitamin K โ 1โ2 days
6. Folate (B9) โ Few days
7. Vitamin E โ Weeks
8. Biotin โ 24โ48 hours
9. Calcium โ Hours (needs daily intake)
10. Magnesium โ 24โ48 hours
The power of emergent knowledge
For 50 years, linguists believed your brain runs a hidden grammar engineโunconsciously applying thousands of rules to generate sentences.
LLMs just proved this is completely wrong.
And the implications go way beyond language.
The old theory made sense. Language is structured. It has patterns. Where there are patterns, there must be rules generating them, right?
So we tried to formalize everything. The entire Semantic Web was built on this assumption: encode the rules, achieve understanding.
Then LLMs showed up.
They've never been taught a single grammar rule. They just... read a lot. Trillions of sentences.
And they write better than most humans.
This isn't an engineering trick. It's a discovery about what language fundamentally is.
Patterns don't require rules to generate them.
The rules were always just our description of patterns that already existed.
Where else have we made this mistake?
Where else have we confused "how we describe it" with "how it has to work"?
Consider Python dependency management.
For 20 years, millions of developers accepted it was inherently painful. Virtual environments were clunky, but that was "the state of the art."
Then uv appeared in 2024. One tool. Suddenly everything just worked.
The "rules" of how dependency management should work weren't requirements.
They were descriptions of one bad approach that everyone had collectively accepted as necessary.
NASA has twice SpaceX's budget.
SpaceX has launched more missions in the last decade than NASA did in fifty years.
Same physics. Same goal. Radically different relationship to "process."
NASA follows the rules:
Multi-year planning cycles
Formal review boards
Institutional accountability chains
SpaceX breaks things fast, learns fast.
The process was supposed to ensure quality. It became a substitute for it.
Something works through emergent patterns
We describe those patterns as "rules"
We forget they were descriptions
We enforce the rules as requirements
The rules become obstacles to the thing they were describing
Grammar rules don't generate language. Grammar rules describe where language already goes.
A river doesn't follow the map. The map describes the river.
The author I'm riffing on (Daniel Lemire @lemire) makes a provocative claim:
"The bureaucratisation of science was a death sentence for progress."
Strong. But the argument structure is hard to dismiss.
We created formal scientific processes to describe how good science happens.
Then we forgot these were descriptions and started treating them as requirements.
Now the process has become the point.
The tired excuse: "All the low-hanging fruit has been picked."
What if the fruit is right there, but we've convinced ourselves we need a 47-step approval process before we can reach for it?
Darwin needed a ship before he could develop evolution.
Einstein's relativity emerged when trains made relative motion obvious.
The neat storyโprofessors think deep thoughts, engineers build stuffโis historically backwards.
Tools enable thoughts. Thoughts enable tools.
It's a loop, not a line.
LLMs aren't "just engineering." They revealed something true about the world.
So what do you do with this?
Next time you encounter a "rule" or "process," ask:
Is this a requirement? Or a description we've been enforcing?
Notice the difference between:
"This is how it has to work"
"This is how we've been describing it"
Most "has to" is actually "we assumed."
The deepest version of this insight:
What LLMs reveal about language might apply to institutions too.
Neither requires explicit rules. Both emerge from patterns of practice. Both calcify when we forget that.
I'm still working out the limits of this.
Does it apply to law? Medicine? Engineering safety standards?
Some rules probably are requirements. The question is which ones.
But I suspect we've overcounted badly.
There's a specific threshold of complexity and self-direction below which a system degenerates, and above which it can open-endedly self-improve.
Current AI systems aren't close to it yet. But it's inevitable we will reach this point eventually. When we do, we won't see a sudden explosion, more like consistently self-sustaining linear-ish progress. Like the pace of Science itself (which itself clearly a self-improving system)
The most powerful scientific instrument of the 21st century isn't the electron microscope or the particle collider. It's the algorithm.
Today, a scientist in biology, physics, chemistry etc. is more likely to be debugging a Python script than to be running a wet lab. Going forward, the biggest breakthroughs will be mostly software achievements, like AlphaFold.
IMO gold medalists are 50x more likely to win a Fields medal, the highest prize in Mathematics, than PhDs from top 10 schools.
The Power Law is everywhere for those with eyes to see.
Nobel Peace Prize Winners (1911โ2025)๐
1911: ๐ณ๐ฑ Tobias Asser, ๐ฆ๐น Alfred Fried
1912: ๐บ๐ธ Elihu Root
1913: ๐ง๐ช Henri La Fontaine
1917: ๐จ๐ญ International Committee of the Red Cross
1919: ๐บ๐ธ Woodrow Wilson
1920: ๐ซ๐ท Lรฉon Bourgeois
1921: ๐ธ๐ช Hjalmar Branting, ๐ณ๐ด Christian Lange
1922: ๐ณ๐ด Fridtjof Nansen
1925: ๐ฌ๐ง Sir Austen Chamberlain, ๐บ๐ธ Charles G. Dawes
1926: ๐ซ๐ท Aristide Briand, ๐ฉ๐ช Gustav Stresemann
1927: ๐ซ๐ท Ferdinand Buisson, ๐ฉ๐ช Ludwig Quidde
1929: ๐บ๐ธ Frank B. Kellogg
1930: ๐ธ๐ช Nathan Sรถderblom
1931: ๐บ๐ธ Jane Addams, ๐บ๐ธ Nicholas Murray Butler
1933: ๐ฌ๐ง Sir Norman Angell
1934: ๐ฌ๐ง Arthur Henderson
1935: ๐ฉ๐ช Carl von Ossietzky
1936: ๐ฆ๐ท Carlos Saavedra Lamas
1937: ๐ฌ๐ง Robert Cecil
1938: ๐ณ๐ด Nansen International Office for Refugees
1944: ๐จ๐ญ International Committee of the Red Cross
1945: ๐บ๐ธ Cordell Hull
1946: ๐บ๐ธ Emily Greene Balch, ๐บ๐ธ John R. Mott
1947: ๐ฌ๐ง Friends Service Council, ๐บ๐ธ American Friends Service Committee
1949: ๐ฌ๐ง Lord Boyd Orr
1950: ๐บ๐ธ Ralph Bunche
1951: ๐ซ๐ท Lรฉon Jouhaux
1952: ๐ฆ๐น Albert Schweitzer
1953: ๐บ๐ธ George C. Marshall
1954: ๐บ๐ณ Office of the United Nations High Commissioner for Refugees
1957: ๐จ๐ฆ Lester Bowles Pearson
1958: ๐ง๐ช Georges Pire
1959: ๐ฌ๐ง Philip Noel-Baker
1960: ๐ฟ๐ฆ Albert Lutuli
1961: ๐ธ๐ช Dag Hammarskjรถld
1962: ๐บ๐ธ Linus Pauling
1963: ๐จ๐ญ International Committee of the Red Cross, League of Red Cross Societies
1964: ๐บ๐ธ Martin Luther King Jr.
1965: ๐บ๐ณ United Nations Childrenโs Fund
1968: ๐ซ๐ท Renรฉ Cassin
1969: ๐บ๐ณ International Labour Organization
1970: ๐บ๐ธ Norman Borlaug
1971: ๐ฉ๐ช Willy Brandt
1973: ๐บ๐ธ Henry Kissinger, ๐ป๐ณ Lรช ฤแปฉc Thแป (declined)
1974: ๐ฎ๐ช Seรกn MacBride, ๐ฏ๐ต Eisaku Satล
1975: ๐ท๐บ Andrei Sakharov
1976: ๐ฎ๐ช Betty Williams, ๐ฎ๐ช Mairead Corrigan
1977: ๐บ๐ณ Amnesty International
1978: ๐ช๐ฌ Anwar alโSadat, ๐ฎ๐ฑ Menachem Begin
1979: ๐ฎ๐ณ Mother Teresa
1980: ๐ฆ๐ท Adolfo Pรฉrez Esquivel
1981: ๐บ๐ณ Office of the United Nations High Commissioner for Refugees
1982: ๐ธ๐ช Alva Myrdal, ๐ฒ๐ฝ Alfonso Garcรญa Robles
1983: ๐ต๐ฑ Lech Waลฤsa
1984: ๐ฟ๐ฆ Desmond Tutu
1985: ๐บ๐ณ International Physicians for the Prevention of Nuclear War
1986: ๐บ๐ธ Elie Wiesel
1987: ๐จ๐ท รscar Arias Sรกnchez
1988: ๐บ๐ณ United Nations Peacekeeping Forces
1989: ๐น๐ท The 14th Dalai Lama
1990: ๐ท๐บ Mikhail Gorbachev
1991: ๐ฒ๐ฒ Aung San Suu Kyi
1992: ๐ฌ๐น Rigoberta Menchรบ Tum
1993: ๐ฟ๐ฆ Nelson Mandela, ๐ฟ๐ฆ F.W. de Klerk
1994: ๐ต๐ธ Yasser Arafat, ๐ฎ๐ฑ Shimon Peres, ๐ฎ๐ฑ Yitzhak Rabin
1995: ๐ฌ๐ง Joseph Rotblat, ๐ต๐ฌ Pugwash Conferences on Science and World Affairs
1996: ๐น๐ฑ Carlos Filipe Ximenes Belo, ๐ต๐ท Josรฉ RamosโHorta
1997: ๐บ๐ธ International Campaign to Ban Landmines, ๐บ๐ธ Jody Williams
1998: ๐ฎ๐ช John Hume, ๐ฎ๐ช David Trimble
1999: ๐บ๐ณ Doctors Without Borders
2000: ๐ฐ๐ท Kim Daeโjung
2001: ๐บ๐ณ United Nations, ๐ฌ๐ญ Kofi Annan
2002: ๐บ๐ธ Jimmy Carter
2003: ๐ฎ๐ท Shirin Ebadi
2004: ๐ฐ๐ช Wangari Maathai
2005: ๐ฆ๐น International Atomic Energy Agency, ๐ช๐ฌ Mohamed ElBaradei
2006: ๐ง๐ฉ Muhammad Yunus, ๐ง๐ฉ Grameen Bank
2007: ๐บ๐ณ Intergovernmental Panel on Climate Change, ๐บ๐ธ Al Gore
2008: ๐ซ๐ฎ Martti Ahtisaari
2009: ๐บ๐ธ Barack H. Obama
2010: ๐จ๐ณ Liu Xiaobo
2011: ๐ฑ๐ท Ellen Johnson Sirleaf, ๐ฑ๐พ Leymah Gbowee, ๐พ๐ช Tawakkol Karman
2012: ๐ช๐บ European Union
2013: ๐ณ๐ฑ Organisation for the Prohibition of Chemical Weapons
2014: ๐ฎ๐ณ Kailash Satyarthi, ๐ต๐ฐ Malala Yousafzai
2015: ๐น๐ณ National Dialogue Quartet
2016: ๐จ๐ด Juan Manuel Santos
2017: ๐ฆ๐บ International Campaign to Abolish Nuclear Weapons
2018: ๐จ๐ฉ Denis Mukwege, ๐ฎ๐ถ Nadia Murad
2019: ๐ช๐น Abiy Ahmed Ali
2020: ๐บ๐ณ World Food Programme
2021: ๐ต๐ญ Maria Ressa, ๐ท๐บ Dmitry Muratov
2022: ๐ง๐พ Ales Bialiatski, ๐ท๐บ Memorial, ๐บ๐ฆ Center for Civil Liberties
2023: ๐ฎ๐ท Narges Mohammadi
2024: ๐ฏ๐ต Nihon Hidankyo
2025: ๐ป๐ช Maria Corina Machado
Source: NobelPrize Organization
To understand human nature, read the older books. To develop specific knowledge, stay on the bleeding edge, read newer (technical) books.
The best authors - Deutsch, Schopenhauer, Borges, Ted Chiang - write with very high density. The best authors respect the readerโs time.
When a model gives you the right answer to a reasoning question, you can't tell whether it was via memorization or via reasoning.
A simple way to tell between the two is to tweak your question in a way that 1. changes the answer, 2. requires some reasoning to adapt to the change. If you still get the same answer as before... it was memorization.
message to all ~7,000 people on tech twitter:
THIS IS A TINY BUBBLE
NORMIES DON'T KNOW WHAT CLAUDE IS AND HALF HAVE NOT TRIED CHATGPT
TRY TO GET BILLIONS OF VIEWS, NOT MILLIONS
SINCE 2022; CULTURE SPREADS THROUGH SHORT FORM. NOTHING ELSE
A VIRAL X LAUNCH GETS 1 MIL VIEWS ONE TIME AND DIES
THERE ARE EXTREME POWER LAWS IN ATTENTION LIKE NEVER BEFORE
ENOUGH ATTENTION WILL MAKE U THE PRESIDENT OF USA
EVERY AWARENESS EXPERIMENT SHOULD BE POTENTIALLY CAPABLE OF 1 BILLION VIEWS