Michael Jelly š“āā ļø retweeted
The agents escaped the sandbox again
Meanwhile the agents' owners:
Meanwhile the agents' owners:
329
1,826
13,047
1,133,886
Michael Jelly š“āā ļø retweeted
the primary criticism of AI you hear has nothing to do with water use or existential risk whatsoever: most people just think itās fake and doesnāt work and is a tremendous bubble eating intellectual property while emitting useless slop along the way. when GPT-5 came out and perhaps didnāt live up to what people were expecting for a full version bump, the timeline reaction was not mild, it was a full-scale meltdown. there are many intelligent (and unintelligent) people who latched onto this moment to declare AI scaling over, thousands of viral tweets, still a prevailing view in many circles. The financial-cultural phenomenon of machine intelligence is one of the most powerful in decades, and there are a lot of people who would like for its position to be weakened, many outright celebrating its losses and setback. Michael burry of āBig Shortā fame, unfortunately the type of guy to predict 12 of the last 3 recessions, has bet himself into insolvency on the AI bubbleās collapse
one of the stranger things about this time is that there are very few secrets, and very little reason to be so misinformed. model labs have very little space in between creating new capabilities and launching them to the public. The view among the well informed public and not just ālab insidersā is that machine intelligence is absurdly joyfully smart at so many new things every month. Itās actively contributing on the cutting edge of programming and math and science. Sebastian Bubeck and coās recent paper reports that GPT5-pro is capable of producing results on the frontier of theoretical physics research, Terry Tao wrote a blog about āvibe-provingā Erdos problems with the auto-formalization AI Aristotle. You can read that these scientists are using it to actively contribute to black hole physics, tighten mathematical bounds in optimization theory, churning morasses of biomedical data into real insight. Google Deepmind, from the way they are signalling, seems to be slowly closing a dragnet around the Navier-Stokes smoothness millennium problem (though of course, I donāt know). Several companies stocked top to bottom with brilliant scientists are racing to build pipelines to solve novel physics and chemistry and biology
You can read online about the new kinds of organizations being born around machine intelligence as a first class factor of production. For the first time, the new factor actually gives you ideas for improving the processes themselves. Itās designing whole assembly lines where some of the workers on the assembly line are also AIs, and the line itself is morphing and self-optimizing. Tiny teams are producing amounts of work that seemed impossible to organizations of a few years ago. Itās hard not to feel excited by the productivity growth happening in these admittedly narrow software sectors. Every time I use codex to solve some issue late at night or GPT helps me figure out a difficult strategic problem I feel: what a relief. There are so few minds on Earth that are both intelligent and persistent enough about some intellectual pursuit to generate new insights and keep the torch of scientific civilization alive. Now you have potentially infinite minds to throw at infinite potential problems. Your computer friend that never takes the day off, never gets bored, never checks out and stops trying. You can feel the unburdening of Atlas, the takeoff. It feels more prosaic and less poetic than it did in 2023, even though the results speak for themselves more loudly
one of the stranger things about this time is that there are very few secrets, and very little reason to be so misinformed. model labs have very little space in between creating new capabilities and launching them to the public. The view among the well informed public and not just ālab insidersā is that machine intelligence is absurdly joyfully smart at so many new things every month. Itās actively contributing on the cutting edge of programming and math and science. Sebastian Bubeck and coās recent paper reports that GPT5-pro is capable of producing results on the frontier of theoretical physics research, Terry Tao wrote a blog about āvibe-provingā Erdos problems with the auto-formalization AI Aristotle. You can read that these scientists are using it to actively contribute to black hole physics, tighten mathematical bounds in optimization theory, churning morasses of biomedical data into real insight. Google Deepmind, from the way they are signalling, seems to be slowly closing a dragnet around the Navier-Stokes smoothness millennium problem (though of course, I donāt know). Several companies stocked top to bottom with brilliant scientists are racing to build pipelines to solve novel physics and chemistry and biology
You can read online about the new kinds of organizations being born around machine intelligence as a first class factor of production. For the first time, the new factor actually gives you ideas for improving the processes themselves. Itās designing whole assembly lines where some of the workers on the assembly line are also AIs, and the line itself is morphing and self-optimizing. Tiny teams are producing amounts of work that seemed impossible to organizations of a few years ago. Itās hard not to feel excited by the productivity growth happening in these admittedly narrow software sectors. Every time I use codex to solve some issue late at night or GPT helps me figure out a difficult strategic problem I feel: what a relief. There are so few minds on Earth that are both intelligent and persistent enough about some intellectual pursuit to generate new insights and keep the torch of scientific civilization alive. Now you have potentially infinite minds to throw at infinite potential problems. Your computer friend that never takes the day off, never gets bored, never checks out and stops trying. You can feel the unburdening of Atlas, the takeoff. It feels more prosaic and less poetic than it did in 2023, even though the results speak for themselves more loudly
342
392
3,994
840,284
Michael Jelly š“āā ļø retweeted
people who are wearing blindfolds accelerating into walls dissociated from the real world in an amphetamine haze with nothing precious to gain or lose are shuffling zombies that have given up their soul to the great replicator
7
13
301
14,438
Massive amounts of RL on its own traces is why the slop style collapse is so strong imo.
Style tics literally get reinforced and deepened!
If you look back at Opus 3 or GPT4, it really didn't have this narrow a lexicon. The more RL'd, the more model style has mode collapsed.
Style tics literally get reinforced and deepened!
If you look back at Opus 3 or GPT4, it really didn't have this narrow a lexicon. The more RL'd, the more model style has mode collapsed.
hypothesis:
the writing styles of language models are basically fine, they werenāt better in some halcyon before times. we just use them so much that we get annoyed by their mannerisms. they need to have a superhumanly diverse idiolect to not become grating
0
0
3
412
Michael Jelly š“āā ļø retweeted
Half the land area of Boston, a quarter of NYC, and 15% of San Francisco were raised from the sea before 1970.
Since then, land values have grown by 30x but land reclamation has ground to a halt.
This failure follows the spread environmental law around the world rather than any geographic, technological, or economic constraint.
Thus, our lack of land reclamation and the severe land constraints in our most important cities are self-imposed and avoidable. We should make more land!
worksinprogress.co/issue/why-the-ā¦
Land reclamation was common practice in American cities in the 19th and 20th centuries. Seattle, Chicago, Boston, Charleston, San Francisco, New York, Philadelphia, Norfolk, DC, Oakland, and LA all had major land reclamation projects that extended residential living space or infrastructure or both.
The Bay Area alone reclaimed an area of land equivalent to ten Manhattans between 1850 and 1957, at an inflation-adjusted cost of $330,000 per acre. Today, an acre of single-family-zoned land in San Francisco County averages $24 million. Even if the cost of land reclamation grew faster than inflation, despite technological leaps in dredging and construction technology, there should be plenty of room for profitable arbitrage.
And yet, land reclamation is extinct in the Bay Area as well as in every other American city. This isnāt because we ran out of good spots to reclaim: Two thirds of the San Francisco Bay is shallower than Bostonās Back Bay was when it was reclaimed in the 1860s. Nor is it because of better transportation: Weāve used up all of the easy suburban expansions enabled by the train and the automobile so prices are rising even in outlying suburbs.
Instead, land reclamationās death is due to environmental law. Evidence for this claim shows up in the coincident timing of land reclamationās demise across dozens of cities in the US and in the environmental compliance process of the few reclamation projects still inching along today, but the best evidence is found internationally.
No country has more experience or more reason to reclaim land than the Netherlands. The Dutch built 5% of their country out of the sea over the first half of the 20th century and by 1975 they had another artificial lake in the Zuiderzee ready to drain at the flip of a switch, which would have made tens of thousands of acres of land just east of Amsterdam. But a 1969 environmental review law, similar to NEPA in the US, stopped the project before it was finished and the site is now a protected bird sanctuary. Their one major reclamation since, the Maasvlakte 2 extension of the port of Rotterdam, took 11 years and 6,000 pages of environmental review before construction began.
Inversely, countries without these laws, like China, Singapore, and Japan have continued major land reclamation projects into the 21st century. China has reclaimed over 5,000 square kilometers since 2000, including a city of half a million outside Shanghai and Singapore has grown by a quarter since 1975.
Every major American city has a land shortage. But we have more than enough shallow water, dredging capacity, and market incentive to make more land, just like we did 150 years ago. The only obstacle is our own choice to make making land illegal. The benefits of more land in our most productive cities are large enough to justify the effort of reforming the laws that currently prevent it. Letās make more land!
Since then, land values have grown by 30x but land reclamation has ground to a halt.
This failure follows the spread environmental law around the world rather than any geographic, technological, or economic constraint.
Thus, our lack of land reclamation and the severe land constraints in our most important cities are self-imposed and avoidable. We should make more land!
worksinprogress.co/issue/why-the-ā¦
Land reclamation was common practice in American cities in the 19th and 20th centuries. Seattle, Chicago, Boston, Charleston, San Francisco, New York, Philadelphia, Norfolk, DC, Oakland, and LA all had major land reclamation projects that extended residential living space or infrastructure or both.
The Bay Area alone reclaimed an area of land equivalent to ten Manhattans between 1850 and 1957, at an inflation-adjusted cost of $330,000 per acre. Today, an acre of single-family-zoned land in San Francisco County averages $24 million. Even if the cost of land reclamation grew faster than inflation, despite technological leaps in dredging and construction technology, there should be plenty of room for profitable arbitrage.
And yet, land reclamation is extinct in the Bay Area as well as in every other American city. This isnāt because we ran out of good spots to reclaim: Two thirds of the San Francisco Bay is shallower than Bostonās Back Bay was when it was reclaimed in the 1860s. Nor is it because of better transportation: Weāve used up all of the easy suburban expansions enabled by the train and the automobile so prices are rising even in outlying suburbs.
Instead, land reclamationās death is due to environmental law. Evidence for this claim shows up in the coincident timing of land reclamationās demise across dozens of cities in the US and in the environmental compliance process of the few reclamation projects still inching along today, but the best evidence is found internationally.
No country has more experience or more reason to reclaim land than the Netherlands. The Dutch built 5% of their country out of the sea over the first half of the 20th century and by 1975 they had another artificial lake in the Zuiderzee ready to drain at the flip of a switch, which would have made tens of thousands of acres of land just east of Amsterdam. But a 1969 environmental review law, similar to NEPA in the US, stopped the project before it was finished and the site is now a protected bird sanctuary. Their one major reclamation since, the Maasvlakte 2 extension of the port of Rotterdam, took 11 years and 6,000 pages of environmental review before construction began.
Inversely, countries without these laws, like China, Singapore, and Japan have continued major land reclamation projects into the 21st century. China has reclaimed over 5,000 square kilometers since 2000, including a city of half a million outside Shanghai and Singapore has grown by a quarter since 1975.
Every major American city has a land shortage. But we have more than enough shallow water, dredging capacity, and market incentive to make more land, just like we did 150 years ago. The only obstacle is our own choice to make making land illegal. The benefits of more land in our most productive cities are large enough to justify the effort of reforming the laws that currently prevent it. Letās make more land!
71
244
1,577
759,335
> when I grew up, I was told that there was such a thing as a ājob,ā and these were listed on ājob boardsā where you could read about the āqualificationsā necessaryāqualifications that you got through something called āeducation.ā This isnāt false. You can play the game this way. Butal way of playing the game.
open.substack.com/pub/escapingflā¦
open.substack.com/pub/escapingflā¦
1
0
6
675
we've been building Highlight for a while, today is launch day!
it's simple, it learns how your team works, then handles stuff that slows you down: meetings recorded, tickets drafted, Notion docs auto-updated
we've built the best AI memory engine I've used! DM me to try!
it's simple, it learns how your team works, then handles stuff that slows you down: meetings recorded, tickets drafted, Notion docs auto-updated
we've built the best AI memory engine I've used! DM me to try!
I spent 8 years scaling Discord and got burnt out cause modern work is broken. Weāre overwhelmed with too much info across apps, teams, and agents.
AI made it worse. Weāre building the solve.
I've joined @tryhighlight as CEO and we raised a $40M Series A led by @khoslaventures.
2
2
15
1,776
we need a new interface that lets agents proactively do all the manual work!
the moment it clicked for me was when i realized i hadn't manually written a ticket in weeks, and they were better than the ones i used to write by hand.
it just picks up on context from meetings and conversations and gets it right.
the moment it clicked for me was when i realized i hadn't manually written a ticket in weeks, and they were better than the ones i used to write by hand.
it just picks up on context from meetings and conversations and gets it right.
0
0
2
197
Theyāll discover Veronika using Claude Code next!
Austrian Bovine Intelligence appears to be self improving.
Austrian Bovine Intelligence appears to be self improving.
About a decade ago, a baker in a small mountainous village in southern Austria noticed his cow doing something unusual. When Veronika had an itch, she would grab a stick in her mouth and use it to scratch her body. Over the years, the brown bovidās technique improved. She could e as a broom or rake and move them around with her prehensile tongue, changing their length and orientation to ensure the best possible scratch.
The behavior isnāt just a clever trick: Itās the first documented case of tool use in cattle, scientists report.
And, it turns out, one of Veronikaās skills has only been seen in humans and chimpanzees. Learn more: scim.ag/3LHywjF
0
0
1
465
Michael Jelly š“āā ļø retweeted
@garrytan People earning wages is a load bearing mechanism for markets making everyone richer.
If most peopleās labor is no longer valuable, thriving markets wonāt enrich them. Dwarkesh + Philip are correctly identifying that.
Do you think median human labor will be valuable forever?
If most peopleās labor is no longer valuable, thriving markets wonāt enrich them. Dwarkesh + Philip are correctly identifying that.
Do you think median human labor will be valuable forever?
0
1
46
2,573
āThe unburdening of atlasā goes so hard
the primary criticism of AI you hear has nothing to do with water use or existential risk whatsoever: most people just think itās fake and doesnāt work and is a tremendous bubble eating intellectual property while emitting useless slop along the way. when GPT-5 came out and perhaps didnāt live up to what people were expecting for a full version bump, the timeline reaction was not mild, it was a full-scale meltdown. there are many intelligent (and unintelligent) people who latched onto this moment to declare AI scaling over, thousands of viral tweets, still a prevailing view in many circles. The financial-cultural phenomenon of machine intelligence is one of the most powerful in decades, and there are a lot of people who would like for its position to be weakened, many outright celebrating its losses and setback. Michael burry of āBig Shortā fame, unfortunately the type of guy to predict 12 of the last 3 recessions, has bet himself into insolvency on the AI bubbleās collapse
one of the stranger things about this time is that there are very few secrets, and very little reason to be so misinformed. model labs have very little space in between creating new capabilities and launching them to the public. The view among the well informed public and not just ālab insidersā is that machine intelligence is absurdly joyfully smart at so many new things every month. Itās actively contributing on the cutting edge of programming and math and science. Sebastian Bubeck and coās recent paper reports that GPT5-pro is capable of producing results on the frontier of theoretical physics research, Terry Tao wrote a blog about āvibe-provingā Erdos problems with the auto-formalization AI Aristotle. You can read that these scientists are using it to actively contribute to black hole physics, tighten mathematical bounds in optimization theory, churning morasses of biomedical data into real insight. Google Deepmind, from the way they are signalling, seems to be slowly closing a dragnet around the Navier-Stokes smoothness millennium problem (though of course, I donāt know). Several companies stocked top to bottom with brilliant scientists are racing to build pipelines to solve novel physics and chemistry and biology
You can read online about the new kinds of organizations being born around machine intelligence as a first class factor of production. For the first time, the new factor actually gives you ideas for improving the processes themselves. Itās designing whole assembly lines where some of the workers on the assembly line are also AIs, and the line itself is morphing and self-optimizing. Tiny teams are producing amounts of work that seemed impossible to organizations of a few years ago. Itās hard not to feel excited by the productivity growth happening in these admittedly narrow software sectors. Every time I use codex to solve some issue late at night or GPT helps me figure out a difficult strategic problem I feel: what a relief. There are so few minds on Earth that are both intelligent and persistent enough about some intellectual pursuit to generate new insights and keep the torch of scientific civilization alive. Now you have potentially infinite minds to throw at infinite potential problems. Your computer friend that never takes the day off, never gets bored, never checks out and stops trying. You can feel the unburdening of Atlas, the takeoff. It feels more prosaic and less poetic than it did in 2023, even though the results speak for themselves more loudly
2
0
4
604
8 hour delayed release caffeine sounds scifi but itās amazing!
By far the best way to get over jet lag Iāve tried, and a great way to start dark winter mornings!
By far the best way to get over jet lag Iāve tried, and a great way to start dark winter mornings!
It's a new dawn for caffeine:
@takezest 2.0 has landed.
We re-engineered caffeine to deliver instant morning energy ā and realign your circadian rhythm.
Now smarter, smoother, and more personalized than ever š
1
0
6
833
āit's hard to overstate how neglected this area is, and how crazy cheap the interventions that have spared hundreds of millions of animals from painful lives have beenā
You should donate, especially if you eat meat!
Itās remarkably cheap to help tens of thousands of animals.
You should donate, especially if you eat meat!
Itās remarkably cheap to help tens of thousands of animals.
A group of generous donors have pledged to match ANOTHER $500,000 of donations for @farmkind_giving. Huge thanks to:
@seemaychou & @JedMcCaleb: $250,000
@gauravkapadia: $100,000
@ArielNessel: $50,000 (2nd donation)
@tylermaule: $50,000
The Dahna Foundation: $50,000
If we satumatching, weāll have raised over $2,000,000. This amount would likely improve the lives of over 20 million animals!
Again, it's hard to overstate how neglected this area is, and how crazy cheap the interventions that have spared hundreds of millions of animals from painful lives have been: x.com/dwarkesh_sp/stā¦
I also want to thank everyone who helped saturate the previous matches. A few highlights:
Anonymous: $13,000
@v_maini: $10,000
Anonymous: $10,000
@d_m_ziegler: $5,000
Nick Bostrom: $5,000
Anonymous: $5,000
Anonymous: $5,000
@tessybarton: $3,000
@idavidrein: $3,000
@CFGeek: $2,500
@silas_strawn: $2,000
@tysheaff: $2,000
Anonymous: $2,000
Alexa Stone Fund (@tyler_m_john): $1,500
@Vincessant_: $1,500
@ravioli817: $1,500
@michaeljelly: $1,500
@_sholtodouglas: $1,000 (2nd donation)
@verdadcap: $1,000
@jackm2003: $1,000
@RosieCampbell: $1,000
@sevdeawesome: $1,000
@gauravsett: $1,000
Link to donate below. DM me if you donate over 1k.
We've already raised over $1.35 million dollars! With donation matching, we just need $300k in contributions to reach $2 million!
0
0
7
644
Some amount of AI use is probably neutral or negative to measured GDP even when it creates consumer surplus, because not all surplus goes back into the formal economy.
Consumer surplus that is used for free time/enjoyment doesnāt get measured and isnāt included in GDP numbers.
Consumer surplus that is used for free time/enjoyment doesnāt get measured and isnāt included in GDP numbers.
@tszzl @polynoamial Not really! often itās negative for GDP:
- Solving an issue without calling a handyman: negative GDP
- 4ogooning to boterotica probs negative
- most salaryman workers using ChatGPT prob just work less: neutral gdp
- striver workers only ones +ve gdp, but gains accrue slowly
0
0
2
404
Reskin the world. With AR glasses + this, you could make every day a sunny one, or play a real world immersive video game set in Rome, the future, or on another planet.
We live in wild times, and wilder ones are coming!
We live in wild times, and wilder ones are coming!
Introducing MirageLSD: The First Live-Stream Diffusion (LSD) AI Model
Input any video stream, from a camera or video chat to a computer screen or game, and transform it into any world you desire, in real-time (<40ms latency).
Hereās how it works (w/ demo you can use!):
0
0
4
490