Kavish Garg

7 followers · 8 following · 554 views

on the atlas · 68

highlights · 244

Unlike chips and large language models, there is no “good enough beats best” when it comes to lithography. This is a process of many moving parts. Precision is measured in atoms; tolerance is infinitesimally lower than it would be for, say, electric vehicles. Constant tweaking requires feedback, which ASML gets from shipping hundreds of data-generating machines every year. China cannot yet match that.
ASML’s hold over the future of chipmaking is weakening
There is no longer evidence that rising government debt generates political pressure for public spending cuts or tax increases.
Whatever happened to prudence?
The result could be a “Jevons paradox”: less analytical effort per decision, but more contested proposals to be considered and authorized, and so more work for consultants and managers.
What do consultants get paid for? - by Luis Garicano
What do consultants get paid for?
What do consultants get paid for? - by Luis Garicano
The group plans to release its second-generation Vision-Language-Action (VLA) software in Europe from 2027
Xpeng says China close to building ‘killer’ rival to Tesla Model Y
an eternal fast-follower will never have sufficient internal coding agents to keep up with compounding gains at the frontier.
The Moonshot - by Anton Leicht - Threading the Needle
Is Marxian economics the only economic lens one can apply to literature? What would a Hayekian literary criticism look like? The place to start is the great Paul Cantor’s pioneering essay on Thomas Mann’s “Disorder and Early Sorrow,” a slight-seeming story set in Weimar Germany during the hyperinflation. Cantor shows that when one reads the novella through Hayek and Mises rather than Marx, the story opens up.
Marginal Revolution
However, modernity is about tiring its opponents by persisting with it.
The Only Indian City Where Life Has Improved
There are anecdotal signs that a large fraction of current token spend is for new work that wasn’t previously paid for rather than replacing existing work.
AI Dark Output: The Visible Cost of Invisible Output
Additionally, it should be noted that Alphabet, through a subsidiary, owns the original makers and primary maintainers of ROS/ROS2.
The Final Offshoring
Because it leased the access from CoreWeave and lots of other providers, Andromeda also didn’t take on the liability and depreciation risks of building and maintaining its hardware or own data centers.
AI Compute Startup Andromeda Raises $60M At $1.5B Valuation
An additional demand: when superintelligence is developed, it’s designed to refuse to crush middle powers. By doing this, the US would credibly bind itself to maintaining the sovereignty of other nations.
LessWrong
The only alternative that makes sense to me is siding with China.
LessWrong
Huge investments mean that the group’s return on capital is modest, at less than 8% a year over the past decade, well below its cost of capital.
Can Mukesh Ambani pull off his biggest gamble yet?
China has urged its companies to adopt ai, but not to lay off workers.
Prepare for an AI jobs apocalypse
The government recently halved its economic-growth projection for 2026 to just 0.5%, owing, in part, to the blockage of the Strait of Hormuz, but also the sluggish pace of domestic reforms.
Friedrich Merz can’t go on like this
However, Nasdaq has loosened its rules as it battled to win the SpaceX listing over its rival NYSE, allowing the stock to join the Nasdaq 100 after just 15 days.
‘Fast entry’ SpaceX, OpenAI and Anthropic IPOs to ignite Wall Street trading frenzy
Even the contrast between old-money nonprofits and venture-backed startups can be recast in terms of the classic Protestant schism over whether to give primacy to charity or vocation.
Think Tanks Have Defeated Democracy
Venture capital funding, with its expectation that most startups will fail, has even begun to resemble a form of philanthropy with West Coast characteristics.
Think Tanks Have Defeated Democracy
In 2016, 1.44% of U.S. GDP was donated to nonprofit organizations, more than any other country and nearly twice as much as New Zealand, the runner up.
Think Tanks Have Defeated Democracy
The opposite has happened. Birinci, Karabarbounis, and See in a 2026 NBER paper show that about half of the American-European hours gap that existed in the 1990s has reversed by the end of the 2010s. Americans work fewer hours per person than they did in 2000, while most Europeans work more.
Luis Garicano 🇪🇺🇺🇦 on X: "We stopped everything to write an answer (link below) to Paul Krugman's two posts of today (one informal, one with a simple model) arguing that Europe is broadly not falling behind the United States. The change measured by the Draghi report, he argues, is mostly due to growth" / X
When the government auctioned a bond in March maturing in 2028, after Mr. Milei’s term ends, investors demanded almost 350 basis points more than a comparable bond maturing before his term ends one year earlier.
Why Argentina’s Comeback Has Stalled - WSJ
he portion of AI6 that was originally assigned to TSMC will reportedly be transferred entirely to Intel Foundry.
Jukan on X: "Interesting. Tesla’s current AI5 chip, codenamed Helios, is reportedly being foundried by TSMC and Samsung, and the next-generation AI6 was also originally expected to be split between those two foundries. However, under strong pressure and encouragement from the Trump https://t.co/ceYAyyB9Va" / X
The origins of Indians
Ancient India was shaped by waves of migration and mixing | Aeon Essays
Policy action is already underway. Full-stack exports feature prominently in the Action Plan and in statements by senior White House staffers alike.
Making AI Export Promotion Work - by Anton Leicht
Now this is easier to do in some cases than others – unless you’re foolish enough to sign away mining rights to raw materials for cheap, raw materials can’t really be reshored to the US –, but it’s a latent risk that will be salient to the involved powers.
Forging the Pax Silica - by Anton Leicht
his is exactly the kind of thing middle power strategists are worried about: if the US mixes international tech policy with its political desiderata, how can we trust the merits of strategic deals?
Forging the Pax Silica - by Anton Leicht
Selling the Pax Silica to someone who was never a US treaty ally before could be the easiest thing in the world, and pitching it to long-standing US allies now would still feel like a downgrade and incur resistance accordingly.
Forging the Pax Silica - by Anton Leicht
One way to understand this, then, is that ‘Pax Silica’ might be the Republican word for ‘allied scale’.
Forging the Pax Silica - by Anton Leicht
Upstream inputs into AI capability, like semiconductor manufacturing and high-quality data, are still very strong outside the US;
Forging the Pax Silica - by Anton Leicht
Instead, it asks them to do what they’re already good at, and closer integrate it into the US alliance.
Forging the Pax Silica - by Anton Leicht
A reduction of alliances to mutual leverage, then, makes the bandwagoning strategy easier, not harder – because it insures the arrangement against unilateral reneging, just as long as middle powers keep up their part of the bargain
Forging the Pax Silica - by Anton Leicht
Between regulatory constraints, electricity grid shortcomings, high energy prices, and possible political backlash against AI, it’s conceivable that Western democracies have enough GPUs, but lack the datacenter infrastructure to run them. An ideal AI oil power faces no such issue – gulf autocracies can wave away political backlash and regulatory oversight, and usually do not lack energy supply
Datacenter Delusions - by Anton Leicht
d some reasons to think you’d need less, e.g. the lack of stockpiling and thus more immediate economic effects of supply changes. Almost no middle power is on track for that position.
Datacenter Delusions - by Anton Leicht
Third, in an increasingly bipolar AI paradigm, you might be able to turn your coat as a nuclear measure. Presumably, most AI oil powers would be dependent on compute imports from a great power. Today that’s mostly from the US, in the future it might also be from China.
Datacenter Delusions - by Anton Leicht
Middle powers engaged in compute buildout do not seem intent on answering these questions – and so I have trouble believing that frontier training is the threshold they’re trying to cross.
Datacenter Delusions - by Anton Leicht
Again, this is most relevant where you want your AI’s decision to be intransparent and unpredictable from the outside, such as in military or broader security applications, or in cases where you want to extensively train on proprietary and confidential data, such as from consumers or from manufacturing processes.
Datacenter Delusions - by Anton Leicht
If you expect most of your inference needs to be serviced through running closed frontier models, you want secure datacenters so that these models will be exported to you without fear of theft.
Datacenter Delusions - by Anton Leicht
ither way, building out a full inference stack is a lasting commitment in need of expensive upkeep, and by no means a necessary imperative. It’ll also almost necessarily be much more costly and economically restrictive than simply importing the intelligence from elsewhere. All in all, a bit like reaching full energy sovereignty today.
Datacenter Delusions - by Anton Leicht
This tracks a general global demand for sovereign supply of vital economic goods: you don’t want to be reliant on intelligence imports just as much as you don’t want to be too reliant on energy imports.
Datacenter Delusions - by Anton Leicht
This piece looks at viable strategic thresholds for compute buildout – and finds that many enthusiastic compute buyers are not on track to deploy and leverage the resources they’re attempting to amass. Merely controlling some compute does not yet make for controlling the universe.
Datacenter Delusions - by Anton Leicht
Datacenter Delusions Simply building compute capacity does not make for a convincing national AI strategy.
Datacenter Delusions - by Anton Leicht
I can’t speak to political dynamics everywhere to the same extent, but by and large I think this sort of plan has the big advantage of a favourable political economy. In short: All this does not require a visible pivot to a tech- or AI-driven economic model.
A Roadmap For AI Middle Powers - by Anton Leicht
Or maybe you won’t win decisively. Maybe the three months of AGI-advantage don’t ramp it up enough. Would you bet the house on that if you were sitting on the NSC?
A Roadmap For AI Middle Powers - by Anton Leicht
Winning, in that sense, just means that your economy and military gets a lot more bang for its buck due to some distribution of efficiency gains.
A Roadmap For AI Middle Powers - by Anton Leicht
When getting ready for increasing AI capabilities, you want to be ready to turn cheap intelligence into valuable growth.
A Roadmap For AI Middle Powers - by Anton Leicht
These state-run projects might not be too keen on sharing — recent policy writing from the US that second-guesses proliferation even through the Five Eyes intelligence alliance should send shivers down many middle powers’ spines.
A Roadmap For AI Middle Powers - by Anton Leicht
in high-frequency financial trading, microseconds of latency advantage justify enormous infrastructure investment; at the frontier of AI development itself, companies treat even marginal capability differences as competitively decisive.
Import Imperatives - by Anton Leicht - Threading the Needle
The optimal unit of frontier progress is a focused, talent-dense team with access to enormous compute.
The Need for an Independent AI Grid - AMP PBC
Bessent has pitched the idea of expanding the list of countries that have permanent swap lines with the US as a way to lock in dollar supremacy.
Swap Lines Explained: How the US Extends Dollar Power Globally - Bloomberg