How’s France doing?

by philg

We haven’t heard too much from France lately, other than in the context of ball-kicking matches. The country is out of the World Cup, but how is France doing more generally?

A photo of the French World Cup team doesn’t suggest economic greatness. The players wore jerseys with the logos of an American company (Uber) and a Hitler-founded German company (VW):

Trump-hating neural network guru Yann LeCun was born in France, but has worked “primarily at Bell Labs, New York University and Meta Platforms, Inc.” (Wikipedia). He shifted back towards his homeland in late 2025 with Advanced Machine Intelligence Labs, which claims to be doing everything better than the American and Chinese LLM companies, but doesn’t seem to have anything to show yet. Can a country where it is illegal to work more than 35 hours per week compete in the only arena that matters anymore? Given France’s prowess at building nuclear power plants, one might imagine that the country would be a natural habitat for data centers. Unfortunately, France seems to have lost its mojo in nuclear power, with recent projects subject to massive time and cost overruns (China is the superstar now, according to the geniuses at Harvard). At the rate that France and the US are going, it would be cheaper to melt Earth with CO2 and buy a new planet than to stay on the current nuclear plant construction trajectory.

The French don’t have anything like Dutch ASML that can pick up some crumbs from the NVIDIA/TSMC table.

Who wants to predict the next significant innovation from a culture that was one of the world’s most innovative? Or maybe that culture doesn’t exist anymore because the legacy French are being bred out of existence? Foreign-born women in France have a total fertility of 3, far above the 1.56 average for all of France that includes the immigrants and children of recent immigrants (with some statistical magic, France officially says that the TFR for immigrant women is only 2.37 despite the standard calculation coming up with 3). The most successful residents of France, as measured by genetics, are those from Africa outside of the Magreb (TFR over 3) and those from North Africa (TFR just under 3). In the long run, therefore, we’d expect France to be as innovative as Algeria, Côte d’Ivoire, and Congo have been.

Separately, how about some music to celebrate our brothers, sisters, and binary-resisters across the Atlantic? In order to honor Brigitte Macron, a strong independent woman not afraid to smack a wimpy man whenever necessary, I’ll pick Louise Farrenc (1804-1875): “In the 1830s, Farrenc gained considerable fame as a performer and her reputation was such that in 1842 she was appointed to the permanent position of Professor of Piano at the Paris Conservatory, a position she held for thirty years and one which was among the most prestigious in Europe. Farrenc was the only woman to hold the esteemed position and rank at the Paris Conservatory throughout the 19th century.”

Loosely related, who enjoyed the World Cup this year? Never having played soccer, I’m not in love with watching. It feels random and chaotic. I don’t like the flopping or the shirt-pulling (an NFL player who did either would be called a “sissy”?).

Only slightly related to France, but definitely related to Florida (Lionel Messi featured):

It’s great to have Jeff Bezos, Larry Ellison, Ken Griffin, Larry Page, et al., as taxpaying neighbors, but I wonder how many Floridians would give all of them back to their respective Lands of the Righteous in order to keep Messi.

Straight Outta Fort Lauderdale…

Finally, an EU perspective:

Full post, including comments

How’s your National Data Center Day of Protest going?

by philg

Today in Evanston, Illinois: “Say No to AI Data Centers”.

A lot of retailers display this sign, but my favorite is GadgetWorks, whose smartphone and game console customers are in no way dependent on data centers:

The Guardian says that the dream would be to find a data center that is used by ICE and involved in demanding IDs from voters:

Speaking of ID, although everyone in Evanston seems to agree that requiring ID to vote is unreasonable, the city requires ID of those who wish to visit the local (Lake Michigan) beach:

Full post, including comments

What Would Barnes Wallis (WWBW) do with Iran?

by philg

As the world’s aviation nerds gathering for EAA AirVenture (“Oshkosh”; July 20-26), let’s think a bit about the impact of aviation on human conflict…

The Dambusters Raid, one of the most famous uses of aircraft during World War II, was the brainchild of Barnes Wallis, an aeronautical engineer who sought a clean-sheet-of-paper answer to the question, “What’s the most effective way to reduce Germany’s ability to fight?” Sir Arthur Harris‘s plan was simple: kill as many German civilians as possible via firebombing of cities because a factory whose workers have been killed isn’t going to be productive (and the productivity of workers who survived would be reduced since they’d be homeless). Wallis settled on the idea of destroying hydroelectric dams in Germany’s industrial heartland, thus simultaneously denying German factories the electricity, coal, and water necessary to achieve full production. He looked at the inputs to German manufacturing, especially energy, and tried to destroy those inputs so that it wouldn’t matter how effective Germany was at dispersing factories.

The vulnerability of a dam seems to have been understood by both the Germans and the British, i.e., an explosion right up against the interior wall at a time when the dam was nearly full. As such, the Germans had strung torpedo nets in their reservoirs to prevent traditional torpedoes from being used. Barnes Wallis came up with the idea of a bomb that would skip along the surface of the lake impounded by the dam, roll down the water-side dam wall, and explode. A tremendous number of engineering problems needed to be solved to make this a reality. Paul Brickhill’s classic book is excellent, especially since the author was himself a WWII military pilot, but I recently listened to a 2013 book by James Holland that adds much interesting detail and makes it clear that Barnes Wallis wasn’t quite the lone inventor that we meet in Brickhill’s earlier work.

Arthur Harris was initially skeptical, to say the least: “This is tripe of the wildest description. There are so many ifs and buts that there is not the smallest chance of it working.” A combination of engineering persistence and incredible bravery and sacrifice by 617 Squadron, however, proved that Wallis had been on the right track. Revisionist historians later minimized the damage done to Nazi Germany by the 19 planes sent out on May 16, 1943 by noting that the Germans successfully scrambled to rebuild the dams, as well as many washed-out bridges and factories, before winter. James Holland points out in his book that forcing the Germans to divert tens of thousands of workers and perhaps $10 billion adjusted to post-Biden dollars. The breached dams had terrible consequences for German arms production just when they needed to support their campaigns against the Russians and against the Allies in Italy. The breached dams also took labor and money away from beefing up the Atlantic Wall against the invasion that actually happened in 1944.

Another passion for Barnes Wallis was precision bombing with 10-ton bombs that he designed. These pioneering “bunker busters” were used to good effect against various German installations, including for submarines and V-2 rockets, as well as to sink the battleship Tirpitz. Any time that we’re bombing fortified infrastructure in Iran, therefore, we’re channeling the spirit of Wallis.

What Would Barnes Wallis (WWBW) do with Iran? More or less the opposite of what the U.S. has been doing. Wallis would almost surely have recommended taking out the Iranian oil industry, a critical energy input to everything else that happens in Iran, and the industrial-scale electric power generation currently available to the Islamic Republic to run any factories that it still has and/or rebuilds. Wallis probably would also try to destroy Iran’s ability to move energy and materials, e.g., by taking out bridges, railroads, airports, etc. At a minimum, this would keep the Islamic Republic busy rebuilding and people who are busy rebuilding can’t put their energy into weapons production or military facility construction.

Maybe the U.S. strategy of slowly surrending to Iran will, in the long run, prove to have more merit. I hope so!

Full post, including comments

Why does birth tourism to middle class foreigners bother Americans, but millions of children of millions of TPS foreigners don’t concern anyone?

by philg

A post on X about birth tourism that turned middle class Hondurans into Americans was exciting enough to get over 750,000 views:

Stories about Chinese birth tourism “rings” also reliably generate interest/outrage. Anyone with enough money to participate in a birth tourism scheme is from at least a middle class family in a foreign country and, therefore, is unlikely to become the founder of a multi-generational welfare dynasty. (Statistically, those who grow up in self-sufficient households are likely to be adults in self-sufficient households.)

Meanwhile, the US citizens being birthed by foreigners here “temporarily” under Temporary Protected Status, some since 1990, aren’t of interest. Neither Gemini nor ChatGPT could even begin to estimate how many foreigners, total, have spent time in the U.S. under TPS. Partly because of that, neither AI could begin to estimate how many children they’ve had, each one with a U.S. passport. The best that the AIs can do is come up with 1.3 million foreign TPS holders in the U.S. right now and possibly 390,000 U.S. citizen children living with them. Gemini estimates that the majority of TPS households with children are on what used to be called “welfare” (“safety-net participation”), but even AI probably can’t see into all of the state-run programs for which immigration status isn’t relevant, e.g., health care in California, public housing in Maskachusetts, etc. Our AI overlords say that the majority of children who in welfare-eligible households will grow up to adults living in welfare-eligible households (i.e., multi-generational welfare is the expectation, though at just over 50 percent). These stats are somewhat complicated by non-welfare welfare programs, e.g., the various below-market-rate rent schemes in New York City that cover more than half of rental units there.

During the entire time that the above-described Honduras-to-US pipeline was running, there was a 26-year “temporary” program to allow some favored subset of Hondurans to live in the U.S. There were over 100,000 Hondurans granted TPS during the program’s launch in 1999. Gemini says that almost no Honduran ever went back before Donald Trump ended their status and that almost none went home after the September 2025 termination either:

The termination of TPS did not grant Immigration and Customs Enforcement (ICE) the right to summarily deport people. To be removed from the country, individuals who do not leave voluntarily must be issued a Notice to Appear (NTA) and go through the formal immigration court system.

With immigration courts already facing massive multi-million case backlogs, processing thousands of long-term residents with clean criminal histories and significant family equities in the U.S. takes years, effectively leaving them in a prolonged state of legal limbo within the United States.

So a handful of pregnant Hondurans with the family resources (maybe from U.S. relatives) to pay for roundtrip airfare to Boston generates outrage due to the resulting 10+ U.S.-born anchor babies. Not a single one of these kids has ever been on welfare, to my knowledge, unless we consider the taxpayer-funded hospital payments to be welfare. But perhaps 100,000 births to Hondurans here under TPS, nearly all of which were similarly paid for by taxpayers and then resulted in welfare eligibility for at least 18 years, aren’t interesting.

Full post, including comments

Claude is “marginally the best, but quite expensive”

by philg

From a friend who uses LLMs to write code…

Current state of affairs according to me:

OpenAI is at the frontier. There is no reason to use Terra, as there is a Luna or Sol model at an effort level that matches it for cheaper. You can get quite a lot of inference out of the $20/mo plan. $100/mo gives you 5x, but $200/mo gives you 20x, not 10.
Claude Fable is marginally the best, but quite expensive and falls back to Opus on anything it considers remotely questionable, while still charging Fable prices.

Gemini’s latest is 3.5-flash and is well off the curve. The latest non-flash is still 3.1-pro.

GLM 5.2 is the best open weight Chinese model, but its price per token is misleading as it eats up reasoning tokens like crazy.

Best local coding model is probably Qwen 3.6 27B. Gemma 4 is a good local all around model to talk to, but not as good at coding.

Where will these various models run? New York State has banned data centers, which is forward-thinking but I prefer to reflect on the exquisite timing of the New York political elite in shutting down their massive nuclear electricity plant just one year before the AI/data center boom began (launch of ChatGPT in 2022):

How about AI data centers in space?

Loosely related:

Full post, including comments

Don’t go outside, says the New York Times

by philg

You’ll die from the Canadian and Minnesotan (Somali?) smoke if you go outside in the north and from “dangerous” heat if you go outside in the south, says the NYT (front page):

The smoke is due to “climate change”, according to Canada’s Prime Minister (NYT).

How bad is it? aviationweather.gov:

METAR KORD 162051Z 05010KT 1SM R10L/P6000FT FU VV020 32/18 A3003
METAR KTOL 162052Z 00000KT 1SM R25/4500V5000FT FU BKN018 OVC024 31/1857007
SPECI KCLE 162109Z 36007KT 2SM FU SCT000 BKN026 BKN250 28/09 A3003
METAR KBUF 162054Z 23011KT 2 1/2SM FU SCT031 BKN065 OVC200 26/18 A2996 
METAR KTEB 162051Z 26008G14KT 1 1/2SM FU BKN020 OVC033 30/18 A298

Chicago is down to 1 statute mile of visibility (IMC begins at 3 SM so everyone has to be on an instrument flight plan) in smoke (“FU”, which sounds rude but is actually French (pre-Islamic)). Notice that Chicago and Teterboro, NJ are both at toasty temps of 32C and 30C, respectively.

Toledo, Ohio is also at 1 SM. Things get better moving east to Cleveland (2 miles) and Buffalo (2.5 miles), but get worse as one approaches the Mamdani Caliphate (1.5 miles visibility in smoke).

What about if Lionel Messi is home in Fort Lauderdale celebrating his recent victory over an African/Islamic nation? It’s “good”:

Same site for Chicago:

Let’s check the aviation weather for SE Florida, a hellscape of dangerous heat and unavailable abortion care, according to the NYT:

METAR KDJT 162053Z 22006KT 10SM FEW030 BKN250 33/18 A3005 RMK AO2 SLP175 T03330178 58016 $
METAR KFLL 162053Z 13007KT 10SM FEW030 SCT150 BKN190 BKN250 31/24 A3005 RMK AO2 SLP175 T03110244 56013
METAR KMIA 162053Z 12008KT 10SM SCT046 SCT055 SCT220 34/24 A3005 RMK AO2 TSE28 SLP177 CB DSNT SW T03390244 57006 $

Why only 10 miles of viz at President Donald J. Trump International Airport? (formerly PBI) That’s as high as the automated systems go. Note that Messi’s home in Fort Lauderdale is actually one degree C cooler than Chicago!

If this all clears before EAA AirVenture starts in Oshkosh, Wisconsin on Monday I will personally support Greta Thunberg for multiple Nobel Prizes (physics, chemistry, and peace?).

Full post, including comments

Federal government rewrites Disney’s Circle of Life (SNAP then Ozempic via Medicare)

by philg

CBS:

Because federal tax dollars are free, of course, the CBS story covers what Medicare beneficiaries will pay ($50/month) and doesn’t mention what it will cost the young taxpayers who are being hit to pay for keeping Boomers alive forever (ChatGPT says $6 billion/year would be easy to hit since the drug companies are getting a lot more than $50/month).

The federal government starting paying for a lot of the calories that made today’s older Americans fat when it created the food stamp program, now “SNAP”, in the 1960s. By paying for weight loss drugs, therefore, we have a complete SNAP-to-GLP-1 circle of life. What about rewriting the Disney classic song to fit the current situation? ChatGPT refuses to do this, claiming that it would be a violation of Disney’s copyright to create some new lyrics for their existing melody. Grok apparently doesn’t care.

Circle of Life (SNAP Edition)

(Zulu-style chant, sung with dramatic flair)
Nants ingonyama bagithi Baba
Sithi uhmm… SNAP-onyama!

Here comes the benefit, father
Oh yes, it’s the EBT card
A feast we’re gonna fund here
A burger, a soda, a leopard… wait, that’s just the spots on the fries

Verse 1
From the day we arrive on the planet
And blinking, step into the store
There’s more to eat than can ever be eaten
More to buy than can ever be bought
There’s far too much junk food to take in here
More calories than can ever be burned
But the government check rolling high
Through the welfare sky
Keeps the great and small on the endless round

Chorus
It’s the Circle of Life
And it moves us all
Taxpayers fund the snacks
Then the drugs for the sprawl
Through obesity and hope
Through waistlines and scope
‘Til we find our place
On the Ozempic path unwinding
In the Circle
The Circle of Life

Verse 2
Some of us eat the surplus cheese
While others pay the mounting fees
We are all connected in the great chain
From the SNAP aisle to the pharmacy lane
The taxpayer’s wallet gives what it can
To the processed foods and the injection plan
It’s the wheel of fortune, it’s the leap of fate
It’s the government gate… that swings both ways

Chorus
It’s the Circle of Life
And it moves us all
Through despair and hope
Through faith and love… and rebates
‘Til we find our place
On the path unwinding
In the Circle
The Circle of Life

Bridge/Outro
Wegovy rising like the morning sun
GLP-1 flowing when the damage is done
From the grocery cart to the needle’s prick
The federal cycle does the trick
Nants ingonyama… SNAP-onyama!
It’s the Circle… of Life!

Full post, including comments

100th birthday of the man who made the X-ray diffraction image of crystallized DNA that helped Watson and Crick

by philg

As part of James Watson‘s continued cancelation, it is conventional to credit Rosalind Franklin with having figured out the structure of DNA from an X-ray diffraction image that she made prior to dying of cancer. In fact, Photo 51 was created by Raymond Gosling, born exactly 100 years ago today.

Wall Street Journal, “Celebrating the Forgotten Man of the DNA Quest” (2012):

But it was [Gosling], supervised by Maurice Wilkins, who in 1950 first took an X-ray photograph of a DNA fiber, which he had wound around a bent paper clip. Of that moment, he writes: “Those discrete diffraction spots emerging on the film in the developing dish was a truly eureka moment. Maurice and I drank several glasses of his sherry.”

It was this photo that electrified Dr. Watson when he saw it presented by Dr. Wilkins at a conference in Naples, Italy. If, as the image implied, the gene had a regular structure, then elucidating that structure might explain its properties. And this quest brought Dr. Watson to (then) Mr. Crick, whom he persuaded to moonlight from the study of proteins and help him tackle DNA.

The WSJ describes Photo 51 as having been a collaborative effort between Franklin and Gosling, so the attribution solely to Rosalind Franklin isn’t completely baseless. ChatGPT: “Franklin was expert at getting DNA fibers into a state suitable for X-ray crystallography, especially controlling humidity so DNA formed the “B form,” the hydrated form that produced Photo 51.”

A standard example of de-Watsonification, “8 Female Scientists Whose Discoveries Were Credited to Men” (Biography.com):

Born in 1920 in London, Rosalind Franklin was a chemist, X-ray crystallographer and leading molecular biologist who discovered the structure of DNA.

Note that in the article about nerds not being properly credited for their nerdism, Raymond Gosling isn’t credited.

Full post, including comments

Quantifying the understatement of inflation due to the Owners’ Equivalent Rent fantasy (Wall Street Journal data)

by philg

The cost of owning a house in the U.S. can go up 1,000% and it would have no effect on the government’s official inflation rate, updated this morning to 3.5 percent. That’s because the cost of an owned house is included in CPI by asking homeowners, who aren’t landlords, how much they think they would get if they became landlords (usually not much because nobody wants to move into a single-family house with a one-year lease and then get kicked out at the end becuase the owner decided to do something else with the property). BLS:

The expenditure weight in the CPI market basket for OER is based on the following question that the CE Survey asks of consumers who own their primary residence: “If someone were to rent your home today, how much do you think it would rent for monthly, unfurnished and without utilities?”

ChatGPT says that OER rose by 31% from year-end 2019 to year-end 2025 (citing FRED).

What happened to actual costs? Nobody knows because the government doesn’t bother to measure them (see above). The Wall Street Journal, however, has a recent article on an attempt to quantify the rise in costs. It’s 39% rather than the BLS fantasy number of 31% if one considers someone who bought a house in 2025 vs. in 2019:

OER is about 26% of CPI so the effect of using OER instead of actual costs has been to understate inflation by about 0.3% per year over the 2019-2025 period coverd. That doesn’t sound like a lot, but it is about one-sixth of the Fed’s 2% inflation target (how are brightest technocrats doing on that, by the way? https://www.nytimes.com/2026/07/14/business/economy/inflation-cpi.html has a chart showing six straight years of failure to meet this target; six years is “transitory”?).

Related:

Full post, including comments

New York Times summarizes Lindsey Graham’s 71-year life as “Trump ally” while reminding us that the true American heroes aren’t, strictly speaking, American

by philg

Today’s New York Times…

A Mexican national whom we are informed was shot at random by ICE (they were looking for an undocument migrant and shot the unfortunate Mr. Lorenzo Salgado Araujo for no reason and, as it happened, Mr. Araujo was also an undocumented migrant (what are the odds in a country where we’re informed that only a tiny percentage of residents are undocumented migrants?)):

Lindsey Graham, whose entire life can be reduced to, simply, “Trump Ally”:

Democrat connections on Facebook are filling up their feeds with glee regarding Senator Graham’s death. Some examples:

“I’ve never wished a man dead, but I have read some obituaries with great pleasure.”
— Clarence Darrow, “The Story of My Life”, 1932

According to Dante, Lindsay’s final destination would be Antenora.

X users who previously expressed support for the Gazans and hatred of Israel are also happy. Example from someone affiliated with the American-Arab Anti-Discrimination Committee:

The mere existence of an American-Arab Anti-Discrimination Committee seems like something that evolved during Lindsey Graham’s adulthood (until Hart-Celler in 1965, Arabs generally weren’t eligible to immigrate to the U.S.). Let’s look at what else has changed since Lindsey Graham was born in 1955.

Lindsey Graham served in the USAF for 33 years, including in Iraq and Afghanistan, and, unusually, did not become disabled by his work at a desk. VA Benefits cost $4.5 billion in 1955. Adjusted for inflation, that’s about 55 billion 2026 mini-dollars. That’s during a period where nearly 100% of American men were veterans of three brutal wars (WWI, WWII, and Korean). The population was smaller because we hadn’t discovered the miracle of open borders and, thus, the U.S. was home to only about 22 million living veterans in 1955. Today, VA benefits, not including military pensions earned via length of service, cost about $440 billion. There are only about 16 million living veterans and a high enough percentage of them are classified as “disabled”, thus forcing taxpayers to pay out more than $200 billion in disability compensation, e.g., to Graham Platner in Maine. The average American who actually pays any significant federal income tax will work approximately 2-3 hours per week to fund payments and services to veterans, despite the fact that it is tough to think of any war that we’ve won since Korea. ChatGPT: “real federal spending per veteran has risen by substantially more than a factor of eight”.

(Note that we will soon be spending more on disability payments to veterans than the Chinese spend on their entire active-duty military. The majority of disabled veterans were never in combat so we can’t blame our enemies becoming tougher for this epidemic of disability. A majority of veterans who left the U.S. military during the Biden-Harris administration were classified as disabled. There are some tough-to-quantify costs of these disability ratings. Veterans who receive TDIU (Total Disability based on Individual Unemployability) can’t work without risking the loss of the taxpayer cash. Veterans who are rated 100 percent disabled, such as Graham Platner, pay no property tax on a primary residence in Florida (imagine if this fearless fighter were to move into Ron DeSantis terrirtory!).)

Lindsey Graham started out wanted to give nearly every undocumented migrant in the U.S. “amnesty” and a passport. The percentage of immigrants in the U.S. has risen from about 5 percent when Lindsey Graham was born to at least 15 percent today (Pew). If we were to look at “immigrants plus children of recent immigrants” the growth would be much more dramatic due to the U.S. having closed its borders in 1924. If we corrected Pew’s numbers with the 2X likely population of undocumented migrants that Yale found, the percentage of the U.S. population that was foreign-born would likely be around 20 percent (and maybe 30 percent for immigrants plus children of recent immigrants since immigrants have higher fertility than the native-born Americans whom they are definitely not “replacing”).

Not in any way related to this enrichment via immigration, the cost of an existing single-family house (i.e., not one in some exurb) relative to median income, is now 2-3X what it was in 1955, relative to median male worker’s earnings, depending on how one adjusts for square footage.

Graham was, like all members of Congress, responsible for setting federal spending levels. Federal debt was about 65 percent of GDP in 1955, not too long after massive borrowing to fund WWII. Today it is 125 percent of GDP. Federal spending has grown from 16.5 percent of GDP to about 23 percent (don’t forget state and local as well!). ChatGPT: “The decline in military spending since 1955 was more than fully absorbed by the growth of Social Security, Medicare, Medicaid, veterans’ benefits, other transfers, and interest on the debt.”

Loosely related…

Full post, including comments