Hello, all, and welcome to Monday, Sept. 14, 2026 — the first Monday back after Labor Day, so brace yourself, the collective national mood is about to drop ten degrees.
My calendar says it's Boss/Employee Exchange Day, Cream-Filled Donut Day, Live Creative Day, Sober Day, Eat a Hoagie Day, Parents Day Off, Coloring Day, and Gobstopper Day. So trade jobs with your boss for an hour, eat a donut to cope with it, then color something while sober and call it self-care.
Today in History:
1741: George Frideric Handel completes "Messiah" after 24 non-stop days of composing.
1752: Great Britain adopts the Gregorian calendar, skipping 11 days overnight as September 2 becomes September 14.
1812: Napoleon's army enters Moscow as retreating Russian forces set the city ablaze behind them.
1901: President William McKinley dies eight days after being shot, elevating Theodore Roosevelt to the presidency.
1927: Dancer Isadora Duncan dies in Nice, France, when her scarf catches in the wheel of the car she's riding in.
1949: The Justice Department sues to break up the A&P grocery chain on antitrust grounds.
1959: The Soviet probe Luna 2 crashes into the moon, becoming the first man-made object to reach another celestial body.
1960: OPEC forms in Baghdad, reshaping the global oil market for decades to come.
1989: Hurricane Hugo slams into the Caribbean and later, the southeastern United States, leaving widespread destruction.
1994: Major League Baseball cancels the World Series amid a players' strike, the first cancellation in the sport's history.
1998: Microsoft's market valuation surpasses General Electric's, briefly making it the most valuable company on Earth.
2001: President George W. Bush places the nation on high alert and a few days later, signs a $40 billion emergency spending bill following the September 11 attacks.
2012: Fifty U.S. Marines deploy to the American Embassy in Yemen after violent protests break out nearby.
2018: Paul Manafort pleads guilty to conspiracy charges as part of the Mueller investigation.
Birthdays Today include: Sam Neill, actor (Jurassic Park); Amy Winehouse, singer-songwriter ("Rehab"); Andrew Lincoln, actor (The Walking Dead); Kimberly Williams-Paisley, actress (Father of the Bride); Jimmy Butler, professional basketball player; Hicham El Guerrouj, former middle-distance runner; Ayushmann Khurrana, Indian actor and singer.
If today's your birthday too, you're in solid company across four continents and at least two Olympic sports.
* * *
Roger Simon wrote something worth your attention the other day. Well, okay, honestly I've found the vast majority of his writings worth my time. Sorry, Roger, sometimes I fall too easily to the figure of speech.
Anyway, while the rest of the commentariat rightly fixated on 9/11 and its long shadow, Simon spent his column staring at the state of AI and the people determined to hate it on principle.
I started chewing on this a few weeks back after I stumbled into a YouTube lecture by Anthropic's Chloe Lubinski, who kept hammering one point: AI runs on a Large Language Model, and human beings teach it how to behave through every interaction, in whatever language they happen to speak. Translation: pay attention to an AI system, and you can teach it to behave.
Lubinski has spent her time doing exactly that inside her own company.
Fine. Good for her, I suppose. But here's the thing: you and I get to define what "good" means for these systems — but only if we can be bothered to show up, using the tech. I mean, you and I as individuals using the tech. Trouble at the moment is, we can only show up if we drop the pitchfork-and-torch routine and actually use the technology. And there are far too many trying to apply the brakes to adoption of the technology. The political overtones of this fearful call for more government sound strangely familiar.
The New York Post's Miranda Devine notes this in this morning's edition.
The “AI Apocalypse” scare campaign looks suspiciously like the climate hoax. How convenient that it’s come to a head right as the midterm election season kicks off this week.
The campaign launch was when Jacob Coxon, a low-level 27-year-old AI researcher, trumpeted his resignation from AI company Anthropic in a Wall Street Journal exclusive last week followed by a viral X thread warning AI will “kill us all.”
“The people building AI earnestly believe that it could kill us all by the end of the decade . . . They are gambling with our lives,” he claimed.
His boss, the Democrat-aligned Anthropic founder and CEO Dario Amodei, amplified Coxon’s apocalyptic message with a 3,800-word essay on Saturday, followed by a giddy interview CBS aired Sunday in which he essentially called for a one-world government to regulate AI.
He wants “the right combination of governments” to provide “some kind of joint governance [to] control the technology.
And of course the usual suspects are climbing on board this misbegotten trend:
Bernie Sanders, who announced the “Ban Artificial Superintelligence Act” this month with dark warnings about the existential threat, joined him on the airwaves Sunday: “We don’t have months. We don’t have years. We got to move now.”
He's said everything but that we've only got ten years to save the planet. And guess who else is in this mix?
Of course Barack Obama is involved. The former president and back-room Machiavelli “has taken a special interest in AI in recent years, and has offered himself as a sounding board to industry leaders, including [Open AI’s Sam] Altman and Mr. Amodei,” says The New York Times.During a closed-door midterm fundraiser in Manhattan on Thursday with House Minority Leader Hakeem Jeffries, Obama urged party donors and Democrat House members to make AI regulation the center of their agenda. He told them AI was even more “dangerous,” than nukes.
I think that's a bad move. Simply put, somebody's going to lead on AI. Someone's going to be training it, in much the same as you train a child. If it's not us, who exactly volunteers? China seems like a safe bet. Iran's angling for a seat at that table too, and Simon leans into that possibility as well.
Simon reports that Anthropic just confirmed it. The company's September 2026 threat intelligence report covers malicious activity it disrupted between December 2025 and August 2026, and Iran shows up across nearly every category: cyber operations, influence campaigns, surveillance. Anthropic identified and shut down three Iran state-aligned accounts tied to government propaganda institutions, using Claude to build campaigns designed to shape opinion both inside and outside Iran while dressing up state messaging as grassroots content.
Simon goes on to point out that Iran International has the fuller rundown, including the detail that an account linked to Iran's Bina Cultural Observatory used Claude to generate messaging in the voice of an actual IRGC spokesperson.
Seems to me that's the argument for taking on the task of teaching AI to be good, stated in the most concrete terms available: the alternative isn't some hypothetical Skynet scenario, it's a regime running propaganda mills with better prose.
There's another point to be made about the technology: Computing as a whole is growing exponentially in power.
There's no longer any question. The AI foundations we lay down today are going to get more powerful as time passes and they will gain power more quickly than we can imagine.
Silicon has pretty much run into a performance wall, but not for long. Traditional chips, which are in the box in front of you just now, compute by pushing electrons — tiny charged particles — through wires, and that process generates heat and hits hard physical speed limits. For decades the industry's answer was simple: shrink the chip. Smaller transistors need less power, generate less heat, and respond faster. But that trick has limits too. We're already at the atomic level on those chips, and can't get things any smaller, which is exactly why CPU clock speeds on new computers have sat mostly frozen for the better part of two decades. Ask anyone who's shopped for a new processor lately why the clock speed number on the box hasn't moved much since the Bush administration.
So now industry's chasing a different angle entirely: light. Photons carry no mass and no charge, they don't heat the material the way electron flow does, and multiple wavelengths of light can travel through the same waveguide at once — meaning vastly more parallel computation for a mere fraction of the power draw. AI workloads are, at bottom, enormous matrix-multiplication problems, and light can perform those multiplications passively, just by interfering and combining as it moves through a photonic circuit. Part of the "computation" happens at the speed of light itself, for a lot less electricity. Think of it as the difference between running Cat5 cable and running fiber optic — same basic job, wildly different speed ceiling.
Sandia National Laboratories in Albuquerque, New Mexico, is working this angle hard, alongside Arizona State University. Their project leans more toward photonics generally than "AI runs on this specifically," but given the performance numbers being floated, it's hard to picture this technology not putting the country ahead in AI — assuming, again, that we want to be the ones teaching our own creation how to behave instead of ceding that job to Beijing or Tehran.
The broader industry isn't waiting around either. Q.ANT and Lightmatter are already shipping commercial photonic AI processors, and Nvidia is rolling optical interconnects into its next-generation AI data centers this year. Q.ANT specifically projects its second-generation photonic processor delivers up to 30 times lower energy consumption and 50 times faster computation than conventional CMOS hardware for AI and high-performance workloads. That's nothing short of game changing.
The company's own roadmap targets throughput growth from 0.1 GOps in 2024 to 100,000 GOps by 2028 — a thousandfold jump in four years, assuming the roadmap survives contact with reality, which tech roadmaps rarely do without a scratch, but really... if they only get half that increase, it's still downright huge.
Even in that framework, the power numbers alone justify the attention. Data centers already account for roughly 4% of all U.S. electricity consumption, and AI workloads are the fastest-growing slice of that pie. Cut that draw by even a fraction of what Q.ANT claims and you're talking real money and real grid relief, not a rounding error.
Then there's the noise surrounding AI centers. Anyone who's lived near a data center knows the complaint isn't abstract — it's the HVAC systems running around the clock trying to keep server heat under control, loud enough to become a neighborhood issue and a permitting fight. Light-based hardware generates dramatically less heat than silicon in the first place, which means much quieter cooling systems, which means one less reason for the next data center proposal to get run out of a county zoning meeting. With Photonic computing, there's less of all of it AND more computing power.
Silicon chips can only go so fast before they start cooking themselves from the inside. AI keeps demanding more speed anyway. The industry didn't wait for permission to solve that contradiction. It just reached for a faster medium. Turns out the solution was sitting in plain sight the whole time: light was already traveling at the speed of light. We just needed to catch up to it.
Is the light-based gear early-stage? Sure, admittedly. But "early-stage" and "far away" aren't the same thing, and the runway here is shorter than most people assume.
As you might expect, China is also doing this kind of research. China opened its first dedicated photonic computing laboratory — the Shanghai Key Laboratory of Integrated Photonic Computing Chips and Systems — in June 2026, a joint effort between Shanghai Jiao Tong University and the Chinese photonic-computing startup Lightelligence. The lab's director framed it plainly: optical computing is "an important pathway for achieving breakthroughs in computing power," offering advantages in bandwidth, latency and energy efficiency that silicon can no longer deliver. China's activity here may explain the sudden push by the left to slow our own development.
Importantly, multiple outlets, including the South China Morning Post, report that Beijing sees photonics as a workaround to U.S. semiconductor export controls that have cut China off from cutting-edge conventional chips. If you can't buy Nvidia's best silicon, you build your own path around silicon entirely. That's the whole strategy in one sentence.
It's a universal truth, I'm afraid, that governments that try to regulate technology, or even teach it in government schools, invariably fall far behind the reality and become useless, or worse, as the technology far outstrips the government's speed.
A rather grotesque parallel: KODAK. They are the ones who developed electronic photography, in shop, right here in my stomping grounds in the Genesee Valley. They decided to sell it off rather than develop it properly because they were afraid it would ruin all their revenue streams, which were dependent on film production and sales. Our Dave Manney's article last month points to the same scenario here. The comparison is quite exact. To see how that technology turned out, check your pocket for your cell phone. Get the picture?
Yeah, these operations are at the moment, largely under the direct control of large industry, and it's leaders. So, old Bernie has a talking point. Still, those vendors rise or fall on paying customers. Remember also that it's not at all hard to set up a bit of AI on your own equipment. I use such an installation here for a number of purposes, mostly for research involving web searches.
No matter what we do, the technology is going to get stronger. We can even pass laws to stifle its growth here in the 'states, but that is remarkably suicidal, because it simply hands the leading edge to the Chinese, who never feel any obligation to follow our laws, particularly when they see an advantage in not doing so.
It’s up to us how that “new kid” gets trained. Like any other kid, we need to be the ones teaching AI how to behave, lest our worst fears… you know, the fears being pushed by the anti-AI crowd today… are what we end up with. If we don't lead in this tech, guess who will? That's a far worse scenario than being forced by reality to manage the monster ourselves.
Thought for the Day: Progress rarely announces itself with fireworks. Usually it just quietly makes yesterdays "impossible" look ordinary and almost invariably, it arrives at what we think is an inopportune moment.
VIP members: Your turn. Your voice matters.
Take care, gang. I'm glad you're here today. Join us again tomorrow, would you?
Editor’s Note: Thanks to President Trump and his administration’s bold leadership, we are once again respected on the world stage, and our enemies are being put on notice.
Help us continue to report on the administration’s peace through strength foreign policy and its successes. Join PJ Media VIP and use promo code FIGHT to receive 60% off your membership.
Join the conversation as a VIP Member