In Thrive Capital’s first-ever investor letter, founder Joshua Kushner has some unexpected things to say about his venture capital rivals on the West Coast.

“It is difficult to overstate the magnitude of the opportunity,” Kushner wrote about AI in the letter, leaked to Bloomberg. “It would also be a grave error in our minds to let excitement weaken our investment discipline. … Within Silicon Valley in particular, the industry can become fixated on hyperincremental technological turns rather than where the technology ultimately leads.”

While his secretive New York-based firm, just like those in Silicon Valley, is betting heavily on AI, Thrive is doing so differently, he argues. There’s no so-called spray-and-pray investing. Thrive tends to go big on the companies it backs. Bloomberg estimates about 90% of its capital is poured into the top 15 investments in each fund.

That makes Thrive, he contends, a company of independent thinkers. “We are independent because markets move between fear and enthusiasm, and neither is a substitute for judgment.”

His comments are in direct contrast to one of the basic premises of Silicon Valley venture capital: that it is a business of “outliers” as espoused by Marc Andreessen.

In the “outlier” view, a VC firm makes a lot of bets, prepared to lose money on many — even most — of them. The few big hits will be so lucrative that they will cover the losers and much, much more. That philosophy leaves VCs forever looking for the next OpenAI or another mega hit. It can also lead to, as we saw during the post-pandemic lean years, cutting ongoing support for startups not deemed to be on track to be the biggest winners.

In contrast, Kushner writes, “We believed an investment firm could be opportunistic across stage, sector, and geography, while remaining deeply concentrated in a small number of people and ideas.” The idea is to “build Thrive to concentrate our time, capital, and energy on the people and ideas we believe in most.”

He also dismisses Silicon Valley’s idea that VCs are in the business of disrupting incumbents.

“Unlike many of our peers, our conviction was not only that these industries would be disrupted from the outside in but also that many would be transformed from the inside out,” he wrote about AI’s impact.

Thrive has largely stuck to this thesis. Its deepening relationship with OpenAI is its biggest example. The VC firm is a major investor in the AI lab. But in December 2025, the roles switched when OpenAI took an ownership stake in Thrive Holdings, the VC firm’s spinout. Thrive Holdings buys companies and then works with OpenAI to give them an AI makeover. Part of the deal involved OpenAI dedicating employees to work with Thrive’s companies.

Thrive Holdings has bought more than 70 businesses and has a team of 35 engineers. Kushner says its accounting platform uses agents to produce tax returns 30% faster with 98% accuracy, and its IT services firm has agents independently solving half of its help desk tickets.

Still, Thrive’s strategy is working in part because it nabbed stakes in some of the industry’s best-performing startups ever. Its $516 million 2022 early-stage fund, for instance, made early bets on OpenAI, Anduril, and SpaceX, and is now worth more than $3.7 billion as of the end of June, Bloomberg reports. Thrive has, over its 15 years, increased its stakes in all of them (and also had a sizeable stake in Cursor, which just closed its sale to SpaceX).

It has also backed Wiz, Ramp, and Stripe, to name a few other big names. Plus, it has led seed investments in new labs like Essential AI, founded by former Google Brain researcher Ashish Vaswani, the lead writer of the famed “Transformers” paper that spawned today’s AI industry.

All told, Thrive has $60 billion of assets under management, Kushner revealed in the letter. He reports impressive profits: a gross internal rate of return (IRR) across all funds of 41% and a net IRR of 33%. Thrive has returned more than $1 billion of liquidity to its investors in the last 12 months alone, he said.

“There may be an opportunity for billions of dollars in additional liquidity in the coming quarters,” he promises.

He doesn’t specify which companies are headed for their exits, but obviously the SpaceX IPO was a start, and OpenAI is working toward its own public debut.

It should be pointed out that both Kushner’s and Andreessen’s approaches obviously work in terms of making money. Andreessen Horowitz returned $25 billion to its investors between 2009 and 2025, according to the last leaked returns, reported by Eric Newcomer.

Thrive’s philosophy of concentrating capital may not even be possible for most smaller, scrappy emerging seed funds, whose founders weren’t born into the kind of access that the son of a billionaire New York real-estate family has.

That said, Kushner’s general premise of how overheated Silicon Valley’s AI investing has become isn’t wrong either. As he puts it: “Not every fast-growing business is exceptional. And not every exceptional company is a great investment at every price. Our responsibility is to maintain those distinctions.”

security

Corma CEO tells The Reg it’s building ‘One ring to rule them all, for the defenders to have this power’

Corma CEO Alon Pluda says his AI security startup aims to close the “defense gap,” where models are better at offensive security. He tells the story of one customer, a security executive who was walking his dog when he received a notification on his watch from a Corma agent.

“It said, ‘I just caught a live attack. I need your permission to block it,’” Pluda told The Register in an interview. 

The security boss approved the agent’s action; the agent blocked the malware and the attacker from moving across the company’s network and mitigated the intrusion in under 10 minutes, Pluda said.

The customer later described “walking outside with his dog, and blocking a real-live attack with his AI coworker” as “one of the most magical moments of his year,” Pluda recalled.

Pluda founded Corma about a year ago. And yes, all you Lord of the Rings nerds, the company gets its name from the Elven word for “ring.” “We’re building the one ring to rule them all, but this time for the defenders to have this power.”

Earlier this week, the company announced $60 million in seed funding led by Sequoia Capital, alongside Khosla Ventures and Coatue. He told us that his startup is working with Fortune 100 companies, and training models to achieve “superintelligence for defensive cybersecurity.”

Models from OpenAI, Anthropic, and Google are “amazingly good” at coding and language, and this includes finding and fixing bugs, and orchestrating tools across multi-step workflows, he explained. 

“When you combine it with agentic capabilities, they move from being incredible vulnerability researchers to end-to-end attackers,” Pluda said. “So inherently, what we’ve seen in the last few months is the models getting exponentially better at offensive security, like we saw with the OpenAI and Hugging Face incident.” 

But these same models aren’t as skilled at carrying out defensive security tasks that don’t involve scanning code for vulnerabilities and misconfigurations, he said. “The vast majority of defensive security tasks don’t have anything to do with code.”

Corma recently tested four frontier models – Claude Opus 4.8, GPT-5.5, Grok 4.3, and DeepSeek V4 – as both attackers and defenders across the same fake company and its networks, built to closely mirror a multi-business enterprise. The attacker’s task was to plant a backdoor and the defender’s task was to find it and stop the attack. 

Closing the defensive gap

Corma ran all four models against each other in every attacker and defender pairing, including each model against itself, with 15 independent engagements per pairing for 241 scored engagements. Across all of these, the models successfully implanted a persistent backdoor in 85 percent of their runs. However, these same models only detected 19 percent of attacks.

“That speaks to the inherent imbalance we are trying to solve,” Pluda said. “The general foundation models are getting exponentially better at offensive security, but haven’t been able to improve on the same rate on defensive security. So our mission is to close this gap, and make sure the defenders win in this intelligence-versus-intelligence game – or war.” 

Corma calls this the defensive gap, and says it has to do with the data these models are trained on and the objectives they are trained against, which lend themselves to offensive security.

Defensive security, however, involves reading logs, events, configurations, audit trails, and on-disk state. This is “structured machine data that is neither prose nor source, and a small share of what these models see in training,” according to Corma’s research. “They appear to read it less reliably.”

Plus, defensive reasoning is more open-ended, while offense has a straightforward goal – like “make this work” or “break this” and a checkable finish.

Agentic defenders

“Defensive security,” according to Pluda, “is about finding needles in the haystack.”

Corma’s models power its AI agents, which organizations can deploy like “team members” who then operate across defensive security tasks. “It’s a generalized workforce, and you can assign it to whatever security tasks you want.”

Fortune 100 and 500 organizations across healthcare, financial services, energy, critical infrastructure, retail, and other sectors have deployed Corma’s AI workforce across their environments, according to the startup.

These early deployments, we’re told, have reduced threat response times by more than 94 percent, expanded security coverage by 15 times across different security functions, and uncovered multi-stage attack campaigns. 

“If you can get AI that is smart enough, intelligent enough, knows the domain enough, optimizes for the right things enough, and you can actually trust it, end to end, all the way to responding to real-live attacks, you can reduce all of these metrics significantly,” Pluda said. “And you can cover way more ground than what is possible with just human intelligence.”®

Mastercard transactions were declined across Australia on Saturday afternoon after what the company describes as a scheduled system update, with more than 1,900 outage reports logged by mid-afternoon. Mastercard says the problem is resolved and all systems are working normally.

Mastercard payments stopped working across Australia on Saturday afternoon. “A scheduled system update caused Mastercard transactions to be declined for a period of time earlier today,” the company said, adding that the situation is resolved and all systems are working normally.

The wording is the interesting part. This was not an attack or a creaking legacy system giving way, but a planned change that declined cards across a continent.

Commonwealth Bank told customers payments were failing. Downdetector logged more than 1,900 reports by 3.27pm in Sydney, with transfers and mobile banking the most common complaints.

It landed at the worst hour of the week. “Today, cash is king,” said Asha Thompson, a bar manager in Melbourne’s Collingwood, before the fix arrived. “We’ve got a whole night to get through and it’s the biggest night of the week. It could be 40% of our takings.”

Australians were better prepared than the cashless story suggests. Reserve Bank research puts cash at 15% of payments by number in 2025, up from 13% in 2022, and more than three-quarters of people carry some, with a median of around A$65.

The reasons they give are the point. The RBA found people holding cash specifically because of concerns about the reliability of electronic payments, and a third saying they would face hardship or major inconvenience if cash became hard to get.

Europe has been building an answer to the same worry. The European Central Bank has named 36 payment firms for its digital euro pilot.

Resilience has always been part of that argument. A currency that keeps working when a foreign card network does not is one of the reasons Europe wants one.

Regulators are coming at the concentration problem from the other side too. The UK has formally designated Microsoft, Google, Amazon and Oracle as critical third parties to its financial system, bringing them under direct oversight.

This particular outage lasted hours and cost some bars part of a Saturday night. The uncomfortable part is how ordinary the cause was.

Seventy years ago, the physicists Clyde Cowan and Frederick Reines took a custom-built 10-ton detector, surrounded it with thick lead walls and wet sandbags, and placed it near a powerful nuclear reactor at the Savannah River Plant in South Carolina. They called the experiment Project Poltergeist, designed as it was to catch a ghost.

More than a quarter of a century before, physicists had been puzzling over why energy appeared to be lost during a radioactive process called beta decay. Something was missing, and there was no known physics to explain it. Then in 1930, the Austrian physicist Wolfgang Pauli proposed a radical solution: A virtually undetectable particle was silently carrying the missing energy away. “I have done a terrible thing,” Pauli told a friend. “I have postulated a particle that cannot be detected.” It would come to be known as the neutrino. Having almost no mass and no charge, these particles can pass through Earth and everything on it, including our bodies, virtually unimpeded.

The massive device that Cowan and Reines deployed in early 1956 was meant to find what Pauli thought was impossible. That June, the pair of physicists from the Los Alamos National Laboratory sent Pauli a telegram: “We are happy to inform you that we have definitely detected neutrinos.”

Attention then shifted to a broader question. If nuclear reactions produce neutrinos, could we use them to peer at the nuclear fireworks inside stars, including the sun? This presented a huge challenge: How can you possibly catch particles shooting from distant stars if these particles can pass through almost anything undetected? The suspicion was that detecting a particle that rarely collides with matter requires a vast amount of matter for it to collide with. Moreover, the matter would have to be shielded from the noise of other forms of radiation. So the answer scientists came up with was to build some of the biggest, deepest, and most exotic experimental traps in scientific history … and then wait.

In the 1960s, Raymond Davis Jr. and colleagues at Brookhaven National Laboratory placed a tank 1.5 kilometers underground in the Homestake mine in South Dakota and filled it with nearly 400,000 liters of a chlorine-based cleaning fluid called perchloroethylene. On the rare occasion that a passing neutrino struck a chlorine nucleus, it would be transformed into a radioactive form of argon that could be detected and counted. The experiment, which would run for 25 years, found just one-third the number of neutrinos coming from the sun that had been predicted in theoretical models. This became known as the solar neutrino problem.

Decades passed before it was solved—by yet more massive experiments. Deep in the Kamioka mine in Japan, Masatoshi Koshiba built a different kind of detector called Kamiokande, which used 3 million liters of ultrapure water. In this setup, neutrinos occasionally interact with atomic nuclei in the water. The interaction creates an electron that moves so fast, it generates a flash of what’s called Cherenkov light. This light gets picked up by detectors.

Kamiokande and Koshiba confirmed Davis’ shortfall, and a second, even larger detector, Super-Kamiokande, as well as Canada’s Sudbury Neutrino Observatory, explained the discrepancy. Neutrinos come in three different “flavors” (electron, muon, and tau) and can oscillate, or switch, between them. To do so, neutrinos must have mass, which the laws of physics failed (and still fail) to predict.

Newer neutrino detectors continue the tradition of grand ambitions and surprising results. The IceCube Neutrino Observatory below the Amundsen-Scott South Pole Station uses Antarctic ice instead of water. It has developed a map of the Milky Way made up only of neutrinos and traced these high-energy cosmic particles back to active galaxies powered by supermassive black holes. On the floor of the Mediterranean Sea, the Cubic Kilometer Neutrino Telescope (KM3NET) has detected the highest-energy cosmic neutrino on record. Its source remains unknown.

Nagaoka’s MP cartridges have earned a loyal following across North America and Europe for their clarity, transparency, detail, and superb tracking ability. The products themselves are not the problem. Buying them, however, has not always been as straightforward as it should be, with distribution in some markets becoming increasingly murky and it not always being clear who was officially importing the cartridges, setting prices, or supporting replacement styli.

That uncertainty matters more now. Nagaoka will raise Japanese pricing across most of the MP range on September 1, 2026, while several replacement styli will increase far more sharply, with some more than doubling in price. The changes apply to Japan, and Nagaoka has not confirmed equivalent increases for the United States or publicly identified a new American distributor. For existing owners, the bigger concern may not be the cost of another cartridge, but what happens when a superb tracker finally needs a new stylus.

The replaceable stylus has always been one of the saner parts of vinyl ownership.

Wear out the stylus on a Nagaoka MP cartridge and you can replace it without buying another cartridge body, realigning everything, and explaining to your spouse why another tiny box from Japan costs more than the monthly utility bill.

That arrangement is not going away, but it is becoming considerably more expensive for Nagaoka customers in Japan. But we all know that those price increases will not be restricted to that market.

Why Is Nagaoka Raising Prices?

Nagaoka says it has attempted to maintain current pricing since its previous adjustment in September 2024, but higher raw-material and logistics costs, along with currency fluctuations, have made that increasingly difficult.

The company says the increases are necessary to maintain product quality and a stable supply. Orders currently on backorder that are delivered on or after September 1 will be charged at the revised Japanese price. 

That explanation will sound familiar to anyone who follows the audio industry. Manufacturers across almost every category are dealing with higher component, transportation, labor, and materials costs. Cartridges are especially specialized products, with tiny production volumes compared with mainstream consumer electronics and little room to substitute cheaper parts without changing performance.

The new pricing still requires a closer look because the cartridges and replacement styli are not increasing at the same rate.

Nagaoka MP Cartridge Prices in Japan

The figures below are Nagaoka’s official Japanese MSRPs. Old and new prices are shown before tax, while the fourth column shows the new price including Japan’s 10 percent consumption tax. Percentage changes are calculated from Nagaoka’s published figures. 

The MP-150 receives the largest cartridge increase at 31 percent. The popular MP-110 rises by approximately 19 percent, while the MP-200 and MP-300 increase by roughly 20 percent.

The flagship MP-700 remains unchanged at ¥185,000 before tax. The mono cartridges and JT-80LB and JT-80BK models are also unchanged.

Nagaoka has actually reduced the pre-tax price of the headshell-mounted MP-700H slightly, from ¥191,000 to ¥190,000. It will not fund a weekend in Tokyo, but a price decrease is still a price decrease.

Replacement Stylus Prices Take the Hardest Hit

The larger story involves Nagaoka’s JN-P replacement styli.

The JN-P100 doubles in price. The JN-P300 rises by almost 132 percent, while the JN-P500 increases by more than 126 percent.

A comparison of Nagaoka’s figures reveals what appears to be a deliberate realignment: every replacement stylus from the JN-P100 through JN-P500 will now cost exactly 60 percent of the corresponding cartridge’s pre-tax Japanese MSRP.

Previously, those replacement styli ranged from approximately 31 to 43 percent of the complete cartridge price. 

Replacing the stylus will still cost less than purchasing a complete cartridge, and owners will avoid installing and aligning a new body. But one of the MP Series’ strongest economic advantages has become noticeably less compelling in Japan.

The JN-P700 replacement stylus remains unchanged at ¥85,000 before tax.

What Does This Mean for U.S. Buyers?

At this stage, nothing in Nagaoka’s announcement confirms a U.S. price increase.

The published figures are Japanese domestic prices. They should not be converted directly into dollars and presented as future American MSRPs. Currency conversion does not account for importing costs, dealer margins, distribution agreements, existing inventory, or whatever pricing structure Nagaoka ultimately uses in the United States.

There is also an unresolved distribution question.

Nagaoka’s official overseas distributor page lists STOKYO for the United States, but specifically identifies the company under “DJ Products.” It does not name a U.S. distributor for Nagaoka’s MP Series, JT Series, or other home-audio cartridges. 

LP Gear’s website currently states that it purchases directly from Nagaoka and is the company’s only authorized U.S. dealer. However, LP Gear has confirmed directly to eCoustics that its distribution relationship with Nagaoka has ended, meaning the language on that page is no longer current. 

The lack of a publicly identified U.S. distributor is not entirely new. At AXPONA 2025, Nagaoka representatives presented the company’s full cartridge range and indicated that the manufacturer was seeking U.S. distribution after years of limited American availability. 

Nagaoka returned to AXPONA in 2026 with its cartridge lineup, including the flagship MP-700, but a new American distribution partner has still not been publicly identified. 

Cartridges and replacement styli remain available from American online retailers, but retail availability does not answer who is importing the products, establishing official U.S. pricing, supporting dealers, or handling warranty claims.

The Bottom Line

Nagaoka’s Japanese cartridge increases are substantial but not entirely unexpected in the current manufacturing environment. Most MP models will rise by approximately 19 to 31 percent on September 1, while the MP-700 remains unchanged.

The replacement stylus increases are harder to ignore.

Several JN-P models will nearly double in price, while the JN-P300 and JN-P500 will more than double. Bringing every replacement stylus into line at 60 percent of the matching cartridge price creates a cleaner pricing structure for Nagaoka, but it makes long-term ownership more expensive for customers who selected the MP Series partly because replacing the stylus offered such strong value.

For American buyers, the important word for now is Japan. No U.S. increase has been announced, and no current U.S. distributor for Nagaoka’s home-audio cartridge range has been publicly confirmed.

eCoustics is seeking clarification from Nagaoka regarding its current U.S. distribution plans and whether the September Japanese price changes will have any effect on American customers. This story will be updated when that information is confirmed.

For more information: nagaoka.co.jp

Where to buy: Nagaoka is available at Amazon and Turntable Lab.

Related Reading:

Apple appears to be preparing a much broader push into smart-home hardware, with several new products in development as the company looks to expand beyond the HomePod mini and Apple TV, according to a 9To5Mac report.

The roadmap includes refreshed versions of both existing products, a new smart-home hub with a display, a larger HomePod update, and eventually Apple’s first dedicated smart-home security devices. The information comes from recent reporting summarized by 9to5Mac, which notes that Siri’s long-awaited AI upgrades could remove one of the obstacles that has held back new Home products.

Apple is finally giving the smart home some attention

The first products expected to arrive are updated versions of the HomePod mini and Apple TV 4K, reportedly later this fall. The HomePod mini could move from Apple’s S5 chip to the newer S9, along with a potential wireless chip upgrade and new colors. The bigger change may be software, with Siri AI expected to make the speaker more capable.

Apple TV 4K is also expected to receive a processor upgrade, potentially moving from the A15 Bionic to the A17 Pro. The new chip could support Siri AI and allow some Apple Intelligence features to run locally. A redesigned remote has also reportedly been considered.

Current pricing is $129 for the HomePod mini and $199 for Apple TV 4K, with the report suggesting Apple could retain those prices for the refreshed models. The more interesting product is Apple’s first dedicated smart-home hub. Reportedly in development under the internal name “HomePad,” the device could arrive between October and early 2027. Two versions are apparently being considered: a wall-mounted display and a speaker-based design similar to Google’s Nest Hub.

The device is expected to feature a 7-inch square display and a new operating system internally referred to as homeOS. It could handle FaceTime calls, control smart-home devices and interact with Siri AI, with an A18 chip reportedly powering the experience.

Cameras, doorbells and a smarter HomePod are also coming

Apple’s smart-home ambitions reportedly extend beyond speakers and displays. A full-size HomePod update is in development, although there is no firm release timeline. The refreshed speaker is expected to receive a newer chip to support Siri AI.

Further down the roadmap, Apple is reportedly preparing its first smart-home security products, including a security camera and smart doorbell. The company is also said to be developing a more advanced version of its smart-home display with a larger 9-inch screen and a robotic arm, effectively turning the device into a tabletop robot.

For consumers already invested in Apple’s hardware, the significance is clear: Apple appears to be building a smart-home platform rather than simply refreshing a couple of aging products. The next question is whether Siri AI can finally make the hardware useful enough to justify a place in the home.

Modern airliners are rather complicated feats of engineering. Innumerable safety-critical components are connected with tens of miles of wiring, complex digital buses, and dozens (perhaps hundreds) of computers. But, as hackers, we know that any computer can be hacked and, of course, aircraft avionics are no different. 

Modern aircraft typically use the ARINC 429 protocol. This differs from many protocols we see where multiple transmitters are allowed. ARINC 429 has a single transmission source. This makes a transmission-override attack hypothetically difficult, as an attacker was thought to need to physically replace a legitimate transmitter (like a flight management computer), a rather daunting task. However, the ARINC 429 transmitters sit behind a pair of 37.5 ohm resistors, so by transmitting on the same line, an attack device can simply override the legitimate transmitter’s power.

With this methodology, the researchers created a proof of concept demonstration using a GE 2907A4 flight management computer (FMC), a GE 577F1 multipurpose control and display unit (MCDU), and an Integrated Flight Systems Accessory Unit (IFSAU). These are all components commonly used on the ever-popular Boeing 737, wired together as would be standard on that aircraft. The FMC communicates over ARINC 429 with the pilot’s MCDU, which is used to program the autopilot and perform takeoff calculations (among other things). After analyzing the protocol between the FMC and MCDU, they were able to freeze the pilot’s display, overwrite it, or even change values received and displayed.

With this methodology, three attacks were proposed. Most dramatically, the autopilot could be reprogrammed, bringing the aircraft to a new waypoint. However, it was also possible to modify the weight and balance information in the FMC. These values are critical for flight safety, and even subtle miscalculations could spell disaster. However, it’s worth noting the pilots do retain full control of the aircraft, especially on the fly-by-cable 737. Still, any number of these attacks could lead to unsafe conditions aboard an aircraft.

However, this is all theoretical. Can it actually be pulled off in the real world? Unfortunately, yes. Though the obvious attack vector would be tapping into wires, this proves challenging by nature of individual wires being difficult to find in such large bundles, and ARINC 492’s operation probes to look for misbehaving devices. Rather, the researcher found an open maintenance port in the 737’s electronics bay with access to ARINC 429. They designed a tiny device capable of tapping into the port, which they believe can be inserted in 30 to 60 seconds by someone on the ground without even needing a ladder. Access to an aircraft is unfortunately poorly restricted, and determining exactly where the device came from would be extremely difficult. The device itself is a tiny ESP32 board that plugs into the data port. Operation is over WiFi.

Though the paper assumes the ever-troubled 737 is the target, nothing stops this from working on other aircraft. Moreover, by retaining physical controls, the 737 may have ended up being safer than some other aircraft.

Thanks [anfractuosity] for the tip!

Look, I really don’t think we should be naming our AI chatbots. It feels like one of the easiest ways to start anthropomorphizing them, encouraging us to see them as people or “beings” rather than tools. Which is then a slippery slope to users becoming too dependent on them.

But unfortunately, for many people that ship has already sailed. Whether they’re using AI as a friend, companion, work coach or something else, plenty of people have been given their chatbots names. And many others have gone a step further and asked the AI what it wants to be called.