Aussie maker Turnah81 wanted to know whether an ordinary person could create a functional car airbag at home. He documented every step in a recent video and followed through with an actual crash test on a homemade rig. The results offer a clear window into both the ingenuity behind the build and the reasons professionals handle these systems.

Turnah81 began with the fundamental problem that all airbags must solve. In a crash, the car comes to a quick stop, but the driver continues forward because a bag must inflate and get in the way of items such as the steering wheel or dashboard before the driver’s body can close the space. Commercial versions use sensors, CPUs, and special inflators to achieve split-second timing, but he believed he could do better by stripping everything back.

Sale

LEGO Technic Ferrari FXX K Toy Car – Building Toy for Girls & Boys, Ages 10+ – Cool Birthday Gift for…

• LEGO FERRARI MODEL CAR KIT – Builders ages 10+ can create the legendary Technic Ferrari FXX K with authentic details and working mechanical features
• AUTHENTIC RACE CAR DETAILS – This supercar building set features working butterfly doors, opening hood, and an engine cover that reveals the…
• CAR MODEL KIT – Young engineers explore real automotive concepts with this educational learning toy as they build the working differential and watch…

The trigger was an inertial fuel cut off switch, which is what mechanics use to shut down the fuel pump after a severe impact. He set it up to close an electrical circuit when it received the correct kind of jolt. His concept lacked microcontrollers and accelerometers, instead opting for a traditional mechanical system that could be built in a workshop by one person. Next came the airbag module, which he salvaged off a wrecked car’s steering wheel. He got the gas generator, but he put the chemical one away and replaced it with a standard pressurized gas cylinder with an electronic valve. When the inertial switch activated, current flowed to the valve, opening it, and the gas simply pushed it into the bag. The cylinder generated a lot of pressure without causing any terrible explosive reactions like the chemical ones. Commercial gas generators are small, only a canister that can be hidden in a steering column, which saves space, but the compressed gas cylinder he used was large and cumbersome. When you see the entire thing in place, it becomes clear why manufacturers prefer the chemical approach; it simply will not fit neatly into a car.

He utilized a pillow case as the airbag material, with reinforcement at the stress areas and some careful sewing to prevent all that gas from rushing in. He put it all on a crash test rig, which is supposed to give you a realistic impression of what happens in a crash without putting a real automobile or person at risk. It works by detecting a collision and opening the valve, allowing gas to enter and inflate the pillow case. The whole thing moves quickly enough to see what’s happening on on camera, but it also demonstrates the practical reality, as it has some significant restrictions. Measuring the timing reveals that the bag inflates just as quickly as the actual thing, but the bag itself is different, as is the pressure profile, which is due to the fact that the gas is cool and expands slower.

Turnah81 conducted a number of studies and openly released the findings, with each trial supporting the premise that the system could be deployed on demand. The accompanying video captures that exact moment when the rig ploughs ahead and the bag whacks the bag where a driver’s chest would normally be. Although the findings are encouraging, the configuration demonstrates why a DIY airbag would always fall short of commercial kit. You have compressed gas cylinders that require constant pressure monitoring and temperature adjustments, which a basic valve cannot do on its own. If you set up the system on a cold winter morning, it will be underpowered. In warmer conditions, however, the bag will overinflate, rendering it incapable of providing adequate cushioning. The inertial switch detects head-on impacts well, but it may miss glancing blows or crashes at an angle, which commercial sensor suites can detect precisely.
[Source]

The University of Michigan invested 20 million dollars in OpenAI before ChatGPT existed, before Microsoft committed billions, and before the company was worth more than some countries. Court documents from the Musk v. Altman trial revealed this week that the stake carries a target redemption value of two billion dollars. A university endowment made a hundred-to-one return on an artificial intelligence company that was, at the time of investment, a nonprofit research laboratory with no commercial product.

The investment appeared in an exhibit filed in the federal trial in Oakland, California, where Elon Musk is suing OpenAI and its leadership for 150 billion dollars, alleging that the company’s conversion from nonprofit to for-profit corporation constituted theft from a charity. The document listing early investors was not the focus of the trial. But the line item, 20 million dollars from the University of Michigan, has become the most consequential revelation for anyone interested in who saw the AI revolution coming and who actually wrote a cheque.

The bet

Michigan’s investment arrived in one of OpenAI’s earliest fundraising rounds, alongside Khosla Ventures at 50 million dollars, Reid Hoffman’s Aphorism Foundation at 50 million, a Y Combinator fund at 10 million, and the trust of Google’s Paul Buchheit at three million. The round predated Microsoft’s initial one billion dollar investment in 2019 and the public release of ChatGPT in November 2022. At the time, OpenAI was a nonprofit whose mission was to ensure that artificial general intelligence benefits all of humanity. It had no revenue model, no consumer product, and no path to a public listing.

University endowments invest in venture capital and early-stage companies as part of their alternative asset allocation, typically committing capital through fund-of-funds structures or direct investments managed by the endowment’s chief investment officer. Michigan’s endowment, which totalled approximately 17.9 billion dollars at the end of fiscal 2025, has been more aggressive than most in its AI allocation. The 20 million dollar commitment to OpenAI was not a rounding error in a portfolio of that size. But it was a bet on a nonprofit research lab at a time when the commercial potential of large language models was understood by almost nobody outside the organisations building them.

The relationship

The OpenAI investment was not Michigan’s only connection to Sam Altman. In 2023, the university committed 75 million dollars to Hydrazine Capital, a venture fund led by Altman. By 2024, Michigan had increased that commitment to 180 million dollars. The Hydrazine investments are separate from the OpenAI stake, distinct vehicles with different structures and return profiles. But the combined exposure, 200 million dollars across a direct investment and a venture fund both connected to the same individual, represents an unusual concentration of a university endowment’s capital in one network.

The Michigan Daily, the university’s student newspaper, reported in 2024 that the endowment had increased its allocation to AI and cryptocurrency investments, generating returns that outperformed the broader market. An opinion column in the same publication argued that the university should scale back its AI investments, citing ethical concerns about the technology the endowment was profiting from.

Musk called himself “a fool” on the stand for funding OpenAI, a characterisation that applies to his own contributions of approximately 50 million dollars to the same nonprofit that Michigan invested in. The difference is that Musk’s contributions were donations to a nonprofit. Michigan’s investment, through the for-profit conversion, became equity in a company now valued at 852 billion dollars.

The conversion

OpenAI’s transformation from nonprofit to for-profit is the mechanism that turned Michigan’s 20 million dollar investment into a two billion dollar stake. In October 2025, OpenAI restructured into OpenAI Group PBC, a public benefit corporation. The OpenAI Foundation retained a 26 per cent stake. Microsoft held 27 per cent. Early investors, including Michigan, saw their positions converted into equity in an entity that could pursue a public listing.

Brockman’s own journals, introduced at trial, described the nonprofit mission as “a lie,” language that Musk’s legal team used to argue that the conversion was premeditated. The conversion is the central issue in the trial. For Michigan’s endowment, it is the event that crystallised the return. Without the for-profit conversion, the 20 million dollar investment would have remained a contribution to a nonprofit with no liquidity path.

OpenAI closed a 122 billion dollar funding round in March 2026 at a post-money valuation of 852 billion dollars. The round included commitments from SoftBank, Andreessen Horowitz, Amazon, and Nvidia. An IPO is anticipated, with internal targets discussed for a filing in the second half of 2026 and a listing that could value the company at one trillion dollars. If Michigan holds its position through a public offering at that valuation, the return would exceed a hundred to one.

The pattern

Stanford’s James Zou is targeting a one billion dollar valuation for an AI physiology startup backed by research published in Nature, one example of a university-to-company pipeline that has produced some of the most valuable AI companies. Google emerged from Stanford. OpenAI’s founding team included researchers from Berkeley and Stanford. The university endowments that invested earliest in these networks have generated returns that dwarf their conventional portfolios.

Michigan’s 20 million dollar investment is exceptional in magnitude but not in kind. University endowments have been allocating to venture capital for decades. Yale’s endowment, under the late David Swensen, pioneered the model of heavy alternative asset allocation that most large endowments now follow. What distinguishes Michigan’s OpenAI bet is not the strategy but the timing and the target. The endowment committed capital to an AI nonprofit before the technology had demonstrated commercial viability, before the industry had attracted mainstream venture capital at scale, and before the word “ChatGPT” existed in any language.

The AI industry’s trajectory in 2025 confirmed what Michigan’s investment office apparently understood years earlier: that large language models would become the most valuable technology platform since the smartphone. The 20 million dollars is now worth two billion. The university that wrote the cheque will have to decide, when OpenAI eventually goes public, whether to take the return or hold the position in a company that is losing 14 billion dollars a year while generating 25 billion in annualised revenue. The bet was prescient. The exit will determine whether it was also wise.

There’s another type of digital scam to be aware of, as per the BBC. It’s called “reservation hijacking.”

The name gives you a clue as to how it works. Essentially, scammers use details about a booking you’ve placed (perhaps with a hotel or airline) to trick you into sending money somewhere you shouldn’t.

While this type of scam isn’t brand new, a recent data breach at Booking.com has raised the risk of people being caught out. With data about you and your reservation, a far more convincing setup can be put in place—why wouldn’t you believe that someone purporting to be an employee from a spa you’ve got a reservation with is telling the truth about who they are, especially if they know the dates of your trip, your phone number, and your email address?

According to Booking.com, no financial information was exposed in the April 2026 hack. However, names, email addresses, phone numbers, and booking details have been leaked. The travel portal says affected customers have been emailed about the heightened risk of scams, so that’s the first thing to check for when it comes to staying safe.

Minimizing the risk of getting scammed by a reservation hijack involves many of the same security precautions you may already be following, and just being aware that this is a way you might be targeted will make a difference.

How Reservation Hijacks Work

Scammers can get hold of your booking details.

Courtesy of David Nield

We’ve already outlined the basics of a reservation hijack, but it can take several forms. As with other types of scams, it tends to evolve over time. The basic premise is that someone will get in touch with you claiming to be from a place you have a reservation with, whether it’s a car rental company or a hotel.

The scammers will try to pull together as much information as they can on you and your booking. Sometimes they’ll target employees of the place you’ve got the reservation with in order to get access to their systems, and other times they may take advantage of a wider data breach (as with the recent Booking.com hack).

They might also get information through other means. Maybe they’ve somehow got access to your email, or to some of your social media posts (where you’ve shared your next vacation destination and a countdown of how many days are left to go). Don’t be caught out if you find yourself speaking to someone who knows a lot about your travel plans.

When photography enthusiasts begin shopping for action cameras, the latest models with all of their bells and whistles typically receive all of the attention. These newer models have fancy resolutions and tons of sensors, but after months of putting the DJI Osmo Action 4 Essential Combo, priced at $198.99 (was $299), through its paces, hiking, bicycling, and even taking it underwater, users have surprisingly come to dub it the most undervalued action cam around.

It handles difficult terrain and quick moves with ease, keeping the film as steady as a rock thanks to its internal stabilizer. Getting pro-quality footage without having to deal with extra equipment and software is a huge plus, and 4K at 120 frames per second in slow-motion is just the icing on the cake. You also get a larger 1.3 inch sensor, which lets in more light and allows you to take stunning images in low light or dense forest.

Sale

DJI Osmo Action 4 Essential Combo, Waterproof Action Camera with 1/1.3″ Sensor, 4K/120fps Video, Stunning…

• Superior Low-Light Performance with 1/1.3-inch Sensor – Struggling with grainy footage in low light? Osmo Action 4 action camera 4K uses 2.4μm pixel…
• Pro-Grade Color Science – Worried your videos lack real-life colors? 10-bit color depth captures subtle shades, while D-Log M preserves more detail…
• See More, Feel More – Want to capture more action and scenery? Shoot in 4K at 120fps with a 155° ultra-wide field of view. Minimal cropping means…

The battery lasts longer than expected, making it ideal for all-day shots, especially on frigid days when other cameras start to droop. The Essential Combo pack includes the main body, a replacement battery, and a quick-release frame that fits into any standard mount, which is really good value for the price.

The D-Log mode provides 10-bit color depth and allows editors to experiment with tones while maintaining quality. The 4:3 aspect ratio allows you to easily upload some amazing tall vertical movies from social media, and the bitrate is sufficient to maintain those fine textures looking great in grass, water, or cloth, often exceeding newer models in real-world use. It also effortlessly pairs with DJI mics, allowing you to capture high-quality audio without much fuss.

It has a water resistance rating of 18 meters on its own and 60 meters with the extra case, so you can go for a splash or dive with complete confidence. The magnetic clip mechanism allows you to swiftly attach and detach from handlebars, chest straps, and helmets. Yes, there is no built-in storage, but most folks will have an SD card lying around somewhere.

When Apple unveils its next macOS at WWDC 2026, a new report says that it will have a slightly redesigned Liquid Glass interface, though really just the same design iterations the company has always done.

Liquid Glass has had vocal critics, but just as with every version of macOS before, Apple is going to refine and mildly redesign it each year. According to Bloomberg, this year’s revision is chiefly concerned with the appearance of different Mac elements with Liquid Glass.

Specifically, the “slight redesign” is to concentrate on improving various readability issues. Where those have arisen so far, it’s been in Liquid Glass’s transparency and shadow effects, so presumably that is what Apple will work on.

This is the same thing Apple does after every significant redesign, starting with the toning down of the Aqua interface in the first years of Mac OS X. It was perhaps most noticeable with iOS 7 which debuted a very flat design that over the next years was slowly improved and clarified.

Speaking of iOS, though, the report also says that Apple’s work on the next version of this will include a benefit for macOS 27. Apple is said to be working to have tabs in Safari automatically organize themselves, as can already be done in rival browsers.

Tab Groups is definitely an area that needs attention, and not only because elements of it are better on other browsers. Many years ago, Apple added an option to Shortcuts on the Mac that was supposed to let users switch automatically between Tab Groups, but to this day the Shortcut action fails with an “internal error.”

It does work perfectly on iOS and iPadOS, though, so hopes that improvements on those platforms will come to the Mac as well could yet be wishful thinking.

Looking for the most recent Strands answer? Click here for our daily Strands hints, as well as our daily answers and hints for The New York Times Mini Crossword, Wordle, Connections and Connections: Sports Edition puzzles.

Some really old-timey words appear in today’s NYT Strands puzzle. I found a few of the answers difficult to unscramble, so if you need hints and answers, read on.

I go into depth about the rules for Strands in this story. 

If you’re looking for today’s Wordle, Connections and Mini Crossword answers, you can visit CNET’s NYT puzzle hints page.

Read more: NYT Connections Turns 1: These Are the 5 Toughest Puzzles So Far

Hint for today’s Strands puzzle

Today’s Strands theme is: A nice medley

If that doesn’t help you, here’s a clue: This and that.

Clue words to unlock in-game hints

Your goal is to find hidden words that fit the puzzle’s theme. If you’re stuck, find any words you can. Every time you find three words of four letters or more, Strands will reveal one of the theme words. These are the words I used to get those hints, but any words of four or more letters that you find will work:

• PODGE, MELD, BEND, SHAME, DOPE, RIDE, HAMS, BARN, DOSE

Answers for today’s Strands puzzle

These are the answers that tie into the theme. The goal of the puzzle is to find them all, including the spangram, a theme word that reaches from one side of the puzzle to the other. When you have all of them (I originally thought there were always eight but learned that the number can vary), every letter on the board will be used. Here are the nonspangram answers:

• JUMBLE, RAGBAG, VARIETY, HODGEPODGE, MISHMASH

Today’s Strands spangram

Today’s Strands spangram is ODDSANDENDS. To find it, look for the O that is five letters down on the far-left vertical row, and wind across.

The idea of using cardboard for a sloppy PC case isn’t new; it’s a time-honored tradition dating back to at least the 1990s. That said, with today’s CNC cutters and other advanced tooling available to hobbyists, you might be curious to see how far you can push the concept. As demonstrated in a recent video by [mryeester], the answer appears to be that good planning and a solid understanding of cardboard’s limitations are as essential as ever.

After having the PC case drawn up in CAD and cut on a professional CNC cutter by a buddy who makes commercial cardboard displays, the installation procedure for the PC components showed where a bit of foresight could have saved a lot of time and effort.

The first problem was that the GPU couldn’t be installed due to wrong measurements on where the IO bracket normally is screwed into the case. Some cardboard cutting later, the GPU slid into place, but of course, there’s no way to screw it down, putting the full weight on the PCIe slot of the mainboard. Fortunately, the mainboard was quite literally bolted into place, and the case consists of multiple layers of corrugated cardboard to add some rigidity.

Next was more carving as the PSU cut-out was designed for an SFX PSU, not an ATX one. After that ordeal, one could say that perhaps a nice thing about a cardboard case is that you get to pick where buttons are located, though this comes with its own logistical issues.

Finally, mounting side panels turned into another chore, with perhaps some engineering possible to make it work better. For example, we recently looked at making cardboard hinges that would look pretty good on a cardboard PC case. You can also waterproof cardboard and make it much stronger, turning a throwaway, temporary cardboard solution into something that will last for years, even with occasional exposure to moisture and a water-cooling leak. (more…)