Sleep apnea is a debilitating disease that many sufferers don’t even realize they have. Those afflicted with the condition will regularly stop breathing during sleep as the muscles in their throat relax, sometimes hundreds of times a night. Breathing eventually resumes when the individual’s oxygen supply gets critically low, and the body semi-wakes to restore proper respiration. The disruption to sleep causes serious fatigue and a wide range of other deleterious health effects.

Treatment for sleep apnea has traditionally involved pressurized respiration aids, mechanical devices, or invasive surgeries. However, researchers are now attempting to develop a new drug combination that could solve the problem with pharmaceuticals alone.

Breathe Into Me

There are a variety of conditions that fall under the sleep apnea umbrella, with various causes and a range of imperfect treatments. Perhaps the most visible is obstructive sleep apnea (OSA), in which the muscles in the throat relax during sleep. Under certain conditions, and depending on anatomy, this can lead the airway to become blocked, causing a cessation of breathing that requires the sufferer to wake to a certain degree to restore proper respiration. Since the 1980s, OSA has routinely been treated with the use of Continuous Positive Airway Pressure (CPAP) machines, which supply pressurized air to the face and throat to forcibly keep the airway open. These are effective, except for one major problem—a great deal of patients hate them, and compliance with treatment is remarkably poor. Some studies have shown up to 50% of patients give up on CPAP treatment within a year due to discomfort around sleeping with a pressurized air mask.
Obstructive sleep apnea occurs when upper airway muscles relax excessively during sleep, ultimately restricting or totally blocking the airway. Credit: Apnimed
Against this backdrop, a simple pill-based treatment for sleep apnea is a remarkably attractive proposition. It would allow the treatment of the condition without the need for expensive, high-maintenance CPAP machines which a huge proportion of patients hate using in the first place. Such a treatment is now close to being a reality, under the name AD109.

The treatment aims to directly target the actual cause of obstructive sleep apnea. OSA is a neuromuscular condition, and one that only occurs during sleep—as those afflicted with the disease don’t suffer random airway blockages while awake. When sleep occurs, neurotransmitter levels like norepinephrine tend to decrease. This can can cause the upper airway muscles to excessively relax in sleep apnea sufferers, to the point that the airway blocks itself shut. AD109 tackles this issue with a combination of drugs—an antimuscarinic called aroxybutynin, and a norepinephrine reuptake inhibitor called atomoxetine. In simple terms, the aroxybutynin blocks so-called muscarinic receptors which decrease muscle tone in the upper airway. Meanwhile, the atomoxetine is believed to simultaneously improve muscle tone in the upper airway by maintaining higher activity in the hyperglossal motor neurons that control muscles in this area.
Results in phase 2 testing showed a marked decrease in AHI compared to those taking a placebo. Credit: research paper
Thus far, clinical testing has been positive, suggesting the synergistic combination of drugs may be able to improve airflow for sleep apnea patients. Phase 1 and Phase 2 clinical trials have been conducted to verify the safety of the treatment, as well as its efficacy at treating the condition. Success in the trials was measured with the Apnea-Hypopnea Index (AHI), which records the number of airway disruptions an individual has per hour. AHI events were reduced by 45% in those taking AD109 when compared to the placebo group in a phase 2 trial featuring 211 participants. It achieved this while proving generally safe in early testing without causing detectable detriments to attention or memory. However, some side effects were noted with the drug—most specifically dry mouth, urinary hesitancy, and a level of insomina. The latter being particularly of note given the drug’s intention to improve sleep.

Testing on AD109 continues, with randomized Phase 3 trials measuring its performance in treating mild, moderate, and severe obstructive sleep apnea. For now, commercialization remains a ways down the road. And yet, for the first time, it appears promising that modern medicine will develop a simple drug-based treatment for a disease that leaves millions fatigued and exhausted every day. If it proves viable, expect it to become a major pharmaceutical success story and the hottest new drug on the market.

hackaday.com/2026/01/22/new-dr…

[Kevin] admits that FreeCAD may not be the ideal tool for editing STL files. But it is possible, and he shares some practical advice in the video below. If you want to get the most out of your 3D printer, it pays to be able to create new parts, and FreeCAD is a fine option for that. However, sometimes you download an STL from the Internet, and it just isn’t quite what you need.

Unlike native CAD formats, STLs are meshes of triangles, so you get very large numbers of items, which can be unwieldy. The first trick is to get the object exactly centered. That’s easy if you know how, but not easy if you are just eyeballing it.

If you use the correct workbench, FreeCAD can analyze and fix mesh problems like non-manifold parts, flipped normals, and other issues. The example is a wheel with just over 6,000 faces, which is manageable. But complex objects may make FreeCAD slow. [Kevin] says you should be fine until the number of faces goes above 100,000. In that case, you can decimate the number of faces with, of course, a corresponding loss in resolution.

Once you are satisfied with the mesh, you can create a real FreeCAD shape from the mesh. The resulting object will be hollow, so the next step will be to convert the shape to a solid.

That still leaves many triangles when you really want flat surfaces to be, well, flat. The trick is to make a copy and use the “refine shape” option for the copy. Once you have a FreeCAD solid, you can do anything you can do in FreeCAD.

We’ve run our share of FreeCAD tips if you want more. There are other ways to tweak STLs, too.

youtube.com/embed/bdt53O5_rsw?…

hackaday.com/2026/01/22/stl-ed…

There’s little about building spacecraft that anyone would call simple. But there’s at least one element of designing a vehicle that will operate outside the Earth’s atmosphere that’s fairly easier to handle: aerodynamics. That’s because, at the altitude that most satellites operate at, drag can essentially be ignored. Which is why most satellites look like refrigerators with solar panels and high-gain antennas attached jutting out at odd angles.

But for all the advantages that the lack of meaningful drag on a vehicle has, there’s at least one big potential downside. If a spacecraft is orbiting high enough over the Earth that the impact of atmospheric drag is negligible, then the only way that vehicle is coming back down in a reasonable amount of time is if it has the means to reduce its own velocity. Otherwise, it could be stuck in orbit for decades. At a high enough orbit, it could essentially stay up forever.
Launched in 1958, Vanguard 1 is expected to remain in orbit until at least 2198
There was a time when that kind of thing wasn’t a problem. It was just enough to get into space in the first place, and little thought was given to what was going to happen in five or ten years down the road. But today, low Earth orbit is getting crowded. As the cost of launching something into space continues to drop, multiple companies are either planning or actively building their own satellite constellations comprised of thousands of individual spacecraft.

Fortunately, there may be a simple solution to this problem. By putting a satellite into what’s known as a very low Earth orbit (VLEO), a spacecraft will experience enough drag that maintaining its velocity requires constantly firing its thrusters. Naturally this presents its own technical challenges, but the upside is that such an orbit is essentially self-cleaning — should the craft’s propulsion fail, it would fall out of orbit and burn up in months or even weeks. As an added bonus, operating at a lower altitude has other practical advantages, such as allowing for lower latency communication.

VLEO satellites hold considerable promise, but successfully operating in this unique environment requires certain design considerations. The result are vehicles that look less like the flying refrigerators we’re used to, with a hybrid design that features the sort of aerodynamic considerations more commonly found on aircraft.

ESA’s Pioneering Work

This might sound like science fiction, but such craft have already been developed and successfully operated in VLEO. The best example so far is the Gravity Field and Steady-State Ocean Circulation Explorer (GOCE), launched by the European Space Agency (ESA) back in 2009.

To make its observations, GOCE operated at an altitude of 255 kilometers (158 miles), and dropped as low as just 229 km (142 mi) in the final phases of the mission. For reference the International Space Station flies at around 400 km (250 mi), and the innermost “shell” of SpaceX’s Starlink satellites are currently being moved to 480 km (298 mi).

Given the considerable drag experienced by GOCE at these altitudes, the spacecraft bore little resemblance to a traditional satellite. Rather than putting the solar panels on outstretched “wings”, they were mounted to the surface of the dart-like vehicle. To keep its orientation relative to the Earth’s surface stable, the craft featured stubby tail fins that made it look like a futuristic torpedo.

Even with its streamlined design, maintaining such a low orbit required GOCE to continually fire its high-efficiency ion engine for the duration of its mission, which ended up being four and a half years.

In the case of GOCE, the end of the mission was dictated by how much propellant it carried. Once it had burned through the 40 kg (88 lb) of xenon onboard, the vehicle would begin to rapidly decelerate, and ground controllers estimated it would re-enter the atmosphere in a matter of weeks. Ultimately the engine officially shutdown on October 21st, and by November 9th, it’s orbit had already decayed to 155 km (96 mi). Two days later, the craft burned up in the atmosphere.

JAXA Lowers the Bar

While GOCE may be the most significant VLEO mission so far from a scientific and engineering standpoint, the current record for the spacecraft with the lowest operational orbit is actually held by the Japan Aerospace Exploration Agency (JAXA).

In December 2017 JAXA launched the Super Low Altitude Test Satellite (SLATS) into an initial orbit of 630 km (390 mi), which was steadily lowered in phases over the next several weeks until it reached 167.4 km (104 mi). Like GOCE, SLATS used a continuously operating ion engine to maintain velocity, although at the lowest altitudes, it also used chemical reaction control system (RCS) thrusters to counteract the higher drag.

SLATS was a much smaller vehicle than GOCE, coming in at roughly half the mass. It also carried just 12 kg (26 lb) of xenon propellant, which limited its operational life. It also utilized a far more conventional design than GOCE, although its rectangular shape was somewhat streamlined when compared to a traditional satellite. Its solar arrays were also mounted in parallel to the main body of the craft, giving it an airplane-like appearance.

The combination of lower altitude and higher frontal drag meant that SLATS had an even harder time maintaining velocity than GOCE. Once its propulsion system was finally switched off in October 2019, the craft re-entered the atmosphere and burned up within 24 hours. The mission has since been recognized by Guinness World Records for the lowest altitude maintained by an Earth observation satellite.

A New Breed of Satellite

As impressive as GOCE and SLATS were, their success was based more on careful planning than any particular technological breakthrough. After all, ion propulsion for satellites is not new, nor is the field of aerodynamics. The concepts were simply applied in a novel way.

But there exists the potential for a totally new type of vehicle that operates exclusively in VLEO. Such a craft would be a true hybrid, in the sense that its primarily a spacecraft, but uses an air-breathing electric propulsion (ABEP) system akin to an aircraft’s jet engine. Such a vehicle could, at least in theory, maintain an altitude as low as 90 km (56 mi) indefinitely — so long as its solar panels can produce enough power.

Both the Defense Advanced Research Projects Agency (DARPA) in the United States and the ESA are currently funding several studies of ABEP vehicles, such as Redwire’s SabreSat, which have numerous military and civilian applications. Test flights are still years away, but should VLEO satellites powered by ABEP become common platforms for constellation applications, they may help alleviate orbital congestion before it becomes a serious enough problem to impact our utilization of space.

hackaday.com/2026/01/22/skimmi…

Although already having entered the territory of ‘retro gaming’, the Sony PlayStation 3 remains a notoriously hard to emulate game console. Much of this is to blame on its unique PowerPC-based Cell processor architecture, which uses a highly parallel approach across its asymmetric multi-core die that is very hard to map to more standard architectures like those in today’s x86 and ARM CPUs. This makes it even more amazing that the RPCS3 emulator team has now crossed the 70% ‘playable’ threshold on their compatibility list.

This doesn’t mean that you can fire up these games on any purported ‘gaming system’, as the system requirements are pretty steep. If you want any kind of enjoyable performance the recommended PC specifications feature an Intel 10th generation 6-core CPU, 16 GB of dual-channel RAM and a NVIDIA RTX 2000 or AMD RX 5000 series GPU or better.

It should be noted here also that the ‘playable’ tag in the compatibility list means that the game can be completed without game breaking glitches. Performance remains an issue, with very creative optimizations through e.g. the abuse of x86 SIMD instructions remaining the topic of research by the emulator developers. Yet as original PS3 hardware gradually becomes less available, the importance of projects like RPCS3 will become more important.

Header: Evan-Amos, Public domain.

hackaday.com/2026/01/22/playst…

The Nintendo Wii first launched in 2006, and quickly became a fixture in living rooms around the world. It offered motion-controlled bowling, some basic internet features, and a pretty decent Zelda game. On top of all that, though, you could also use it to order a pizza, as [Retro Game Attic] demonstrates.

The Wii used to organize different features of the console into “channels.” Way back in the day, you could install the Demae Channel on your Wii in Japan, which would let you order fast food from various outlets using the Demaecan service.

The Demae Channel service was discontinued in 2017. However, it has since been resurrected by WiiLink, which is a homebrew project which replicates the functionality of the original Nintendo WiiConnect 24 and Wi-Fi Connection servers. As it stands, you can load the WiiLink version of the Demae Channel (or Food Channel) on to your Wii, and use it to order pizza from your local Domino’s Pizza. It only works in the United States and Canada right now, and there are no other restaurants available, at least until further development is completed to add JustEat compatibility. It’s not entirely clear how much of the functionality was recreated from the original Demae Channel; what is clear is that plenty of custom development has been done on the WiiLink version to integrate it with modern delivery services.

What’s so exciting about this is how well it actually works. The app perfectly nails the classic Wii Channel visual style. It also seems to integrate well with the Domino’s API for digital orders, even displaying simple updates on holiday opening hours and order times. Pricing data and images of the pizzas are all available right in the app, and you can even make modifications. It might be a gimmick… but it actually works. Notably, though, the app avoids any stickiness with handling payment—thankfully, pay-on-delivery is still legitimate in the pizza world in 2026.

Will this revolutionize how you order pizza on a daily basis? Probably not. Is it one of the coolest Wii hacks we’ve seen in a while? Undeniably. Video after the break.

youtube.com/embed/_Z-yPjYAkjY?…

hackaday.com/2026/01/22/orderi…

Creative clocks are a dime a dozen, even clocks that use binary have been created in nearly every format. [typo] promises a clever adaptation to the binary format, and it promises a more usable display. Using a combination of both traditional binary and digital gradients creates a usable and yet still nerdy fun clock.

[typo]’s clock fits the traditional binary counting method with the hours on the left side of its face. On the other hand, its right side presents a lighting gradient depending on the completion of the hour. While this is simple in principle, [typo] chose to correct what many don’t consider when deploying visual gradients. The human eye doesn’t see everything exactly as it is, which creates a rough logarithmic curve that gets corrected for in the binary/digital hybrid clock.

If you want something more mobile and still have that smidge of difficult time telling you, check out this minimalist wrist watch!

youtube.com/embed/3vhNw6UBNwo?…

hackaday.com/2026/01/21/binary…

One of the good things about simulating circuits is that you can easily change component values trivially. In the real world, you might use a potentiometer or a pot to provide an adjustable value. However, as [Ralph] discovered, there’s no pot component in LTSpice. At first, he cobbled up a fake pot with two resistors, one representing the top terminal to the wiper, and the other one representing the wiper to the bottom terminal. Check it out in the video below.

At first, [Ralph] just set values for the two halves manually, making sure not to set either resistor to zero so as not to merge the nets. However, as you might guess, you can make the values parameters and then step them.

By using .step you can alter one of the resistor values. Then you can use a formula to compute the other resistor since the sum of the two resistors has to add up to the pot’s total value. That is, a 10K pot will have the two resistors always add up to 10K.

Of course, you could do this without the .step and simply change one value to automatically compute both resistors if you prefer.

We’ve done our own tutorials with .step and parameters if you want a little more context. You can even use this idea to make your own custom pot component.

youtube.com/embed/da3TbdDrjXo?…

hackaday.com/2026/01/21/simula…

Normally, if you want to blast out samples to a DAC in a hurry, you’d rely on an FPGA, what with their penchant for doing things very quicky and in parallel. However, [Anabit] figured out a way to do the same thing with a microcontroller, thanks to the magic of the Raspberry Pi Pico 2.

The design in question is referred to as the PiWave 150 MS/s Bipolar DAC, and as the name suggests, it’s capable of delivering a full 150 million samples per second with 10, 12, or 14 bits of resolution. Achieving that with a microcontroller would normally be pretty difficult. In regular linear operation, it’s hard to clock bits out to GPIO pins at that sort of speed. However, the Raspberry Pi Pico 2 serves as a special case in this regard, thanks to its Programmable I/O (PIO) subsystem. It’s a state machine, able to be programmed to handle certain tasks entirely independently from the microcontroller’s main core itself, and can do simple parallel tasks very quickly. Since it can grab data from RAM and truck it out to a bank of GPIO pins in a single clock cycle, it’s perfect for trucking out data to a DAC in parallel at great speed. The Pi Pico 2’s clock rate tops out at 150 MHz, which delivers the impressive 150 MS/s sample rate.

The explainer video is a great primer on how this commodity microcontroller is set up to perform this feat in detail. If you’re trying for accuracy over speed, we’ve explored solutions for that as well. Video after the break.

youtube.com/embed/UaaveknkjBM?…

hackaday.com/2026/01/21/drivin…

Well there’s your problem. (Credit: Mark Funeaux, YouTube)
Akin to the razor-and-blades model, capsule-based coffee machines are an endless grind of overpriced pods and cheaply made machines that you’re supposed to throw out and buy a new one of, just so that you don’t waste all the proprietary pods you still have at home. What this also means is a seemingly endless supply of free broken capsule coffee makers that might be repairable. This is roughly how [Mark Furneaux] got into the habit of obtaining various Nespresso VertuoLine machine for attempted repairs.

The VirtuoLine machines feature the capsule with a bar code printed on the bottom of the lip, requiring the capsule to be spun around so that it can be read by the optical reader. Upon successful reading, the code is passed to the MCU after which the brewing process is either commenced or cruelly halted if the code fails. Two of the Vertuo Next machines that [Mark] got had such capsule reading errors, leading to a full teardown of the first after the scanner board turned out to work fine.

Long story short and many hours of scrubbed footage later, one machine was apparently missing the lens assembly on top of the photo diode and IR LED, while the other simply had these lenses gunked up with spilled coffee. Of course, getting to this lens assembly still required a full machine teardown, making cleaning it an arduous task.

Unfortunately the machine that had the missing lens assembly turned out to have another fault which even after hours of debugging remained elusive, but at least there was one working coffee machine afterwards to make a cup of joe to make [Mark] feel slightly better about his life choices. As for why the lens assembly was missing, it’s quite possible that someone else tried to repair the original fault, didn’t find it, and reassembled the machine without the lens before passing the problem on to the next victim.

youtube.com/embed/v424-YxSAbM?…

hackaday.com/2026/01/21/repair…

This week Jonathan chats with Nicholas Adams about OpenRiak! Why is there a Riak and an OpenRiak, which side of the CAP theorem does OpenRiak land on, and why is it so blazingly fast for some operations? Listen to find out!

• linkedin.com/groups/14551410/
• github.com/openriak/
• join.slack.com/t/postriak/shar…
• files.tiot.jp/riak/
• tiot.jp/riak-docs/
• meetup.com/tokyo-openriak-meet…

youtube.com/embed/SOFBTvvtqLg?…

Did you know you can watch the live recording of the show right on our YouTube Channel? Have someone you’d like us to interview? Let us know, or have the guest contact us! Take a look at the schedule here.

play.libsyn.com/embed/episode/…

Direct Download in DRM-free MP3.

If you’d rather read along, here’s the transcript for this week’s episode.

Places to follow the FLOSS Weekly Podcast:

• Spotify
• RSS

Theme music: “Newer Wave” Kevin MacLeod (incompetech.com)

Licensed under Creative Commons: By Attribution 4.0 License

hackaday.com/2026/01/21/floss-…

If you follow electronics history, few names were as ubiquitous as RCA, the Radio Corporation of America. Yet in modern times, the company is virtually forgotten for making large computers. [Computer History Archive Project] has a rare film from the 1970s (embedded below) explaining how RCA planned to become the number two supplier of business computers, presumably behind behemoth IBM. They had produced other large computers in the 1950s and 1960s, like the BIZMAC, the RCA 510, and the Spectra. But these new machines were their bid to eat away at IBM’s dominance in the field.

RCA had innovative ideas and arguably one of the first demand paging, virtual memory operating systems for mainframes. You can hope they were better at designing computers than they were at making commercials.

The BIZMAC was much earlier and used tubes (public domain).
In 1964, [David Sarnoff] famously said: “The computer will become the hub of a vast network of remote data stations and information banks feeding into the machine at a transmission rate of a billion or more bits of information a second … Eventually, a global communications network handling voice, data and facsimile will instantly link man to machine — or machine to machine — by land, air, underwater, and space circuits. [The computer] will affect man’s ways of thinking, his means of education, his relationship to his physical and social environment, and it will alter his ways of living. … [Before the end of this century, these forces] will coalesce into what unquestionably will become the greatest adventure of the human mind.”

He was, of course, right. Just a little early.
The machines in the video were to replace the Spectra 70 computers, seen here from an RCA brochure.
The machines were somewhat compatible with IBM computers, touted virtual memory, and had flexible options, including a lease that let you own your hardware in six years. They mention, by the way, IBM customers who were paying up to $60,000 / month to IBM. They mentioned that an IBM 360/30 with 65K was about $13,200 / month. You could upgrade with a 360/30 for an extra $3,000 / month, which would double your memory but not double your computing power. (If you watch around the 18-minute mark, you’ll find the computing power was extremely slow by today’s standards.)

RCA, of course, had a better deal. The RCA 2 had double the memory and reportedly triple the performance for only $2,000 extra per month. We don’t know what the basis for that performance number was. For $3,500 a month extra, you could have an RCA 3 with the miracle of virtual memory, providing an apparent 2 megabytes per running job.

There are more comparisons, and keep in mind, these are 1970 dollars. In 1970, a computer programmer probably made $10,000 to $20,000 a year while working on a computer that cost $158,000 in lease payments (not to count electricity and consumables). How much cloud computing could you buy in a year for $158,000 today? Want to buy one? They started at $700,000 up to over $1.6 million.

By their release, the systems were named after their Spectra 70 cousins. So, officially, they were Spectra 70/2, 70/3, 70/5, and 70/6.

Despite all the forward-looking statements, RCA had less than 10% market share and faced increasing costs to stay competitive. They decided to sell the computer business to Sperry. Sperry rebranded several RCA computers and continued to sell and support them, at least for a while.

Now, RCA is a barely remembered blip on the computer landscape. You are more likely to find someone who remembers the RCA 1800 family of CPUs than an actual RCA mainframe. Maybe they should have throw in the cat with the deal.

Want to see the IBM machines these competed with? Here you go. We doubt there were any RCA computers in this data center, but they’d have been right at home.

youtube.com/embed/GygR1Nbrhfc?…

hackaday.com/2026/01/21/retrot…

[Brainiac75] is a fan of fiber optic lamps, except for one thing—they’re often remarkably dim. Thus, they set out to hack the technology to deliver terrifying amounts of light while still retaining their quirky charm.

Older fiber optic lamps use a dim filament lamp or halogen lamp to light them up. They also often feature a spinning color disk to vary the light patterns, which does have the side effect of absorbing some of the already-limited light output.

When it came to upgrading his own decades-old lamp, [Braniac75] decided to initially stick within the specs of the original halogen lamp. The fixture was rated for 12 volts at 5 watts, with a GU4/GZ4 compatible base, and white light was desired so the color wheel could still do its thing. Swapping out the original 5 W halogen for a 2.5 W LED unit brought a big upgrade in brightness, since the latter is roughly equivalent to a 20 W halogen in light output. Upgrading to a 4.2 W LED pushed things even further, greatly improving the look of the lamp.

The video also explores modding a modern fiber optic lamp, too. It was incredibly cheap, running off batteries and using a single color-changing LED to illuminate the fibers. [Braniac75] decided to try illuminating the plastic fibers with an RGB stage lighting laser rig—namely, the LaserCube Ultra 7.5 W from Wicked Lasers. With this kind of juice, the fiber lamp is eye-searingly bright, quite literally, and difficult to film. However, with the laser output dialed way down, the lamp looks amazing—with rich saturated colors dancing across the fiber bundle as the lasers do their thing.

If you’ve ever wanted to build a fiber lamp that doesn’t look like a cheap gimmick, now you know how. We’ve looked at weird applications for these lamps before, too.

youtube.com/embed/5uoKlGzmDgg?…

hackaday.com/2026/01/21/fiber-…