The lovely thing about the x86 architecture is its decades of backwards compatibility, which makes it possible to run 1990s operating systems on modern-day hardware, with relatively few obstacles in the way. Recently [Yeo Kheng Meng] did just that with Windows 98 SE on a 2020 ThinkPad P12s Gen 1, booting it alongside Windows 11 and Linux from the same NVMe drive.

Naturally, after previously getting MS-DOS 6.22 from 1994 running on a 2020 ThinkPad X13, the step to doing the same with Windows 98 SE wasn’t that large. The main obstacles that you face come in the form of UEFI and hardware driver support.

Both ThinkPad laptops have in common that they support UEFI-CSM mode, also known as ‘classical BIOS’, as UEFI boot wasn’t even a glimmer yet in some drunk engineer’s eye when Win98 was released. After this everything is about getting as many hardware drivers scrounged together as possible.

[Yeo] ended up having to bodge on a USB 2.0 expansion card via a Thunderbolt dock as Win98 doesn’t have xHCI (USB 3.0) support. With that issue successfully bodged around using a veritable tower of adapters, installing Windows 98 was as easy as nuking Secure Boot in the BIOS, enabling UEFI-CSM along with Thunderbolt BIOS assist mode and disable Kernel DMA protection.

Because UEFI-CSM implementations tend to be buggy, the CREGFIX DOS driver was used to smooth things over. Another issue is the same that we chuckled about back in the day, as Windows 98 cannot address more than 512 MB of RAM by default. Fortunately patches by [Rudolph Loew] helped to fix this and some other smaller issues.

Unfortunately neither Intel nor NVIDIA have released Win98 drivers for quite some time, so there’s no graphics acceleration beyond basic VESA support and the SoftGPU driver. Disk access goes via the BIOS too rather than using an NVMe driver, so it’s not as zippy as it could be, but for Win9x it’s quite usable.

Finally ACPI wasn’t recognized by Win98, but it’s only fair to blame that on the complete flaming train wreck that is ACPI rather than anything to do with Windows. This particular issue was worked around by configuring the BIOS to support S3 power state and with that making Win98 happy again.

It’s honestly quite a shame that UEFI-CSM is largely ignored by new systems, as it makes installing even Windows 7 basically impossible, and thus creating probably the largest split within the x86 ecosystem since the arrival of AMD64/x86_64.

Musician Brian Eno’s Oblique Strategies are like a Tarot card deck full of whimsical ideas meant to break up a creative-block situation, particularly in the recording studio. They’re loads of fun to pick one at random and actually try to follow the advice, as intended, but some of them are just plain good advice for creatives.

One that keeps haunting me is “Honor thy error as a hidden intention”, which basically boils down to taking a “mistake” and seeing where it leads you if you had meant to do it. I was just now putting the finishing touches on this week’s Hackaday Podcast, and noticed that we have been honoring a mistake for the past 350-something shows. Here’s how it happened.

When Mike and I recorded the first-ever podcast, I had no idea how to go about doing it. But I grew up in Nashville, and know my way around the inside of a music studio, and I’ve also got more 1990s-era music equipment than I probably need. So rather than do the reasonable thing, like edit the recording on the computer, we recorded to an archaic Roland VS-880 “Digital Studio” which is basically the glorified descendant of those old four-track cassette Portastudios.

If you edit audio in hardware, you can’t really see what you’re doing – you have to listen to it. And so, when I failed to notice that Mike and I were saying “OK, are you ready?” and “Sure, let’s go!”, it got mixed in with the lead-in music before we started the show off for real. But somehow, we said it exactly in time with the music, and it actually sounded good. So we had a short laugh about it and kept it.

And that’s why, eight years later, we toss random snippets of conversations into the intro music to spice it up. It was a mistake that worked. Had we been editing on the computer, we would have noticed the extra audio and erased it with a swift click of the mouse, but because we had to go back and listen to it, we invented a new tradition. Honor thy error indeed.

MasterClass has dropped the price of its annual memberships by 50 percent, making it a more affordable way to explore its full library of courses across cooking, business, photography, writing and more. With the discount applied, you’ll spend $10 per month when billed annually for the Premium tier, with lower-cost Standard and Plus options available depending on how many devices you want to use at once and whether you need offline viewing.

There are more than 200 classes on MasterClass now, and many are led by big names at the top of their fields. Depending on your interests, you might pick up cooking tips from Gordon Ramsay, learn storytelling from Margaret Atwood, explore business strategy with Richard Branson or get insights into performance and mindset from athletes and entertainers. The catalog spans everything from film and TV to wellness, music, science, leadership and photography, so it’s easy to dip in and find something that holds your attention.

MasterClass

Get 50 percent off all MasterClass subscription tiers.

Advertisement

Over the years, MasterClass has expanded into a broad learning platform that feels like part streaming service and part educational library. The catalog spans categories like food, film, music, wellness, sports and entrepreneurship, with lessons designed to be short and easy to follow. Since every plan includes access to the full course library, the choice mostly comes down to convenience. The Standard tier supports one device and doesn’t include offline mode, the Plus plan allows streaming on two devices with downloads, and the Premium tier increases that to six devices with offline access as well.

Classes are structured to be approachable whether you want to learn a new skill or just explore a topic out of curiosity. Lessons are broken into bite-sized segments, so it’s easy to watch one or two at a time on a commute or in the evening. Members also get access to a global community, occasional newsletters and the ability to switch between video and audio listening on supported classes.

If you’ve been considering trying MasterClass, this deal makes it a more affordable way to see if it works for you. A subscription can also double as a thoughtful gift, which is one reason it has appeared in Engadget’s roundup of the best subscription gifts to send to loved ones. With the current 50 percent discount applied across all tiers, it’s a relatively low-cost way to get access to a large library of professionally produced courses that you can watch at your own pace throughout the year.

I used to be a big SmartThings user back in the day. Powered by its hub, it was the powerhouse of smart home automation, but since Samsung has moved away from the Classic app and the hub is now made by Aeotec, it has lost some of its more powerful features, such as WebCore for scripting.

I use Amazon Alexa for voice commands, as it’s the best voice assistant and has the widest range of devices, and I have been using Apple Home for Automations, as the devices I want to use are either compatible or can be added via HomeBridge.

As good as both platforms are (I particularly like the simplicity of Apple Home for quick device control), what’s missing is the power of the old SmartThings. That’s why I’m moving my automations to Homey.

MicroVision, a solid-state sensor technology company located in Redmond, Wash., says it has designed a solid-state automotive lidar sensor intended to reach production pricing below US $200. That’s less than half of typical prices now, and it’s not even the full extent of the company’s ambition. The company says its longer-term goal is $100 per unit. MicroVision’s claim, which, if realized, would place lidar within reach of advanced driver-assistance systems (ADAS) rather than limiting it to high-end autonomous vehicle programs. Lidar’s limited market penetration comes down to one issue: cost.

Comparable mechanical lidars from multiple suppliers now sell in the $10,000 to $20,000 range. That price roughly tenfold drop, from about $80,000, helps explain why suppliers now are now hopeful that another steep price reduction is on the horizon.

For solid-state devices, “it is feasible to bring the cost down even more when manufacturing at high volume,” says Hayder Radha, a professor of electrical and computer engineering at Michigan State University and director of the school’s Connected & Autonomous Networked Vehicles for Active Safety program. With demand expanding beyond fully autonomous vehicles into driver-assistance applications, “one order or even two orders of magnitude reduction in cost are feasible.”

“We are focused on delivering automotive-grade lidar that can actually be deployed at scale,” says MicroVision CEO Glen DeVos. “That means designing for cost, manufacturability, and integration from the start—not treating price as an afterthought.”

MicroVision’s Lidar System

Tesla CEO Elon Musk famously dismissed lidar in 2019 as “a fool’s errand,” arguing that cameras and radar alone were sufficient for automated driving. A credible path to sub-$200 pricing would fundamentally alter the calculus of autonomous-car design by lowering the cost of adding precise three-dimensional sensing to mainstream vehicles. The shift reflects a broader industry trend toward solid-state lidar designs optimized for low-cost, high-volume manufacturing rather than maximum range or resolution.

Before those economics can be evaluated, however, it’s important to understand what MicroVision is proposing to build.

The company’s Movia S is a solid-state lidar. Mounted at the corners of a vehicle, the sensor sends out 905-nanometer-wavelength laser pulses and measures how long it takes for light reflected from the surfaces of nearby objects to return. The arrangement of the beam emitters and receivers provides a fixed field of view designed for 180-degree horizontal coverage rather than full 360-degree scanning typical of traditional mechanical units. The company says the unit can detect objects at distances of up to roughly 200 meters under favorable weather conditions—compared with the roughly 300-meter radius scanned by mechanical systems—and supports frame rates suitable for real-time perception in driver-assistance systems. Earlier mechanical lidars, used spinning components to steer their beams but the Movia S is a phased-arraysystem. It controls the amplitude and phase of the signals across an array of antenna elements to steer the beam. The unit is designed to meet automotive requirements for vibration tolerance, temperature range, and environmental sealing.

MicroVision’s pricing targets might sound aggressive, but they are not without precedent. The lidar industry has already experienced one major cost reset over the past decade.

“Automakers are not buying a single sensor in isolation… They are designing a perception system, and cost only matters if the system as a whole is viable.” –Glen DeVos, MicroVision

Around 2016 and 2017, mechanical lidar systems used in early autonomous driving research often sold for close to $100,000. Those units relied on spinning assemblies to sweep laser beams across a full 360 degrees, which made them expensive to build and difficult to ruggedize for consumer vehicles.

“Back then, a 64-beam Velodyne lidar cost around $80,000,” says Radha.

Comparable mechanical lidars from multiple suppliers now sell in the $10,000 to $20,000 range. That roughly tenfold drop helps explain why suppliers now believe another steep price reduction is possible.

“For solid-state devices, it is feasible to bring the cost down even more when manufacturing at high volume,” Radha says. With demand expanding beyond fully autonomous vehicles into driver-assistance applications, “one order or even two orders of magnitude reduction in cost are feasible.”

Solid-State Lidar Design Challenges

Lower cost, however, does not come for free. The same design choices that enable solid-state lidar to scale also introduce new constraints.

“Unlike mechanical lidars, which provide full 360-degree coverage, solid-state lidars tend to have a much smaller field of view,” Radha says. Many cover 180 degrees or less.

That limitation shifts the burden from the sensor to the system. Automakers will need to deploy three or four solid-state lidars around a vehicle to achieve full coverage. Even so, Radha notes, the total cost can still undercut that of a single mechanical unit.

What changes is integration. Multiple sensors must be aligned, calibrated, and synchronized so their data can be fused accurately. The engineering is manageable, but it adds complexity that price targets alone do not capture.

DeVos says MicroVision’s design choices reflect that reality. “Automakers are not buying a single sensor in isolation,” he says. “They are designing a perception system, and cost only matters if the system as a whole is viable.”

Those system-level tradeoffs help explain where low-cost lidar is most likely to appear first.

Most advanced driver assistance systems today rely on cameras and radar, which are significantly cheaper than lidar. Cameras provide dense visual information, while radar offers reliable range and velocity data, particularly in poor weather. Radha estimates that lidar remains roughly an order of magnitude more expensive than automotive radar.

But at prices in the $100 to $200 range, that gap narrows enough to change design decisions.

“At that point, lidar becomes appealing because of its superior capability in precise 3D detection and tracking,” Radha says.

Rather than replacing existing sensors, lower-cost lidar would likely augment them, adding redundancy and improving performance in complex environments that are challenging for electronic perception systems. That incremental improvement aligns more closely with how ADAS features are deployed today than with the leap to full vehicle autonomy.

MicroVision is not alone in pursuing solid-state lidar, and several suppliers including Chinese firms Hesai and RoboSense and other major suppliers such as Luminar and Velodyne have announced long-term cost targets below $500. What distinguishes current claims is the explicit focus on sub-$200 pricing tied to production volume rather than future prototypes or limited pilot runs.

Some competitors continue to prioritize long-range performance for autonomous vehicles, which pushes cost upward. Others have avoided aggressive pricing claims until they secure firm production commitments from automakers.

That caution reflects a structural challenge: Reaching consumer-level pricing requires large, predictable demand. Without it, few suppliers can justify the manufacturing investments needed to achieve true economies of scale.

Evaluating Lidar Performance Metrics

Even if low-cost lidar becomes manufacturable, another question remains: How should its performance be judged?

From a systems-engineering perspective, Radha says cost milestones often overshadow safety metrics.

Advertisement

“The key objective of ADAS and autonomous systems is improving safety,” he says. Yet there is no universally adopted metric that directly expresses safety gains from a given sensor configuration.

Researchers instead rely on perception benchmarks such as mean Average Precision, or mAP, which measures how accurately a system detects and tracks objects in its environment. Including such metrics alongside cost targets, says Radha, would clarify what performance is preserved or sacrificed as prices fall.

IEEE Spectrum has covered lidar extensively, often focusing on technical advances in scanning, range, and resolution. What distinguishes the current moment is the renewed focus on economics rather than raw capability

If solid-state lidar can reliably reach sub-$200 pricing, it will not invalidate Elon Musk’s skepticism—but it will weaken one of its strongest foundations. When cost stops being the dominant objection, automakers will have to decide whether leaving lidar out is a technical judgment or a strategic one.

Advertisement

That decision, more than any single price claim, may determine whether lidar finally becomes a routine component of vehicle safety systems.

From Your Site Articles

Related Articles Around the Web

from the they’ll-keep-doing-that-until-there-are-real-consequences dept

Judge Boasberg got his vindication in the frivolous “complaint” the DOJ filed against him, and now he’s calling out the DOJ’s bullshit in the long-running case that caused them to file the complaint against him in the first place: the JGG v. Trump case regarding the group of Venezuelans the US government shipped off to CECOT, the notorious Salvadoran concentration camp.

Boasberg, who until last year was generally seen as a fairly generic “law and order” type judge who was extremely deferential to any “national security” claims from the DOJ (John Roberts had him lead the FISA Court, for goodness’ sake!), has clearly had enough of this DOJ and the games they’ve been playing in his court.

In a short but quite incredible ruling, he calls out the DOJ for deciding to effectively ignore the case while telling the court to “pound sand.”

On December 22, 2025, this Court issued a Memorandum Opinion finding that the Government had denied due process to a class of Venezuelans it deported to El Salvador last March in defiance of this Court’s Order. See J.G.G. v. Trump, 2025 WL 3706685, at *19 (D.D.C. Dec. 22, 2025). The Court offered the Government the opportunity to propose steps that would facilitate hearings for the class members on their habeas corpus claims so that they could “challenge their designations under the [Alien Enemies Act] and the validity of the [President’s] Proclamation.” Id. Apparently not interested in participating in this process, the Government’s responses essentially told the Court to pound sand.

From a former FISC judge—someone who spent years giving national security claims every benefit of the doubt—”pound sand” is practically a primal scream.

Due to this, he orders the government to work to “facilitate the return” of these people it illegally shipped to a foreign concentration camp (that is, assuming any of them actually want to come back).

Believing that other courses would be both more productive and in line with the Supreme Court’s requirements outlined in Noem v. Abrego Garcia, 145 S. Ct. 1017 (2025), the Court will now order the Government to facilitate the return from third countries of those Plaintiffs who so desire. It will also permit other Plaintiffs to file their habeas supplements from abroad.

Boasberg references the Donald Trump-led invasion of Venezuela and the unsettled situation there for many of the plaintiffs. He points out that the lawyers for the plaintiffs have been thoughtful and cautious in how they approach this case. That is in contrast to the US government.

Plaintiffs’ prudent approach has not been replicated by their Government counterparts. Although the Supreme Court in Abrego Garcia upheld Judge Paula Xinis’s order directing the Government “to facilitate and effectuate the return of” that deportee, see 145 S. Ct. at 1018, Defendants at every turn have objected to Plaintiffs’ legitimate proposals without offering a single option for remedying the injury that they inflicted upon the deportees or fulfilling their duty as articulated by the Supreme Court.

Boasberg points to the Supreme Court’s ruling regarding Kilmar Abrego Garcia, saying that it’s ridiculous that the DOJ is pretending that case doesn’t exist or doesn’t say what it says. Then he points out that the DOJ keeps “flagrantly” disobeying courts.

Against this backdrop, and mindful of the flagrancy of the Government’s violations of the deportees’ due-process rights that landed Plaintiffs in this situation, the Court refuses to let them languish in the solution-less mire Defendants propose. The Court will thus order Defendants to take several discrete actions that will begin the remedial process for at least some Plaintiffs, as the Supreme Court has required in similar circumstances. It does so while treading lightly, as it must, in the area of foreign affairs. See Abrego Garcia, 145 S. Ct. at 1018 (recognizing “deference owed to the Executive Branch in the conduct of foreign affairs”)

Even given all this, the specific remedy is not one that many of the plaintiffs are likely to accept: he orders that the US government facilitate the return of any of those who want it among those… not in Venezuela. But, since most of them were eventually released from CECOT into Venezuela, that may mean that this ruling doesn’t really apply to many men. On top of that Boasberg points out that anyone who does qualify and takes up the offer will likely be detained by immigration officials upon getting here. But, if they want, the US government has to pay for their plane flights back to the US. And, in theory, the plaintiffs should then be given the due process they were denied last year.

Plaintiffs also request that such boarding letter include Government payment of the cost of the air travel. Given that the Court has already found that their removal was unlawful — as opposed to the situation contemplated by the cited Directive, which notes that “[f]acilitating an alien’s return does not necessarily include funding the alien’s travel,” Directive 11061.1, ¶ 3.1 (emphasis added) — the Court deems that a reasonable request. It is unclear why Plaintiffs should bear the financial cost of their return in such an instance. See Ms. L. v. U.S. Immig. & Customs Enf’t (“ICE”), 2026 WL 313340, at *4 (S.D. Cal. Feb. 5, 2026) (requiring Government to “bear the expense of returning these family units to the United States” given that “[e]ach of the removals was unlawful, and absent the removals, these families would still be in the United States”). It is worth emphasizing that this situation would never have arisen had the Government simply afforded Plaintiffs their constitutional rights before initially deporting them.

I’m guessing not many are eager to re-enter the US and face deportation again. Of course, many of these people left Venezuela for the US in the first place for a reason, so perhaps some will take their chances on coming back. Even against a very vindictive US government.

The frustrating coda here is the lack of any real consequences for DOJ officials who treated this entire proceeding as a joke—declining to seriously participate and essentially daring the court to do something about it. Boasberg could have ordered sanctions. He didn’t. And that’s probably fine with this DOJ, which has learned that contempt for the courts carries no real cost.

Unfortunately, that may be the real story here. Judge gets fed up, once again, with a DOJ that thumbs its nose at the court, says extraordinary things in a ruling that calls out the DOJ’s behavior… but does little that will lead to actual accountability for those involved, beyond having them “lose” the case. We’ve seen a lot of this, and it’s only going to continue until judges figure out how to impose real consequences for DOJ lawyers for treating the court with literal contempt.

Filed Under: cecot, donald trump, due process, el salvador, james boasberg, pam bondi, pound sand, trump administration, venezuela