DeepRare, an agentic AI system integrating 40 specialised tools, outperformed medical specialists in identifying rare conditions in a head-to-head study published in Nature.

For millions of people with rare diseases, the path to diagnosis is a labyrinth. Patients bounce between generalist GPs and specialists across years, sometimes decades, piecing together symptoms that fall outside textbook presentations.

Eighty per cent of rare diseases have a genetic origin, yet most go undiagnosed until too much biological damage has occurred. The bottleneck is not lack of data, it’s finding the needle in the medical haystack.

A new study published in Nature this month suggests that artificial intelligence may accelerate that hunt. Researchers at Shanghai Jiao Tong University’s School of Artificial Intelligence and Xinhua Hospital developed DeepRare, an AI system designed to mimic how human doctors reason through diagnostic uncertainty.

In a head-to-head comparison with five experienced physicians, each with more than a decade of practice, the system achieved higher accuracy across the board.

The numbers are striking. DeepRare correctly identified the disease on its first suggestion 64.4 per cent of the time, compared to 54.6 per cent for the doctors. When given three suggestions instead of one, the AI system achieved diagnostic success in 79 per cent of cases versus 66 per cent for the human specialists.

Crucially, the physicians endorsed the AI’s reasoning 95.4 per cent of the time, suggesting the system not only reaches correct conclusions, but does so in ways that experienced clinicians find persuasive and medically sound.

What distinguishes DeepRare from earlier diagnostic AI is its architecture. Rather than applying a black-box classification model, the system integrates 40 specialised digital tools and follows an explicitly reasoned workflow.

It forms diagnostic hypotheses, tests them against patient evidence, searches global medical literature databases, analyses genetic variants, and revises its conclusions iteratively before ranking possibilities.

The process mirrors the cognitive steps a human diagnostician takes, but with access to the entirety of medical knowledge and computational speed humans cannot match.

The system has already moved beyond the laboratory. Since July 2025, DeepRare has been deployed on an online diagnostic platform, with more than 600 medical institutions worldwide registered to use it.

The research team plans to validate the system further using 20,000 real-world cases and to launch a global rare disease diagnostic alliance. Notably, the authors emphasise that the system is not intended to replace clinicians, but to augment diagnostic workflows, a safeguard that acknowledges both the technical limits of AI and the irreducible human element in medicine.

The implications for patients are profound. Approximately 300 million people worldwide are affected by rare diseases, and the average diagnostic odyssey stretches to five years or longer.

Each year of diagnostic delay is a year of uncertainty, wrong treatments, and accumulating organ damage. An AI system that can trim weeks or months from that timeline, and surface possibilities that might otherwise be overlooked, could reshape the early experience of living with a rare condition.

• University of Missouri researchers claim DNA hard drives can store, erase, and rewrite repeatedly
• Frameshift encoding converts binary data into DNA sequences for molecular storage
• Nanopore sensors read DNA sequences by detecting subtle electrical signal changes

The University of Missouri has announced progress on what it calls a “DNA hard drive,” claiming it can store, erase, and rewrite information repeatedly.

Unlike conventional HDDs or cloud storage, which rely on magnetic or solid-state media, this approach leverages the molecular stability of DNA.

With the RAM and storage crisis hitting personal computing very hard – along with new software increasingly suffering the effects of metastasizing ‘AI’ – more people than ever are pining for the ‘good old days’. For example, using that early 2000s desktop PC with Windows 98 SE might now seem to be a viable alternative in 2026, because it couldn’t possibly make things worse. Or could it? As a reality check, [SteelsOfLiquid] over on YouTube gave this setup a whirl.

The computer of choice is a very common Dell Dimension 2100, featuring a zippy 1.1 GHz Intel Celeron, 256 MB  of DDR1, and a spacious 38 GB HDD. Graphics are provided by the iGPU in the Intel i810 chipset, all in a compact, 6.9 kg light package. As an early Windows XP PC, this gives Windows 98 SE probably a pretty solid shot at keeping up with the times. At least the early 2000s, natch.

Of course, there is a lot of period-correct software you can install, such as Adobe Photoshop 5, MS Office 97 (featuring everyone’s beloved Clippy), but a lot of modern software also runs, with the Retro Systems Revival blog documenting many that still run on Win98SE in some manner, including Audacity 2.0. This makes it totally suitable for basic productivity things.

Gaming on Win98 is naturally limited to games from around that early 2000s time period or before, but the gaming library even for just Win98 and MS-DOS is pretty massive, so as long as you’re fine not playing the latest and greatest games, this is also pretty easy.

Where things get dicey is of course with using the modern Internet, as you need a modern browser and support for the latest TLS encryption features to not have many websites throw a hissy fit. Using Frog Find and similar proxies that target retro computing help here, fortunately.

Previously we covered ways that you can use Discord even on Windows 95 and Windows NT 3.1, others have ported .NET applications to Windows 9x, got Win98 up and running on a 2020-era system, and you can totally use modern YouTube in even the Netscape 2.x browser using an NPAPI plugin.

Although there are many arguments to be made for using at least a Windows version with an NT kernel over the 9x one, it’s hard to deny that software Back Then™ was less complex, less resource-hungry and still got all the things done. Maybe it is worth another look, before the AI Crisis forces us all back on Windows XP systems like the one featured in this video.

from the working-for-the-weekend dept

For many years now we’ve had two regular posts that come out on the weekends: our This Week In Techdirt History posts on Saturdays, and our Funniest/Most Insightful Comments Of The Week posts on Sundays. Sometimes we switch it up a little bit, replacing the history post with a special promotion or (as will be happening again soon!) with our Winner Spotlight posts for our annual public domain game jam.

This has been rolling along smoothly and we don’t plan any major changes, but we occasionally discuss whether there are tweaks we should make, and since I’m on vacation this week (and thus not around to write today’s comments post) we thought it would be a good time to check in with our readers and find out what you think.

In general, the weekend posts don’t get a lot of traffic, though that’s never been our reason for doing them anyway: they are for our community of commenters and long-time readers who we know appreciate them. Nevertheless, the relatively small audience for these posts is why we try not to expend too much time and effort creating them every week, as we’re all very busy around here and there are always a lot of projects on the go!

So with that in mind, we want to know: do folks like these posts? Is there stuff we can do to improve or refresh them going forward? Is there something else entirely you’d like to see as a weekly feature? Sound off in the comments, and we’ll be discussing the possibilities over the coming weeks.

There’s one change that we’ve already talked about and will likely be rolling out after the upcoming run of game jam winner spotlight posts: since Techdirt’s history has grown so long, and since frankly nobody wants to be repeatedly confronted with the fact that 2021 is already five years behind us, we’re going to ditch the “five years ago” section in the history round-ups and instead look at posts from ten, fifteen, and twenty years ago. Next year when we celebrate our 30th birthday, we might even change that up again to ten, twenty, and thirty year sections. Reaching further back just seems more fun and interesting than reminiscing on the recent past.

Also on the history posts, there’s a question I’d love some reader feedback on: is the “paragraph format” with a brief summary of some posts from the week worth keeping? Or would it be just as good (perhaps better) to go with just a bullet-list of selected headlines? Or something else?

As for the comments posts, these are our longest-running and most important tradition, as we’ve always valued our community of commenters and we want to highlight your contributions. However, as you know if you read them every week, lately it’s often been a struggle to populate the “funny” side of the list, and admittedly the quality of the Editor’s Choice selections can vary a lot, as it depends on how much time we (and especially I) have had to follow and engage in the comments on a given week.

So again we want your feedback: should anything change about the comment posts? Should we pivot away from Editor’s Choice and instead feature the top three winners-by-vote in each section (maybe retaining a single Editor’s Choice comment on each side, or just one for the week?) And how should we handle it when it’s a slow commenting week and there just aren’t enough highly-voted comments to make a post? (On especially slow weeks on the funny side, sometimes the “winners” are really just the chance recipients of a trickle of idle votes, and not really comments that make any sense to feature).

There’s also a similar question about the comments posts as there was about the history posts: is the brief written summary (in which I try to explain the context of a comment if and when it’s absolutely necessary to understand the comment itself) worthwhile? Or would a more dry bullet list that simply quotes and links to the winning comments and their authors suffice, leaving readers to go check out the full context as they wish?

Beyond these specific questions, we’re open to any and all thoughts on how the weekend posts should change (if they should change at all) and suggestions about anything else you might like to see. We’re not in a rush to make any changes, but it’s worth getting the conversation started. Let us know in the comments. (And if any suggestions are especially popular and get lots of votes, maybe they’ll turn up in next week’s winning comments post!)