If you’re of a certain vintage, you have probably looked at some of the microcomputers on the market these days and thought “that would have been a decent workstation back in the day!”. We certianly have, and so did [Roberto Alsina]. Rather than allow himself to contemplate his age and threfore rapidly approaching mortailty, [Roberto] wrote a useful operating system called ESP-Osito for the Cheap Yellow Display, which he refers to as “the cheapest computer”. He’s not wrong, and it’s certainly a better use of time than an existential crisis.

He explains some of his reasoning behind the project in an accompanying blog post, but on the project page he compares it to a Palm Pilot– it’s on quick, apps load quick, and the API is simple enough for easy app creation in a few hundred lines of C, unlike certain pocket computers we won’t name. Sure, there’s no multitasking, but when apps jump from SD card to run in memory in microseconds, who cares? Saving the current state of the app back to SD means the experience is virtually identical from a user perspective anyway.

As this is a one-man show for now, the app store won’t quite rival your smart phone– but there’s everything you’d expect on the 90s-era computer this has the horsepower of: a serial terminal, a text editor, a file explorer, a calculator, a clock, but also some things that aren’t so retro. The clock app gives weather info via futuristic wireless networking, the reader app takes Markdown text, and the chat app connects to an LLM somewhere instead of your friends on IRC. The blackberry keyboard option gives it a feel of a slightly different vintage. You can also play snake, because no computer is complete without games. The OS and all its applications are released under the MIT license on GitHub, and [Roberto] is actively looking for collaborators.

If you doubt the workstation comparison at the start of this article, this CYD runs Macintosh System 3 via a 68k emulator. That’s got old-school cred, but there’s something great about having retro constraints with modern code on modern hardware. In that way, ESP-Osito is similar to the 3D graphics engine behind this Wipeout clone.

A recent Indian court ruling against Google’s keyword advertising practices has gained fresh attention after founders said competitors have long used the system to siphon off customers and force companies to pay to protect their own brands.

The ruling, delivered by the Delhi High Court on May 22 in a trademark dispute involving bathroom fittings maker Hindware, found Google liable for trademark infringement over its keyword advertising practices and awarded the company ₹3 million (around $31,600) in nominal damages.

In her 163-page judgment (PDF), Justice Mini Pushkarna rejected Google’s argument that it was merely a passive intermediary in serving ads on its search platform. The judge said Google, through its AdWords platform, allowed Hindware’s rivals to use “Hindware” as a keyword to target users searching for the brand.

“Google by selling the trademark of the plaintiff [Hindware] as a keyword without any authorization for commercial gains is infringing the plaintiff’s right to exclusive use of its trademark under Section 28 of the Trade Marks Act,” the judge said.

The judgment drew attention on Friday after Indian entrepreneurs, including Zerodha founder Nithin Kamath and Zoho founder Sridhar Vembu, publicly backed the ruling, arguing that competitors have long used Google’s advertising tools to divert traffic from established brands and force companies to spend money protecting their own names.

Kamath, who said Zerodha had faced the issue for more than a decade, wrote on X: “Whenever someone searches for ‘Zerodha,’ the traffic should rightfully come to Zerodha. But what often happens is that the first couple of results on Google Search are ads, leading the customer to a competitor’s website.”

Google, for its part, said its Ads policy on trademark keywords “does not allow competitor advertisers to use trademarked terms in the ad-text of an ad” and that the policy is applied globally.

“We look forward to continuing to align our operations with local legal frameworks while maintaining strict standards to protect our users’ long-term interests,” a Google spokesperson said in a statement to TechCrunch.

India is a key market for Google, with more internet users than any country other than China, making court decisions affecting its search and advertising businesses particularly significant.

Legal experts, however, said the implications of the ruling may be narrower than some of the public reaction suggests.

“The judgment per se will require platforms to relook at their processes to see if their automated tools encourage or offer trademarked terms to advertisers at large,” said Aprajita Rana, a partner at AZB & Partners.

Nonetheless, Rana told TechCrunch that the decision does not have a “far-reaching impact” on online platforms’ liability in India, as courts have already established that internet companies can lose legal protections when they play an active role in unlawful activity.

“What’s important in this case is how providing access to trademarked terms, even in ad curation that’s between online platforms and advertisers and not known to customers, can amount to a participative activity for platforms,” Rana said.

They Khan’t get away with that.

Plastics, oil, petrol– the modern world is entirely dependent on hydrocarbons. The good sources are slowly running low and supply is increasingly complicated by geopolitical factors we really don’t want to get into, but hey! It’s just hydrogen and carbon, right like it says in the name. How hard could it be to roll your own at home. Well, if you’ve got a lab like [Marb]’s Lab on YouTube, it might just be doable, as he demonstrates in his latest video.

The Fischer-Tropsch reaction was discovered back in 1925 in Germany by a couple of gents named Fischer and Tropsch. In the unpleasantness that followed later, Germany made good use of their process on an industrial scale, since they had ample coal and no oil on hand. Coal-rich South Africa has also made us of it, particularly during the Apartheid-era trade restrictions. Every so often the idea of industrializing the process comes up in the USA, but there’s still enough oil there it doesn’t make sense economically.

Those nations all have something in common: they’re all coal-rich countries, and that makes sense because coal is easily converted to carbon monoxide and hydrogen– a combo known as syngas– and it just so happens that those are the feedstock for this reaction. The actual chemistry going on inside is quite complex, but conceptually it is pretty simple: hydrogen and carbon monoxide mix over a hot metal catalyst, and combine to form various hydrocarbons.

In [Marb]’s glassware-based demonstration, the catalyst is Cobalt (III) Oxide on silica gel– a lovely, cancer-causing substance that must be prepared for each use, as it lasts but 24 hours before further oxidization ruins it. That’s in spite of purging the system with argon– a necessary step if one does not wish to explode. The yield isn’t amazing, and [Marb] isn’t sure exactly what mix of hydrocarbons he has created– although they smell like gasoline and burn like the dickens, so mission accomplished.

This might seem like the furthest thing from green, but if you use solar power to run the process and something like woodgas– which is syngas by any other name– as a feed-stock, then you’ve got a carbon neutral energy storage medium.

Thanks to [Markus Bindhammer] for the tip!

from the hallucinating-your-arrest dept

This article is republished from The Conversation under a Creative Commons license. Read the original article.

In Baltimore County, Maryland on Oct. 20, 2025, a 17-year-old student named Taki Allen was sitting outside his high school after football practice when an artificial intelligence-enhanced surveillance camera falsely identified the Doritos bag in his pocket as a gun. Within moments police cars arrived, officers drew their weapons and Allen was forced to his knees and handcuffed while they searched him. All they found was a crumpled bag of chips. The AI’s misidentification and the human decisions that followed turned a normal evening into a traumatic confrontation.

On Dec. 24, 2025, Angela Lipps, a Tennessee grandmother, was released after spending five months in jail because facial recognition software had incorrectly connected her to fraud crimes in North Dakota, a state she had never visited. Police had arrested her at gunpoint while she was babysitting her four grandchildren.

These are unfortunate examples of how AI can lead to mistreatment of people because of technical flaws as well as misplaced human faith in the technology’s supposed objectivity. These cases involve different tools, but the underlying issue is the same. AI systems produce probabilities, and people treat them as certainties.

We are researchers who study the intersection of technology, law and public administration. In researching how police departments use AI and how digital technologies operate in a democratic society, we have seen how quickly the shift from probabilistic prediction to operational certainty happens in practice.

AI policing tools are used in dozens of U.S. cities, although no public registry tracks the full footprint. The tools ingest historical crime data and score neighborhoods on predicted risk so officers can be routed toward the resulting hot spots. The mechanism is straightforward, but its consequence is not. Once a system signals a possible threat, the question is no longer how certain the prediction is but what to do about it. A statistical output turns into a deployment decision, and the uncertainty that produced it gets lost on the way.

A matter of probabilities

When generative AI models such as ChatGPT or Claude respond to human requests, they are not searching a database and pulling out facts. They are predicting the most likely answer based on patterns in data they have been trained on. When asked, “Who invented the light bulb?” the models do not go to a source or fact-check a finding. They generate a statistically probable answer which is “Thomas Edison.” The reply might be right, but it might not capture the full story – such as Joseph Swan’s parallel invention at the same time as Edison’s. The danger arises when people believe that the model is retrieving truth rather than generating likelihoods.

This distinction matters. The most probable response is not the same as a factually verified answer, complete with context.

This reality can be highly problematic for policing and law. For example, when law enforcement agencies use AI systems trained on geographical data to estimate where criminal activity is likely to occur, the algorithms analyze historical crime data and geographic patterns. These systems generate statistical risk scores or heat maps for locations based on prior incidents. But such predictions may have little bearing on who was involved in a new crime in the area, even if an algorithm generates information that sounds authoritative.

Some researchers have argued that predictive policing systems do not increase the likelihood that racial minorities will be arrested more often relative to traditional policing practices. The broader concern, however, is not limited to measurable disparities in arrest outcomes alone. It is about how probabilistic predictions can become standardized operational decisions absent further verification.

Artificial intelligence researchers caution against using these models in isolation for crime and legal proceedings or decision-making. Research at the University of Virginia’s Digital Technology for Democracy Lab with police chiefs shows that some law enforcement groups follow strict policies that dictate when technology is used in tandem with, or in place of, human discretion, while others have no such policy.

What most users do not realize is that AI systems rarely produce binary answers: yes or no, a positive identification or a negative one. They generate probabilities. Some systems assign scores that assess the system’s confidence in a prediction. In those cases, engineers set a confidence threshold, a level of certainty that determines when the system should trigger an alert about a possible threat. You can think of this threshold as settings on a control knob. A 95% confidence level, for example, indicates that the model considers its interpretation to be highly likely.

A low threshold catches more potential threats but increases false alarms. A high threshold reduces mistakes but risks missing real dangers. Either way, these algorithmic thresholds are often invisible to the public and are set quietly by vendors or agencies, even though they shape when police action begins.

Where to draw the line

In medicine, these kinds of trade-offs are explicit. Diagnostic tools are calibrated on the relative harm of different errors. In infectious disease settings, for instance, systems that detect infections are often designed to accept more false positives to avoid missing contagious individuals. Then medical professionals look into the human cases. And the algorithm-based decisions are subject to professional standards, ethics reviews and regulatory oversight.

In policing, an AI system must balance false positives, where the system flags a threat that does not exist, and false negatives, where it fails to detect a real danger. The trade-off carries significant consequences. A lower threshold may generate more alerts and allow officers to intervene earlier, but it also increases the risk of mistaken identifications, which happened to Angela Lipps, or escalated encounters like the one Taki Allen experienced. A higher threshold may reduce wrongful interventions but could allow legitimate threats to go undetected.

Some law enforcement agencies argue that acting on imperfect signals is preferable to missing serious risks. But lowering the bar for algorithmic alerts based on probabilistic estimates effectively expands the number of people subjected to police attention. It is important to realize that these thresholds are not neutral features of the technology; they are choices embedded by the creators in the model’s code. Decisions about where to draw the line determine when an algorithmic suspicion becomes a real-world police action, even though the public rarely sees or debates how those thresholds are set.

Limits of optimization

Developers often use several methods to determine where to set a confidence threshold. Techniques such as “receiver operating characteristic curve analysis” examine how changing the threshold for an alert alters the balance between correctly identifying real events and mistakenly flagging harmless ones. Precision–recall analysis examines a similar trade-off, asking how accurate the system’s alerts are relative to the number of incidents it successfully detects.

These approaches could help calibrate systems more responsibly by testing how often an algorithm wrongly flags people or locations. Fine-tuning can improve system performance. But the techniques cannot resolve the underlying question of how much algorithmic uncertainty society is willing to tolerate.

In law, legal standards of proof determine how convincing evidence must be before a judge or jury can rule in favor of a plaintiff or defendant. Courts use formal standards of proof depending on the stakes, such as probable cause, preponderance of the evidence and beyond a reasonable doubt. These standards reflect a societal judgment about how much uncertainty is acceptable before exercising legal authority. A court does not accept a guess or a prediction; it follows a process to weigh evidence. Unlike humans, an AI model does not usually say, “I’m not sure.” A model typically has confidence in its reply, even when the answer is incorrect.

Stakes are rising as AI enters the courtroom, law enforcement, the classroom, the doctor’s office and the public sector. It is important for people to understand that AI does not know things the way many assume it does. It does not distinguish between “maybe” and “definitely.” That is up to us. We believe that technologists should design systems that admit uncertainty and need to educate users about how to interpret AI outputs responsibly.

Maria Lungu is a Postdoctoral Researcher of Law and Public Administration at University of Virginia and Steven L. Johnson, is Associate Professor of Commerce at University of Virginia

Filed Under: ai, arrests, policing, wrongful arrest

Sidephone has built an Android phone around a swappable USB keypad, entering a niche between the distraction-heavy smartphone and the stripped-back dumbphone by offering physical input options without sacrificing app compatibility.

We saw the device at Beyond Expo 2026, where its keypad system came in two layouts, a QWERTY configuration for text-heavy users and a numpad layout for those who prefer a more traditional input style, with a growing selection of cases and colour options.

That restraint extends to the camera, which prioritises simplicity over the high-resolution computational photography that dominates modern flagships, producing shots designed to be printed and kept rather than shared online.

Running Android without Google or Google Play services, the device supports thousands of existing apps that function independently of Google’s infrastructure, covering essentials like calls, SMS, camera, files, and a gallery app out of the box.

Advertisement

Advertisement

Privacy-focused app ecosystem

Beyond the defaults, Sidephone ships with several optional app bundles focused on privacy and open-source software, including the full Proton suite covering mail, VPN, calendar, drive, and password management, as well as Signal, WhatsApp, and a browser choice between Vivaldi and Firefox.

The Essentials bundle adds Maps powered by HERE, LocalSend for cross-platform file sharing, a local music player via Auxio, and a voice recorder, while the Misc bundle extends further with AntennaPod for podcasts, BreezyWeather, and Obtanium for installing apps directly from developer release pages.

Support for third-party messaging apps, including Telegram, Threema, and WeChat, is also listed as a work in progress, alongside compatibility with music streaming services such as Apple Music, Spotify, and Tidal.

For anyone considering making the switch, more information and ordering are available on the Sidephone website, with the broader community active on Reddit at r/Sidephone.

Advertisement