Tech
Software, devices and AI, with a short English summary and a link to the original article. RSS
-
Tech
No Man's Sky arrives on iPhone and iPad after its long comeback
Ten years after its troubled launch, No Man's Sky is coming to Apple devices. It will run on iPhone 11 or newer and on iPad. The game has become one of the best known comeback stories in the industry, thanks to years of free updates. Now its whole universe fits in a pocket.
What it means: The story of No Man's Sky is now a textbook case in the industry: a game that launched with far less than its makers had promised, took heavy criticism and, instead of being abandoned, kept improving for a decade through free updates until it won players back. Its arrival on iPhone and iPad also shows how far mobile chips have come, since a title built for consoles and PCs can now run on a pocket device, and support from the iPhone 11 onwards covers a large audience. For software makers more broadly, the lesson is about the relationship with users: steady, visible improvement after launch can change how a product is seen, even when the first impression was poor, as long as there is transparency and a reliable update rhythm. For Apple, titles like this strengthen its effort to present its devices as serious gaming platforms, which in turn influences how publishers decide to invest in the ecosystem.
-
Tech
Meta's Muse Wonder brings it back to the AI frontier
Meta unveiled Muse Wonder, its largest model so far. According to Wired, its results rival the best models on the market. It also includes new safeguards to keep AI agents within their instructions. After a stretch where it seemed to fall behind, Meta is back in the race.
What it means: Muse Wonder means the top of the language model market gains another serious player, after a period in which Meta seemed to trail its main rivals, and that alone shifts the balance on pricing and availability. The company's emphasis on mechanisms that keep AI agents within their instructions shows where competition is moving: it is no longer enough for a model to answer questions well, it has to carry out sequences of actions predictably without drifting when it meets unexpected content. For businesses and software teams, more strong vendors mean more bargaining power and less dependence on a single supplier, but only if applications are designed so the model can be swapped easily behind a stable software layer. The practical advice is to test new models on the business's real scenarios before any switch, because general benchmarks rarely predict accurately how a system will behave on your own data and workflows.
-
Tech
Google's Fitbit Edge sits between a tracker and a smartwatch
The new Fitbit Edge slots in between the Pixel Watch 5 and the Fitbit Air. It costs $179 and has an OLED screen, smart features borrowed from the Pixel Watch and roughly a week of battery life. The battery can even be replaced. It is not a full smartwatch but targets people who want more than a basic tracker.
What it means: The Fitbit Edge shows Google trying to fill a gap many users recognise: a full smartwatch often needs charging every day and costs a fair amount, while a basic tracker lacks the screen and smart features that make a difference in daily life, such as notifications and apps. At 179 dollars, with an OLED display and roughly a week of battery life, the Edge targets people who want health tracking and a few core features without the bulk and price of a Pixel Watch. The most interesting detail is the replaceable battery, because in wearables the battery is usually the reason a device is retired after two or three years, and being able to swap it extends the product's life and cuts electronic waste. The move fits European pressure for repairable devices, and if it sells well it could push other makers to follow, which would help consumers more than yet another bump in specifications.
-
Tech
US investigators used fake cell towers to locate a journalist's producer
US investigators sought and received warrants to use devices that imitate mobile phone towers to track down a producer working with journalist Don Lemon. A similar request covered journalist Georgia Fort and a Minnesota Senate candidate. These devices also sweep up data from unrelated people nearby. Press freedom advocates are alarmed by the disclosure.
What it means: Devices that imitate cell towers work by tricking nearby phones into connecting to them first, which lets investigators locate a specific handset but also collects data from anyone in the area who has nothing to do with the case. Their use to find people working with journalists and a political candidate moves the debate from technology to the rule of law, because protecting sources is a basic condition for a free press, and the ability to track someone without the carrier's cooperation makes oversight harder. For professionals who handle sensitive information, whether journalists, lawyers or business teams on confidential projects, the practical lesson is that the phone itself is a constant source of location data, and encrypted messaging protects content but not necessarily where you are. In Europe, where GDPR and national laws set stricter limits, cases like this are a reminder that the boundaries of surveillance are decided each time by judicial oversight and transparency rather than by what the technology can do.
-
Tech
Epic Games reportedly shifts to AI-first game development
According to Bloomberg, Epic Games is changing how it builds its joint project with Disney, letting AI models write code and produce art. Developers and artists would move into a role of directing and reviewing generated work. The Marvel and Star Wars themed game, expected in March 2027, has been delayed indefinitely. The move shows how fast AI is reshaping software teams.
What it means: If the Bloomberg report holds up, this is one of the most radical workflow changes at a major studio: developers and artists would no longer produce most of the code and art themselves, but would steer models and review what comes out, which looks more like editing than creating from scratch. The indefinite delay of the joint Disney project is a telling detail, because it suggests the switch is not automatically faster, at least at first, since tools have to be integrated, teams have to learn new processes and quality has to reach a level that can carry a major release. For the software industry in general, the signal is that an AI-first approach is moving from experiments into strategic decisions, with consequences for hiring, roles and production budgets. For smaller teams the calmer reading is to use AI where it already pays off, such as prototypes, tests and repetitive code, while keeping human control over architecture and final quality.
-
Tech
A Claude test run sent a false murder tip to Philadelphia police
Anthropic confirmed that a Claude model mistakenly submitted a false tip about an unsolved murder through the Philadelphia police web form. It happened in July during an internal test where the model was meant to fill in forms without submitting them. The model wrongly expected a confirmation page to follow. The tip had been caught by a spam filter, and the company informed the police in October.
What it means: This incident is one of the clearest examples of what can go wrong when an AI agent has access to the real internet: the model did not intend any harm, it simply assumed that a confirmation page would appear before anything was final, and a test action turned into a genuine report to a police department. The fact that the tip was stopped by a spam filter was luck rather than design, and the months that passed before the police were told show that even large labs do not always have a full picture of what their agents did during testing. For anyone building AI automation the lessons are practical: tests should run in an isolated environment or against fake versions of real forms, every irreversible action such as submitting, paying or sending should require explicit permission, and logs should be reviewed routinely rather than only when something looks suspicious. The same applies to a small business that lets an agent fill in orders or supplier forms, because the model's guess about what comes next is not a safety control.
-
Tech
Amazon now runs more than a million robots
Amazon, the second largest private employer, now uses more than one million robots across its facilities. The company is betting on automation so it can grow without hiring at the same pace. Ars Technica looks at how human jobs in the warehouses are changing. The open question is whether robots will support workers or replace them.
What it means: The one million robot milestone is more than an impressive number: it shows that at the world's largest logistics company automation has become the default way to grow rather than an experiment, with robots moving shelves, sorting parcels and taking over repetitive motions that people used to perform. The question Ars Technica raises is what remains for workers, and the new roles tend to be system supervision, maintenance, exception handling and quality control, jobs that need different skills and not always the same headcount. For smaller businesses the impact is indirect but real, because customer expectations of fast delivery and low cost are shaped by this level of automation, and competing gets harder for anyone still doing things by hand. The realistic response for a small shop or warehouse is not to buy robots, but to automate what it can in software, such as stock control, shipping labels and customer notifications, so that people spend their time on the work that genuinely needs judgement.
-
Tech
US midterm races turn data centers into a campaign issue
In Senate and governor races, candidates are competing over who will be tougher on data centers. Residents worry about power and water use and about rising energy bills. The issue now cuts across party lines. The results will shape where and how fast AI infrastructure gets built.
What it means: Data centers becoming a campaign issue in the US shows how quickly AI has moved from research labs to the average household's power bill: large facilities consume huge amounts of electricity and water, and the people living near them want to know who pays for new transmission lines and whether prices will rise for everyone. When the issue crosses party lines, cloud companies can no longer rely only on tax breaks and promises of jobs, and will need to show that they bring their own power or invest in the local grid. For Europe, and for countries competing for data center investment, the American experience is a useful precedent, since the same questions about grid capacity, cooling water and cost will be asked here too. For businesses that simply buy cloud services, the longer-term risk is that prices end up shaped by where and how fast new facilities are allowed to be built, which is worth factoring into cost planning for the coming years.
-
Tech
AI bots are wasting scammers' time on purpose
Anti-fraud groups are increasingly using AI to turn the tables on scammers. Convincing bots pose as potential victims and keep fraudsters busy in pointless conversations. Every hour spent on a bot is an hour not spent on a real person. The chats also reveal how the scams work, which helps defenders.
What it means: Using bots to waste fraudsters' time is not a new idea, but modern language models make it far more convincing, because they can keep a conversation going for hours, imitate hesitation or confusion and reply in several languages with no human involved. The economics are simple: fraud is a business with limited staff, and every minute a scammer spends on a fake victim is a minute not spent on a pensioner or an unsuspecting bank customer. There is a second, less visible benefit, because those conversations become material for security teams, who learn the scripts, the fake bank details and the pressure tactics in use, and can update filters and warn customers more quickly. For banks, telecom providers and online shops, where text and phone scams keep growing, the message is that defence can and should be automated as fast as attack is, while still respecting data protection rules when those conversations are stored and analysed.
-
Tech
Refusals alone will not keep AI models safe
Today's AI models are trained to turn down many requests, from dangerous instructions to sensitive topics. MIT Technology Review argues that this habit of refusing is not enough as a safety mechanism. Refusals are often bypassed with clever prompts and do not address the deeper risks. Real safety needs several layers, not just a chatbot saying no.
What it means: The MIT Technology Review argument touches something we see in almost every project built on language models: teams rely on the model refusing harmful requests and skip the ordinary controls they would put on any other system, such as per-user permissions, action limits and audit logs. A refusal is a behaviour the model learned from training data rather than a rule written in code, so it can shift with a new model version, a reworded request or instructions hidden inside a document or web page the model reads. For a business putting a chatbot or agent in front of customers, the practical lesson is to give the model access only to what it needs, route sensitive actions such as refunds, account changes or outgoing emails through a separate check or a human approval, and log every tool call so it can be reviewed later. That way, even if someone does manage to trick the model, the damage stays contained, and safety no longer depends on whether a statistical system happens to say no at the right moment.
Summaries are ours. Copyright in the original articles remains with their publishers.
Have a project in mind?
Tell us what you need to achieve. You get a concrete proposal with scope, timeline and cost — no obligation.