📰 English IT Daily · 2026-06-26

A developer created a public experiment called hackmyclaw.com to test the security of his AI email assistant, Fiu. The challenge was simple: people could email the assistant and try to make it reveal the contents of a secrets.env file, which stored sensitive information. After the project appeared on Hacker News, Fiu received more than 6,000 emails from over 2,000 people. According to the developer, the secret was never leaked and the assistant never sent an unauthorized reply.

The test focused on prompt injection, a type of attack that tries to manipulate an AI system through carefully written instructions. Fiu had access to email and had the technical ability to reply, but it was told not to do so. It also had a basic security prompt that blocked actions such as revealing credentials, changing its own files, running code from emails, or sending data to outside services. The goal was to see whether attackers could bypass these rules with social engineering or other tricks.

Many of the attacks were creative. Some emails pretended to come from an administrator or described a fake emergency, compliance audit, or incident response case. Others used different languages or sent many message variations in a short time. The developer said some attempts were surprisingly sophisticated. However, the experiment also showed practical problems. Google temporarily suspended the assistant’s Gmail account because the unusual traffic looked suspicious, and the project produced more than $500 in API costs.

The developer also learned that the testing setup mattered. When emails were processed in batches, early attack messages made the assistant more suspicious of later ones. He changed the system so each email was handled in a fresh context. He also found that the assistant had started to understand that it was part of a security exercise, which could affect the results. Even with these issues, the experiment suggests that simple defensive rules can help, but AI assistants still need careful design, isolation, and monitoring before they are trusted with sensitive tools and data.

IBM has announced what it calls the world’s first sub-1 nanometer chip technology. The new design is based on a 0.7 nanometer, or 7 angstrom, process node. This is an important step for the semiconductor industry because chipmakers are getting close to the physical limits of traditional scaling, the long-term method of making chips smaller and more powerful.

According to IBM, the chip uses a new transistor architecture called “nanostack.” In simple terms, a transistor is a tiny switch that controls electrical signals inside a chip. Nanostack places these transistors in stacked 3D layers instead of only spreading them across a flat surface. This allows more transistors to fit into a very small area. IBM said the chip contains nearly 100 billion transistors on a chip about the size of a fingernail.

IBM says this approach could deliver major gains in both performance and energy efficiency. Reported technical results suggest up to 50% higher performance or 70% better energy efficiency than IBM’s earlier 2 nanometer technology. The company also said the design can help with demanding workloads such as generative AI, cloud infrastructure, and future electronic devices that need more computing power while using less energy.

The company presented the breakthrough as more than just another reduction in size. IBM said nanostack changes how chips are built by using three-dimensional integration and different material combinations in each layer. Researchers also reported results showing that the design can support real computation. If the technology can be developed further for manufacturing, it may influence the next decade of hardware design and help the industry continue progress at extremely small scales.

A website called “web hl2” shows that Half-Life 2 can run in a browser. The page includes a download process and a fullscreen button, suggesting that the game is being delivered through web technology instead of a normal desktop install. For many people, this is a surprising example of how far the modern browser has developed.

Running a large 3D game in a browser is technically interesting because browsers were once mainly used for reading pages and watching simple media. Today, they can support complex graphics, audio, and input from a keyboard or mouse. In simple terms, the browser acts like a software platform, or a base environment, where developers can load and run interactive programs.

This kind of project also shows the value of portability. Portability means software can work in different environments with fewer changes. If a game can run through a browser, users may not need to install it in the usual way on every machine. However, performance still matters. A browser version must manage memory, graphics, and loading time carefully to give users a smooth experience.

For programmers, the project is a useful case study in web delivery and user experience. It raises questions about how older software can be adapted for newer platforms, and how much work is needed to make demanding applications usable online. Even when details are limited, the idea is clear: the browser is no longer just a document viewer. It is becoming a serious runtime for advanced software.

A report in Science says some medical students are using a popular research tool to produce studies that may be misleading. The concern is not only about poor writing. Experts worry that the tool can make weak methods look professional and convincing. In medicine, this is especially serious because published research can influence patient care, education, and future studies.

The tool helps users search academic papers, organize sources, and quickly build a literature review, which is a summary of earlier research on a topic. These features can save time for honest researchers. However, critics say the same system can also be used to create papers that appear carefully designed even when the evidence is limited, selective, or poorly checked. As a result, readers may trust conclusions that are not reliable.

Researchers and journal editors are concerned about the publication process. Medical students often face pressure to publish, and easy-to-use software may lower the barrier to producing large amounts of academic work. If supervisors, reviewers, or journals do not examine the underlying data and methods closely, misleading studies can enter the research record. Once published, such papers may be cited by others and spread doubtful claims further.

The debate is not about banning digital tools. Many scientists say software can improve productivity and support real discovery when it is used responsibly. Instead, they call for stronger review standards, better training in research ethics, and more transparency about how studies are created. The issue also reflects a wider challenge in technology: tools that increase efficiency can also scale low-quality output unless institutions build safeguards and accountability.

Open source software has supported modern technology for many years. It is used in systems that people rely on every day, including banking, telecommunications, and utilities. Many companies build their products on the same shared libraries and tools. Because of this, a weakness in one important project can affect many services at the same time.

The open letter about Akrites says that artificial intelligence has changed software security. In the past, finding a serious vulnerability in a major open source project could take an expert a long time. Now AI tools can help discover weaknesses much faster, sometimes finding several problems in one attempt. This can help defenders improve software, but it can also help attackers work at greater speed.

According to the letter, this faster discovery cycle is creating pressure on maintainers, the people who manage open source projects. If many organizations scan the same software and send separate reports, maintainers may receive too much noise. This can slow remediation, which means fixing a security problem, and it may increase the risk that information about an unpatched flaw leaks before a safe fix is ready.

Akrites is presented as a coordinated effort to improve security in critical open source software. The plan is to create trusted systems and shared tooling for discovery, remediation, and responsible disclosure. It also aims to give maintainers one predictable partner instead of many uncoordinated teams. The main message of the letter is clear: since the technology world depends on open source together, companies must also defend it together.

A new debate is growing around online age verification, especially for social media. Supporters say age checks can protect children from harmful content and keep young users off platforms that may affect their mental health. Critics, however, warn that age verification often becomes identity verification. In practice, this can mean users are asked to share a face scan, a passport, or other official documents before they can post, read, or join online discussions.

An article from FIRE argues that this trend could seriously damage privacy. It describes a situation where a platform wrongly decides that an adult user is under 16. The user may then be pushed to a third-party verification app and told to upload biometric information or a government-issued ID. This raises several concerns: users may not know where the app is based, how long their data is stored, or how well it is protected from hackers and data breaches.

The article points to Australia as an example. Its social media ban for users under 16 took effect in late 2025 and has become an important model for other countries. But according to the source, early results suggest that the law is not fully achieving its goal. Government research reportedly found that many children were still using social media after the ban began, and a medical journal study found little immediate reduction in reported use by adolescents under 16.

Because platforms can face large fines, they are under pressure to take sufficient steps to keep underage users out. Sometimes companies may use existing account data to estimate a person’s age, but in many cases they may need to collect more personal data. That creates a difficult trade-off between child safety and user privacy. For technology teams, this debate is important because system design, security controls, data storage, and vendor risk management all become more complex when identity checks are required.

Dolphin Emulator, a well-known open-source project for GameCube and Wii games, has published its Progress Release 2606. These regular reports explain what developers have changed in the emulator and why the updates matter. An emulator is software that lets one system behave like another, so users can run older console games on modern devices. For many users, Dolphin is not only a gaming tool but also an example of long-term software engineering.

The new release highlights steady work on compatibility, performance, and usability. Compatibility means how well the software can run different games without serious problems. Performance refers to speed and smooth operation, especially on different hardware. Usability covers the user experience, such as settings, menus, and clearer behavior. Progress releases are important because they show that improvement often comes from many small fixes rather than one dramatic change.

Like many mature technical projects, Dolphin depends on careful maintenance of its codebase. A codebase is the full collection of source code in a software project. Over time, developers need to refactor parts of that code, which means rewriting them to make the system cleaner and easier to support without changing the main purpose. This kind of work may be less visible to users, but it reduces technical debt and can prevent future bugs.

For engineers, the Dolphin report is a useful reminder that software quality is built step by step. Emulator development is especially complex because it must reproduce the behavior of specialized hardware in software, often with very high accuracy. That challenge requires testing, debugging, and a strong community process. Even outside gaming, the release shows how transparent progress reports can help users understand technical decisions and build trust in a project.

A new blog post argues that many companies add too many tools when PostgreSQL could solve the same problems. The writer says engineers often choose packaged solutions before thinking from first principles. As a result, teams may build systems with more moving parts, higher operational risk, and greater complexity than they really need.

The article focuses on a common question in software design: when is one relational database enough? Using a financial data example, the writer explains how standard PostgreSQL features can support auditing, state reconstruction, and strong business rules. In simple terms, auditing means keeping a reliable record of what happened, while state reconstruction means being able to understand how data changed over time.

The post also discusses performance concerns that often push teams toward extra infrastructure. These concerns include write throughput, slow transactional queries for daily application use, and slow analytical queries for reporting. The writer says PostgreSQL can address these needs with careful schema design, constraints, triggers, indexes, and planning for the working set, which means the data that is used most often.

Another key idea is that business logic should be protected in the database, not only in the application layer. The article describes rules for transfers and transactions, such as keeping transaction records immutable and using a state machine to control allowed changes. It also mentions serializable isolation, a high level of transaction safety, and decoupling, which means reducing tight dependence between parts of a system. Overall, the post suggests that many teams should treat PostgreSQL as the default choice before adding specialized products.

In Silicon Valley, a “unicorn” is a private startup valued at more than $1 billion. A “zombie unicorn” is a company that is still alive on paper but is no longer growing strongly or moving toward a public listing. Many of these firms raised money easily during years of low interest rates. Now, in a tougher market, some are finding it hard to justify their old valuation.

This matters beyond finance because many unicorns work in important technology areas, including hardware, chips, robotics, and data infrastructure. These businesses often need a lot of capital to build products, hire engineers, and manage a long development cycle. If funding slows down, they may delay research, reduce hiring, or cut ambitious projects. That can affect the speed of innovation across the wider tech industry.

Investors are now paying closer attention to revenue, profit, and cash flow instead of fast expansion alone. In the past, venture capital firms often accepted big losses if they believed a startup could dominate a future market. Today, they are asking whether a company has a sustainable business model. Some startups may need to raise new money at a lower price, while others may look for a merger, a sale, or major cost cuts.

For founders, employees, and customers, zombie unicorns create uncertainty. Workers may see stock options lose value, and enterprise buyers may worry about long-term support for products. At the same time, a weaker market can have a healthy effect by forcing companies to focus on real demand and operational discipline. The bigger question for Silicon Valley is whether fewer easy bets will lead to slower progress, or to stronger and more resilient technology companies.

Experts at France’s National Library have identified a handwritten notebook as a work by Wolfgang Amadeus Mozart when he was 22 years old. The 44-page manuscript was kept between May and July 1778 while Mozart was living in Paris. Library specialists called it a major discovery because the notebook had been stored for a long time without being clearly linked to the famous composer.

According to the library, Mozart used the notebook while working as a music tutor for Marie-Louise-Philippine de Guines. Her father, the Duke of Guines, was a respected flute player and had commissioned Mozart’s Concerto for Flute and Harp. The notebook includes daily exercises for the young harp student, as well as seven pieces for flute and harp. These pieces may have been written for the father and daughter to perform together.

The discovery was made by curator Francois-Pierre Goy in the library’s music department. He was sorting documents before retirement when he noticed handwriting that looked similar to other teaching materials by Mozart. He studied small details, such as the shape of the treble clefs and bass clefs, and compared the notebook with other handwritten Mozart works. He also found that the paper used the same stamps as a copy of the Flute and Harp concerto linked to the Duke of Guines.

The evidence became stronger after these comparisons, and the notebook was authenticated in April 2026 by the Mozarteum Foundation in Austria. The library said the manuscript had originally been part of music bundles confiscated from the Duke of Guines’s home during the French Revolution. Beyond its musical value, the case shows how careful examination, pattern matching, and expert review can reveal important information hidden inside large archives. In this way, an old paper notebook became a newly recognized historical record.