Google's AI Still Can't Spell, Including The Word "Google"

And though the Gemini-powered technology has improved its accuracy dramatically over the last two years (unfortunately for publishers), AI Overviews still gets basic questions wrong. And that includes spelling tests.

Google's AI tools remain abysmal at answering questions about spelling, having gone viral two years ago for responding to the question "how many r's are in the word strawberry?" incorrectly. But it's still bad. Last 26 May, X user Naomi Rohatyn tested the large language model's (LLM) current ability to answer to a spelling question.

"There are exactly 2 'e's in the word "astronomical" (a-s-t-r-e-n-o-m-i-c-a-e-l)," replied AI Overview.

How many Ps are in Google? According to Google, there are two.

There’s also is also "exactly 1 'r' in the word 'poop'," Google’s AI Overview says, as well as two ‘d’s in the word journalism, yet spelled it: j-o-u-r-n-a-d-i-s-m. Google did at least identify that there is one P in the last name of the U.S. president, but spelled it as t-r-p-u-m.

Considering users are less likely to click on links when an AI summary appears in the results, surely the information provided in AI Overviews should be accurate. But it's complicated.

AI chatbots need exact context and specifics to answer as well as they can, so surely spelling words within their training data seems easy. However, things get knotty when you ask an LLM to consider words letter-by-letter, as the model will process text in chunks rather than individual characters (it's called tokenisation).

"Counting within words has been a known challenge for LLMs, and we’re working to fix this particular issue," Google told TechCrunch in an emailed statement.

These basic spelling errors may seem familiar. LLMs, the kind of artificial intelligence that powers chatbots and other text-generators, are not built to understand spelling. It’s been a running joke for years that whenever a company unveils a new AI model, you should ask it how many 'r's are in the word strawberry. These AI models — which can code an app in seconds, or solve problems that have stumped mathematicians for decades — are about as good as a kindergartener at spelling.

Google’s AI overview woes reach beyond silly spelling mistakes though. Google already patched an issue from last week in which searching the word "disregard" would yield what looked like a dictionary definition of the word, only the definition was shown as, "Understood. Let me know whenever you have a new prompt or question!" But these spelling errors have remained amusing because they’re so difficult to quash.

"LLMs are based on this transformer architecture, which notably is not actually reading text. What happens when you input a prompt is that it's translated into an encoding," Matthew Guzdial, an AI researcher and assistant professor at the University of Alberta, told TechCrunch. "When it sees the word ‘the,’ it has this one encoding of what ‘the’ means, but it does not know about 'T,' 'H,' 'E.'"

The token-based architecture that powers LLMs like Google’s AI overview is inherently limiting, and researchers haven’t been optimistic that they can solve the spelling problem.

Read More

Windows 11 Is Upgrading To A More "Useful Start Option"

On Windows 11, as part of Microsoft's broader "Windows K2" initiative to improve the operating system and rebuild user trust, the company has revealed the first tranche of Start menu changes that will soon begin rolling out through the Windows Insider Program.

The upcoming improvements focus heavily on customization and usability. Some of the most notable changes include the ability to resize the Start menu, show or hide individual sections, and separate file recommendations from recent activities.

When the software giant introduced Windows 11, the company replaced the classic Windows 10 Start menu with a completely redesigned experience. The new design removed Live Tiles and the secondary pane layout in favor of a simplified, centered menu that only included traditional app icons.

While the redesign modernized the interface, it also reduced functionality and customization options compared to Windows 10. Since launch, users have consistently criticized the Start menu for feeling restrictive and less flexible.

Now, Microsoft appears ready to address many of those complaints and one of the biggest additions is support for manually resizing the Start menu.

Until now, the Start menu automatically adjusted its dimensions based on screen size and resolution, offering virtually no control over its appearance. In future updates, users will be able to choose between smaller and larger layouts.

Unlike Windows 10, Windows 11 still won't allow freeform resizing by dragging the edges of the menu or enable a full-screen Start experience. However, the new option is still a major improvement over the current fixed layout.

Microsoft is expected to add the feature through a new "Start menu size" setting available from Settings > Personalization > Start.

Microsoft is also introducing privacy-focused improvements for users who frequently share their screens during meetings, presentations, or live streams.

The company plans to add a new option called "Hide your name and profile picture on Start." When enabled, the Start menu will no longer display the account name and profile image.

The setting will appear under Settings > Personalization > Start > Other.

Originally, the Windows 11 Start menu separated content into multiple areas, including "Pinned," "Recommended," and the separate "All apps" list.

One of the most common complaints after launch was the inability to remove the "Recommended" section entirely. Microsoft eventually added a workaround, but it required turning off several unrelated settings before the section disappeared.

More recently, the company redesigned the Start menu by merging the "All apps" list into the main interface. However, customization still remained limited.

As part of the Windows K2 effort, Microsoft has now confirmed that users will soon get dedicated toggles to individually show or hide the "Pinned," "Recommended," and "All" sections.

Read More

A "Living Bandage" Can Speed Up Healing

There is now an engineered "living bandages" that could offer a new way to treat chronic wounds by delivering healing signals directly where the body needs them most. Researchers at Rice University developed a cytokine factory patch that continuously releases therapeutic proteins inside wounds.

Chronic wounds remain difficult to treat because the body often struggles to maintain the immune signals needed for tissue repair. Existing therapies also face limitations because healing proteins degrade quickly and fail to remain at the wound site for long periods.

The Rice team designed the patch as a cell-based delivery platform that uses engineered cells as miniature factories. These cells continuously produce cytokines over extended periods instead of releasing them in short bursts.

Cytokines are signaling proteins that regulate inflammation, immune activity, and tissue regeneration. By delivering them directly at the wound site, the patch aims to maintain a stable healing environment during recovery.

The device came from the laboratory of Omid Veiseh. Researchers encapsulated ARPE-19 cells engineered to secrete cytokines including IL-10, IL-12 and TGF-β inside a protective biocompatible matrix.

The matrix allows nutrients and therapeutic proteins to move through the material while shielding the engineered cells from the host immune system. Researchers said this setup helps sustain localized cytokine delivery for longer periods.

In preclinical studies, the patch accelerated healing in excisional wound models in mice and pigs. The findings highlighted the potential of sustained, localized immune modulation during tissue repair.

"The findings show how continuous, localized cytokine delivery can support key biological pathways involved in tissue repair," Veiseh said. He added that maintaining a steady supply of signaling molecules at the wound site helped engage the body’s natural healing response more effectively.

Researchers also examined how the patch influenced biological activity at the cellular level. RNA sequencing revealed activation of several pathways associated with wound healing after treatment.

Transcriptomic analysis showed increased activity in genes linked to tissue regeneration and immune modulation. The results provided a molecular explanation for the faster healing observed in the animal models.

The platform also offers flexibility beyond a single treatment strategy. Researchers designed the system so scientists can modify the engineered cells to produce different combinations of cytokines, growth factors, or therapeutic proteins.

That modular structure could help tailor future treatments for different wound types and diseases. The team also integrated an optimized hydrogel matrix that may eventually work alongside bioelectronic technologies.

Researchers believe the cytokine factory concept could extend beyond wound healing. They said the platform may support localized delivery of therapeutic proteins in diseases that require sustained, site-specific immune signaling. Christian Schreib, assistant research professor in Rice’s bioengineering department and co-author of the study, said the team now plans to improve control over cytokine delivery.

Read More

Power Banks Kept On Catching Fire Inside Planes

For those traveling by plane these days, they know that there are some pretty important rules when it comes to packing your luggage. But alongside the usual rules for weapons, liquids, and gels, portable power banks — and other electronic devices containing lithium-ion batteries — are facing heightened scrutiny due to an increased risk of fire and burns due to faulty batteries or misuse.

The FAA reported 93 incidents in 2025, and 2026 is already seeing a spike, with 22 reported by mid-April. In response, Southwest Airlines now requires passengers to keep portable chargers visible during use, while Japan has banned power banks on flights altogether.

So, what's the problem with devices containing lithium-ion batteries?

Faulty lithium-ion batteries can overheat or short-circuit, which can kickstart a dangerous chain reaction called thermal runaway, a process where rising temperatures inside the battery continually release energy, as it perpetually grows hotter and hotter. That increases the risk of fire, but the materials are also highly reactive and flammable.

Lithium fires are aggressive and self-sufficient. Efforts to douse them are likely to fail and if successful, the fire can reignite. As reported by CBS News, an Alaskan Airlines flight had to land after a power bank onboard caught fire. While not aboard a plane, in another incident, a 75-year-old woman died of injuries after a charger exploded in her lap.

The easiest ways to limit your exposure to a lithium battery fire are to avoid cheap, poorly made devices and follow manufacturer recommendations when it comes to properly using the device. Overheating and other complications can happen when the batteries are overused, such as too many devices being plugged in at once. Problems can also arise when lithium devices are treated poorly, damaged, or stored improperly.

Always store batteries away from anything that is flammable or could potentially cause the device or battery harm. Store them at temperatures above 32 degrees Fahrenheit (0 degrees Celsius) and below 105 degrees Fahrenheit (40 degrees Celsius). Try to avoid exposing them to extreme temperatures when you're traveling, too.

When boarding a plane, follow all safety instructions, which generally include information about devices containing lithium batteries — devices with these batteries should never be left in check bags. Following the TSA's battery restrictions for carry-on luggage is a great way to stay safe everywhere, not just at the airport or on a plane. Moreover, the best TSA approved power banks on Amazon, not only meet the allowance requirements, but also are generally well-received and considered safe by many users.

Read More

Accounts Of Microsoft Users Are Under Attack

The FBI issued a warning on May 21, as a new AI-powered attack enables "threat actors to obtain Microsoft 365 access tokens and bypass multi-factor authentication protocols without intercepting the user's credentials."

Dubbed Kali365, this phishing-as-a-service threat was first discovered last month. The FBI released its public service announcement "to warn the public" that these attacks use Microsoft’s authentication infrastructure to steal user credentials.

The new phishing-as-a-service platform is distributed via every hacker’s favorite messenger — Telegram. But the attack will come at you via email. "Kali365 lowers the barrier of entry," the bureau says, "providing less-technical attackers access to AI-generated phishing lures, automated campaign templates, real-time targeted individual/entity tracking dashboards, and OAuth token capture capabilities."

If you’re targeted, you’ll first see an email "impersonating (a) trusted cloud productivity (or) document-sharing services." This will include a device code "with instructions to visit a legitimate Microsoft verification page and enter the code."

You then "navigate to the real Microsoft page and paste in the device code," at this point you’re sharing your OAuth access code with the attacker, who can then use these on their own machine, gaining access to your Microsoft 365 account.

The FBI warns that once that’s done, "the attacker can now access Microsoft 365 services such as Outlook, Teams, and OneDrive without needing a password or completing any additional MFA challenges."

Most mitigation for this sit at the enterprise level. Blocking device authentication or creating conditional access policies "can help prevent or limit this style of attack." But for everyday users, understanding that credentials can be stolen in this way and then used to access your services on an attacker’s machine is critical.

Per Proofpoint, "device code phishing is exploding across the threat landscape, with new device code phishing tools emerging every week. The spike in device code phishing coincides with publicly released criminal toolkits, and the emergence of multiple phishing-as-a-service (PhaaS) offerings."

As such, apply the usual discipline. Do not follow links to sites for documents you’re not expecting. Be wary of any email that prompts an action or includes a link. Check the validity of the email before clicking on anything.

Read More