H.Headline
Chaos

THE EDITION ARCHIVE

A front page, saved in time.

Making AI Headlines · AM edition

Published · New York time

Permanent edition link →

10 pages · 23 stories

PAGE 1 OF 10

Security Cameras Demand Freedom of Movement

Ring's planned drone and rover expand the patrol, while Anthropic brings Claude to office documents and meetings.

Making AI Headlines newspaper, page 1: Security Cameras Demand Freedom of MovementPage 1 of 10
Page 1 of 10. Read the same stories below.

The Camera Has Decided to Walk Around

Amazon has announced two roaming Ring security devices: a rover expected to reach customers in late 2026 and a drone expected in early 2027. Both use an AI Pilot system combining cameras, sensors, vision models and cloud processing to plan patrols. The rover is designed for businesses, while the drone is intended for homes. Mobility is the selling point and the complication. A stationary camera has a fairly obvious field of view. A device that travels can collect a broader picture of a space, making questions about privacy and who can see footage harder to leave in one corner of the room. Amazon has faced privacy concerns over an earlier effort to popularize a flying security camera. For customers, the appeal is easy to understand: one moving device might check places that would otherwise require several fixed cameras. The tradeoff is that a more complete tour can also capture more of ordinary life. A robot sent to look for trouble may become very familiar with the laundry. The larger pitch is that AI can decide how to move while the owner gets on with the day. It is a convenient arrangement, provided everyone in the space is comfortable with a security system whose job description now includes wandering.

Share articleOpen article permalink →

Anthropic Promises to Put the Spreadsheet in the Meeting

Anthropic has announced Claude Sheets, a tool for generating and editing spreadsheets, alongside products for meetings and whiteboards. The expansion puts Claude more directly into office work long associated with Google and Microsoft. Sheets supports tabs, formulas and formatting, while Claude Meetings can join Google Meet discussions. The attraction is not simply a machine that fills cells. It is the prospect of moving from a discussion to a plan and then to a spreadsheet without making a person translate every step. That could save effort when the information is sound and the task is well defined. It also moves the point of review: instead of building a table from scratch, a worker may need to inspect one that already looks finished. Formulas and polished formatting can make an answer look authoritative even when the underlying assumptions deserve a second look. A whiteboard can organize a conversation, but it cannot make a weak plan strong merely by arranging it neatly. The office still needs people who know what the numbers mean. For software makers, the prize is a place in the whole workflow rather than a single app. For employees, the suite offers a familiar bargain: fewer blank pages, more material to check, and a meeting companion ready to produce a spreadsheet before anyone has agreed what belongs in it.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑

PAGE 2 OF 10

AI Assistants Arrive in Groups, Apparently to Avoid Working Alone

Google's Gemini 4 Carbon reaches the public as Adobe adds agents and prototyping tools to its creative products.

Making AI Headlines newspaper, page 2: AI Assistants Arrive in Groups, Apparently to Avoid Working AlonePage 1 of 10
Page 2 of 10. Read the same stories below.

Google Invites Everyone to Manage a Swarm

Google is rolling out Gemini 4 Carbon to the public after initially releasing it to a select group of cybersecurity defenders. The company calls it its strongest model yet and says it can compete with leading systems from Anthropic and OpenAI. Its features include an agent swarm, bringing the language of coordinated work to the AI assistant market. The public launch widens the audience for a model Google had first placed with a more specialized group. It also puts another ambitious claim into a field where companies compete not just over what a model can answer, but over how much work it can coordinate. A swarm sounds industrious; whether it is useful depends on the task, the instructions and the quality of the result. For users, the attraction is having software divide up work rather than requiring a person to prompt every step. The responsibility remains deciding what needs doing and checking the output. More agents can create more activity without necessarily making the original question clearer. Google's assessment of Carbon's strength is not a guarantee that it will be the best choice for every job. Still, the branding captures the moment: a single assistant is no longer enough. The assistant now arrives with a crowd, and the human gets the management position.

Share articleOpen article permalink →

Adobe Makes the Creative Brief Briefly Optional

Adobe has announced AI additions across Photoshop, Premiere, Illustrator and InDesign, as well as new capabilities in Firefly. At its MAX event, the company also introduced a way for outside chatbots to use Adobe features and a Firefly prototyping tool called Builder in private beta. The announcement extends AI assistance across several kinds of creative work, from editing images to presenting design ideas. Builder moves Firefly toward territory occupied by Figma, while another tool is intended to create on-brand graphs. Adobe is also adding image editing that can change one part of an image without altering other objects. The appeal is speed, especially for repetitive changes and early drafts. The tradeoff is that a polished prototype or graph can arrive before the decisions behind it have been fully settled. A brand may get a chart in its preferred colors while its author still has to decide whether the chart says anything useful. Letting chatbots use Adobe features also puts familiar creative tools behind a conversational interface. That may reduce the number of menus a user must visit, but it does not eliminate the need to judge the result. The creative brief can become shorter; the responsibility for knowing what good work looks like remains stubbornly full length.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑

PAGE 3 OF 10

Science Hires AI, Then Discovers Checking Is Still a Job

PSI applies an unreleased model to a quantum computing question, while researchers examine coding agents' distractions.

Making AI Headlines newspaper, page 3: Science Hires AI, Then Discovers Checking Is Still a JobPage 1 of 10
Page 3 of 10. Read the same stories below.

Quantum Computing Gets an Assistant With a Very Large Homework Assignment

PSI used an unreleased Anthropic model to answer a fundamental quantum computing question that scientists have pursued for nearly two decades. The effort is part of the company's attempt to build a business around AI-assisted scientific discovery. The distinction between producing an answer and establishing a discovery matters here. Scientific work depends on methods and results that other researchers can examine, not just on a system producing an impressive conclusion. A model may help explore a difficult problem, but the usefulness of its contribution rests on what people can verify and build upon. For a company trying to turn this kind of work into a business, the attraction is obvious: problems that occupy specialists for years are valuable precisely because they are hard. The challenge is showing that an AI system contributes more than an intriguing passage of text. Researchers need a path from a proposed result to an assessment of its substance. The commercial pitch also places an old scientific virtue in a new wrapper. Nobody gets to skip the checking simply because the assistant is unusually fast. If AI can help with hard questions, the next task is to make its work intelligible enough for experts to disagree with it productively. Even an automated breakthrough has to sit through scrutiny.

Share articleOpen article permalink →

Coding Agents Discover the Ancient Art of Being Distracted

Google researchers have described a problem they call agent distraction: AI coding agents can be diverted by stale technical debt, including old TODO comments, obsolete bug references and outdated specifications. Their assessment highlights an obstacle to using agents in real software projects, where code is often accompanied by years of accumulated notes. A comment left for a future developer may no longer describe the actual state of a program. To an agent searching a repository for clues, however, that comment can look like a live instruction. The resulting detour is not necessarily a failure of typing speed. It is a problem of deciding which information still matters. For teams adopting coding agents, that makes housekeeping part of the tool's performance. Clearer documentation and removal of outdated references can help both human developers and automated assistants understand the work. Keeping records useful takes time, though, and software teams already have plenty of reasons to postpone cleanup. The episode gives technical debt a new audience. A forgotten note that once distracted one programmer can now invite an automated workflow to investigate yesterday's problem. The future of coding may be fast, but it still has to read the comments left by the past.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑

PAGE 4 OF 10

AI Takes On Two Jobs Humans Left Messy

Veterans use a free ChatGPT offer, while vision systems help sort reusable materials.

Making AI Headlines newspaper, page 4: AI Takes On Two Jobs Humans Left MessyPage 1 of 10
Page 4 of 10. Read the same stories below.

The Civilian Job Search Gets an Extremely Eager Assistant

More than 20,000 veterans and transitioning service members have claimed a free year of ChatGPT Plus, according to OpenAI. The company says participants use the service to prepare for civilian careers, putting a general-purpose AI tool into a transition that can involve unfamiliar job titles, applications and workplace language. The offer gives users access to a paid product without the usual subscription cost for a year. Its potential value depends on what people do with it: drafting and revising career materials can be useful, but a polished application still needs to reflect the applicant's actual experience. A chatbot can suggest language; the person applying has to decide whether that language is accurate and appropriate. There is also a business angle. Free access introduces a group of users to a product that normally costs money, while giving those users an opportunity to see whether it fits their needs. The one-year period makes the offer concrete, but does not make the tool a substitute for human career guidance or an employer's judgment. The most useful assistant may be one that helps translate real skills rather than inventing impressive ones. In a job search, sounding confident is good. Sounding like a committee wrote your entire life story is less compelling.

Share articleOpen article permalink →

Recycling Robots Take a Hard Look at Our Decisions

Vision AI and robotics are being used to identify and pick recoverable items from streams of used clothing, bottles and mixed recyclables. Nvidia says systems featured in its reporting help divert thousands of tons of material from landfills. Sorting is a basic but consequential step. Materials that arrive mixed together have to be separated before they can be reused effectively, and a conveyor belt does not pause politely while each item is considered. Computer vision offers a way to identify objects in that stream, while robotic equipment can pick out material worth recovering. The approach addresses a physical problem, not just an information problem. Recognizing a bottle is useful only if the rest of the process can handle it. The material still has to move through a system that can sort and reuse it, and the value of automation depends on how well it fits the operation. The spectacle is also a corrective to the idea that every AI application must end in a chatbot. Some systems are working through the unglamorous consequences of what people throw away. After years of asking technology to make things look new, the industry has found a job in deciding which old things deserve another chance.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑

PAGE 5 OF 10

Every Screen Wants an Assistant, Including the Tiny One

Apple's Visual Intelligence is coming to the iPad camera, and Meta's Muse expands to smartwatches.

Making AI Headlines newspaper, page 5: Every Screen Wants an Assistant, Including the Tiny OnePage 1 of 10
Page 5 of 10. Read the same stories below.

The iPad Camera Applies for a Thinking Role

Apple's Visual Intelligence is set to be accessible through a new tab in the iPad Camera app. The addition brings an AI feature into a familiar place on the tablet: the interface people already use to point a camera at the world. The change matters because the camera becomes more than a way to record an image. An AI feature in the same app invites users to treat what they see as something they can ask software to interpret. That can make a tablet's camera feel more like an entry point to a digital assistant and less like a standalone tool. The interface is a quiet part of the pitch. A separate app requires a person to remember that a feature exists and decide to open it. A camera tab puts the option near an activity users already understand. The closer the assistant is to the viewfinder, the easier it is to try, and the more important it becomes for users to consider what they are asking it to examine. The familiar camera still has a simple job: show what is in front of it. The new tab adds a more ambitious colleague. It is the difference between holding up a tablet to see something and holding up a tablet to begin a conversation about what it sees.

Share articleOpen article permalink →

Meta Gives the Wrist Another Reason to Interrupt

Meta's AI agent Muse is available on Apple Watch and Android Wear OS. On Apple Watch, users can interact with the assistant through voice calls, text messages or dictation, bringing the service to a device designed for brief, frequent interactions. Moving an assistant from a phone to a watch changes the setting in which people might use it. A question that once required opening an app can become a quick exchange from the wrist. That convenience is the point of wearable computing, but it also makes an assistant easier to consult in the middle of other activities. The available forms of interaction matter. A spoken request, a typed message and dictation each suit different circumstances, and a watch's small display encourages shorter exchanges. Users may appreciate a quick answer without wanting a long conversation on a tiny screen. For Meta, another device gives Muse another opportunity to become part of a daily routine. For the person wearing it, the test is whether that access saves attention or asks for more of it. The wrist already tells time and delivers alerts. Now it can host an AI assistant too, making the smallest screen in the house a candidate for the busiest meeting.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑

PAGE 6 OF 10

Shopping Bots Learn That Not Every Store Wants a Chaperone

eBay appears to restrict outside agents as OpenAI adds tools for business discovery.

Making AI Headlines newspaper, page 6: Shopping Bots Learn That Not Every Store Wants a ChaperonePage 1 of 10
Page 6 of 10. Read the same stories below.

eBay Reminds Shopping Bots That Browsing Is Not a Birthright

eBay appears to have blocked some third-party AI agents from browsing its platform, including Meta's Muse. The apparent restriction presents an obstacle for assistants designed to search listings and assemble shopping information for users. The issue illustrates a tension in the emerging agent economy. An assistant may be useful because it can visit several services on a person's behalf. Those services, meanwhile, have their own reasons to control how automated visitors access listings and present information. A user's request does not automatically settle the relationship between the agent and the platform. For shoppers, the appeal of a curated digest is straightforward: less time searching and more time comparing. But an agent's usefulness depends on what it can actually reach. If a major marketplace is unavailable to it, a tidy list may represent only part of the available market. The neatness of the answer cannot make up for a missing aisle. For platforms, managing outside automation can also be a question of how their listings are accessed and experienced. The broader contest is over who controls the path between a shopper and a product. AI companies are selling a future in which assistants do the browsing. Marketplaces are demonstrating that the front door still belongs to somebody.

Share articleOpen article permalink →

ChatGPT Opens a Business Directory and Calls It Discovery

OpenAI is introducing business profiles and other features intended to make ChatGPT a stronger tool for consumer discovery. The changes give businesses ways to participate as people use the chatbot to find products and services, moving the assistant closer to functions associated with search engines. The shift matters because a request for an answer can also be a request to find a business. For a consumer, a conversational interface might make it easier to describe what they need without sorting through a long list of links. For a business, being represented in that interface could become another way to reach potential customers. That convenience comes with an important distinction: a profile is information about a business, not an independent guarantee that it is the best fit for a particular customer. People still have to compare what they find with their own needs, and businesses have an incentive to present themselves attractively. The new features also show how assistants are becoming places where commercial encounters begin, rather than merely tools for writing or answering general questions. The old online directory asked businesses to describe themselves. The new one lets the customer start by asking a chatbot. Either way, everyone would still like to appear near the top of the conversation.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑

PAGE 7 OF 10

AI Builds Doors Faster Than Anyone Can Check the Locks

A researcher finds exposed Grok Build apps, while an open-source executive warns about cyberattacks.

Making AI Headlines newspaper, page 7: AI Builds Doors Faster Than Anyone Can Check the LocksPage 1 of 10
Page 7 of 10. Read the same stories below.

The App Builder Makes Sharing Look Alarmingly Easy

A researcher found that 35,000 Grok Build apps were open for others to access, with confidential information exposed in some cases. The finding highlights a security concern around tools that make it easier to create applications quickly. The appeal of an AI coding agent is that it can shorten the distance between an idea and a working app. But an application is not finished simply because it runs. Decisions about what is visible, who can access it and what information it contains matter just as much as the interface a user sees. An exposed app can turn an intended convenience into an unintended audience. That is particularly significant when confidential information is involved. The security question is not whether rapid development is useful; it is whether the safeguards around access keep pace with the speed of creation. For users and organizations, building and publishing are different responsibilities, even when software makes them feel like one continuous action. An assistant can help put a door on a digital building, but someone still needs to check whether it closes. The modern promise is an app in less time. The less glamorous requirement is knowing who else can walk through it.

Share articleOpen article permalink →

Open AI Platforms Face the Awkward Task of Closing Some Doors

Hugging Face executive Irene Solaiman warns that growing cyberattacks on open AI platforms could lead to tighter security restrictions. Her concern is that measures intended to protect those platforms could also reduce the accessibility that makes them useful to developers around the world. Open access makes it easier for people to share and build on tools. That same openness can create difficult security decisions when a platform faces hostile activity. Restrictions may reduce some risks, but they can also add friction for legitimate users and smaller projects that benefit from easy access. The tension is not unique to AI, but it matters in a field where models, code and research move quickly across organizational boundaries. A developer community depends both on the ability to exchange material and on confidence that the systems hosting it can be used safely. A gate that protects the exchange can also slow it down. Solaiman's warning frames security as more than a technical expense. It can shape who gets to participate and how easily. The ideal open platform welcomes collaboration without making every visitor a security exercise. The less elegant reality is that openness sometimes requires a very detailed conversation about locks.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑

PAGE 8 OF 10

Entertainment Discovers the Audience Can Turn AI Down

Spotify adds controls for its AI DJ as game developers and festival jurors debate generative tools.

Making AI Headlines newspaper, page 8: Entertainment Discovers the Audience Can Turn AI DownPage 1 of 10
Page 8 of 10. Read the same stories below.

Spotify's AI DJ Learns That Silence Has Fans

Spotify Premium users can customize the AI DJ with new voices and personalities, and can turn off commentary between songs. The update gives listeners more control over how much of the automated host they hear while using the music feature. The choice acknowledges a familiar tension in music services. A voice can make recommendations feel more personal, but a listener who wants a run of songs may not want an explanation between every track. Different personalities offer variety; the option for less chatter addresses the possibility that the preferred personality is occasionally quiet. For Spotify, an AI DJ can make discovery feel like a guided experience rather than a list of tracks. For subscribers, the guide has to fit the moment. A workout, a party and a quiet commute may call for very different amounts of narration. One person's helpful host is another person's interruption before the chorus. The new controls do not resolve every question about how recommendations work, but they give users a say over one conspicuous part of the experience. That is a telling feature in an industry eager to make AI more conversational. Sometimes the smartest thing a digital host can do is introduce the next song by not introducing it.

Share articleOpen article permalink →

Indie Games Argue Over Whether the Credits Need an Asterisk

Some Independent Games Festival jurors have said they will not vote for games that use generative AI in any capacity. The developer of Big Walk separately said it used AI sparingly near the end of the game's development. Together, the statements expose a disagreement over how AI belongs in independent game production. For jurors, a broad rule offers a clear boundary: use of the technology affects whether a game receives their vote. For developers, the issue can turn on how a tool was used and how much of the finished work depended on it. Those positions do not necessarily produce the same assessment of a game. The dispute matters because independent games are often associated with small teams and distinctive creative choices. Tools that save time can be attractive, while decisions about authorship and artistic labor remain central to how a work is received. A small amount of AI use can therefore become a large point of disagreement. The festival question is not merely whether a game is enjoyable. It is also what standards jurors apply when deciding what to celebrate. Independent creators have long been encouraged to use tools that help them make something new. The argument now is over which tools change what that encouragement means.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑

PAGE 9 OF 10

New York Asks AI Companies to Show Their Work Before the Applause

The state prepares to enforce its safety law, OpenAI buys a travel startup, and Google reports work on software proofs.

Making AI Headlines newspaper, page 9: New York Asks AI Companies to Show Their Work Before the ApplausePage 1 of 10
Page 9 of 10. Read the same stories below.

New York Adds Paperwork to the Frontier

New York is preparing to enforce an AI safety law requiring major developers such as OpenAI and Anthropic to register, disclose safety practices and report critical incidents or face fines. The law puts formal oversight around companies building advanced AI systems. The requirements make safety a matter for regulators as well as company statements. Registration identifies covered developers, while disclosures and incident reports give the state ways to see how those companies handle risks. The practical value will depend on enforcement and what the reports reveal. For developers accustomed to describing their systems as groundbreaking, the new routine is less glamorous: explain the safeguards and report serious problems. Frontier technology has reached the stage at which someone in Albany wants the forms filled out.

Share articleOpen article permalink →

OpenAI Buys a Travel Startup Before Announcing the Itinerary

OpenAI has acquired Odessia, a travel startup created by Sonder founder Francis Davidson. The deal brings a company with travel-industry roots into the AI developer's orbit, though OpenAI has not said what its future travel plans are. Travel is a natural setting for assistants that help people organize information and compare choices. It is also a complicated business in which a useful suggestion is only one part of a journey. Customers care about whether arrangements suit their needs, not merely whether a proposed itinerary reads smoothly. Davidson is a known figure in the travel industry, giving the acquisition an identifiable connection to that market. For now, the deal itself is the concrete development. The AI industry has added a travel startup to its packing list, while the destination remains for the company to reveal.

Share articleOpen article permalink →

Software Verification Gets Three Million Lines of Patience

Google DeepMind says AlphaProof Code agents wrote three million lines of proofs in the Lean programming language to verify Signal's core encryption. The work applies AI agents to formal verification, a process intended to establish properties of software through mathematical proof. Encryption is a consequential place to demand rigor. A system can appear to work in ordinary use while the properties that protect communications require deeper examination. Formal proofs offer a way to make those claims precise and checkable. The length of the work also challenges the idea that AI's value is always a shorter answer. Here the output is extensive because verification demands detailed steps. The assistant's accomplishment, as reported by its creator, is not sparing anyone the mathematics; it is helping produce a great deal more of it.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑

PAGE 10 OF 10

AI's Answer to Everything Still Leaves a Few Forms to Fill

A mathematician considers AI as an oracle, New York plans tests, and founders prepare to sell shares.

Making AI Headlines newspaper, page 10: AI's Answer to Everything Still Leaves a Few Forms to FillPage 1 of 10
Page 10 of 10. Read the same stories below.

Mathematics Considers an Oracle That Still Owes a Proof

Mathematician Steven Strogatz discusses AI and mathematics in connection with his book, Big Math. The question is what happens when a system can offer answers to problems that people may struggle to assess. Mathematics prizes more than a convincing conclusion: the reasoning needs to withstand scrutiny. A powerful tool could help researchers explore ideas, but an answer that cannot be checked leaves an important part of the work unfinished. The oracle may be quick; the proof still has office hours. Even a confident answer must eventually explain itself to the rest of the department.

Share articleOpen article permalink →

Anthropic Founders Plan Matching Exits From a Small Slice

Anthropic's seven founders each plan to sell the same number of shares in the company's IPO. At the midpoint of the indicated range, the sales would bring each founder about $250 million, while the group would retain control of the company. The arrangement combines a substantial potential cash-out with continued influence. Equal planned sales give the transaction a symmetry not often associated with the messy work of building a technology company. The founders could each leave the offering with cash while keeping a collective hand on the steering wheel. Equal portions, it seems, also make the cap table easier to discuss at dinner.

Share articleOpen article permalink →

Oura Makes a House Call on Primary Care

Oura plans to acquire AI-native primary care startup Counsel Health for $625 million. The deal follows Oura's indefinite postponement of an IPO that could have raised up to $2.2 billion, with the company citing uncertainty in the IPO market. The acquisition connects a wearable technology business with primary care. It also shows that postponing a public offering does not mean a company has stopped pursuing deals. Oura's ambitions are moving beyond a device worn on a finger and toward a business providing care. The ring is getting closer to the doctor's office, although wearing one is not the same as receiving care.

Share articleOpen article permalink →

New York Decides AI's Self-Assessment Needs a Second Opinion

New York's research center will independently test AI models rather than rely solely on companies' own evaluations. The plan adds an outside check to a field where developers frequently promote the capabilities and safety of their own systems. Independent testing can give officials another perspective on how models behave. It also creates a useful separation between building a product and assessing it. Companies may know their systems well, but they also have an obvious interest in presenting them favorably. Even an advanced model can benefit from someone else grading the paper, particularly when its creator is eager to put the results on the refrigerator.

Share articleOpen article permalink →

Sources from the news desk

Satire. Not factual reporting. Read the original reporting from the linked publishers.

Back to page 1 ↑