The feature lets users see Spotify’s brief summary of their interests in music, podcasts and audiobooks, then send a note asking for more or less of particular sounds, topics or moods. Spotify says the notes guide recommendations on the Home feed; they do not act as strict filters.

You can tell Spotify what has changed

To find it, open the Spotify mobile app, tap your profile picture and choose Taste Profile. Under “Tell us more,” a listener might ask for more ’90s alternative rock, a different mix for morning workouts, or more history podcasts and fewer technology shows.

Spotify says changes typically begin appearing on Home within a few hours. Notes can be edited or deleted, so the profile can change as a listener’s interests do. The company describes this as “gentle guidance,” not a replacement for its existing recommendation system.

That distinction matters. Spotify can already observe what people play, skip, save or replay, but those actions do not always explain why someone listened. A child’s music, sleep sounds or a temporary research binge can become misleading signals. A written instruction gives listeners a direct way to add context that may not be obvious from their listening history.

Taste Profile builds on narrower controls Spotify already offers, including excluding specific tracks or playlists from a profile and keeping children’s and family music from influencing a parent’s recommendations. The new feature makes that input broader: users can ask for interests they have not yet explored, as well as correct ones they have outgrown.

The algorithm still makes the final picks

A natural-language note is not a promise that every recommendation will match. Spotify still decides which songs, podcasts and audiobooks fit a request and how prominently to rank them. Taste Profile changes the information the system uses; it does not let listeners inspect or directly edit every part of the ranking model.

That makes the feature a steering wheel, not full manual control. It may help when the service has inferred a lasting preference from a short-lived habit, but Spotify has not said that a note will affect every playlist, search result or part of the app. Its announcement specifically describes changes to the Home feed.

Spotify and YouTube are making recommendations more conversational

The timing points to a wider design shift. On September 23, YouTube also announced Custom Feeds, which will let viewers describe an interest and save the resulting video recommendations in dedicated homepage tabs. FreshWire’s YouTube Custom Feeds report explains how that feature differs from Spotify’s approach: YouTube is building separate feeds, while Spotify is adjusting the recommendations on Home.

Both companies are asking people to explain what they want instead of relying only on past behavior. That could make personalization feel more understandable and easier to correct. It does not make the systems transparent, though. Users still cannot see exactly how each instruction changes the ranking or how long its influence lasts.

The useful test will be whether the recommendations improve after the novelty wears off. If the feature can recognize that a listener’s interests have changed—and make it easy to say so—it gives people a practical new control. If prompts only nudge a few rows on Home, its impact will be smaller than the idea suggests.

Sources

• Spotify: How to Use Our Taste Profile Feature to Shape Your Spotify Homepage
• TechCrunch: Spotify brings Taste Profile to U.S. Premium listeners
• YouTube: Custom Feeds and viewer features

The post Spotify’s Taste Profile Gives Premium Users More Control Over Recommendations appeared first on The FreshWire.

The film stars Sam Rockwell as private eye Raymond Gunn, Scarlett Johansson as megastar Venus Nova, and Tom Waits as Gunn’s one-eyed alien partner, Eye-Ra. Netflix says Ray Gunn will play in select theaters, including limited 70mm engagements, beginning December 4, then stream on Netflix from December 18.

A 1939 vision of the future

Ray Gunn is set in Metropia, a vast city imagined as if people in 1939 were predicting the future. Its streamlined Art Deco towers, flying cars and alien residents sit alongside murder, corruption and a detective who still works the old-fashioned way.

Bird has described the premise as “The Maltese Falcon meets Buck Rogers.” In Netflix’s synopsis, Raymond Gunn is drawn into a case involving aliens, murder and Venus Nova, a multimedia star whose fame brings trouble with it. The trailer’s central contrast is already clear: a retro future full of advanced machines, and a private eye who keeps doing the legwork himself.

That tension extends to the character. In Metropia, tiny “pry bots” handle much of a detective’s routine work, but Gunn and Eye-Ra continue to investigate in person. The setup gives Bird a science-fiction world while keeping the story rooted in a familiar hard-boiled mystery.

A project older than The Iron Giant

Netflix describes Ray Gunn as a project Bird has carried for more than 30 years, dating back to before The Iron Giant, his 1999 animated feature. The idea stayed with him while he went on to direct The Incredibles, Ratatouille and other live-action and animated films.

That long gestation is part of the film’s appeal, but it also makes the trailer a first test of a concept that has been changing alongside the animation industry. Rather than adapting an existing franchise, Bird is bringing an original detective story to a large-scale animated production at Skydance Animation.

Bird has also said he wants to reach viewers who assume animation is mainly for children. The noir influences, adult crime story and retro-science-fiction setting give Ray Gunn a different pitch from a conventional family comedy. Whether audiences follow that invitation is still an open question; the trailer is a first look, not a verdict on the finished film.

A limited 70mm run before Netflix

The December release gives the film a short theatrical lead-in. Netflix’s wording is specific: the movie will appear in select theaters, with limited 70mm presentations, on December 4. It reaches the streaming service two weeks later, on December 18.

The 70mm engagements fit the film’s emphasis on a densely designed city and cinematic scale, though Netflix has not announced a broad theatrical rollout. For a project that Bird has wanted to make for decades, the trailer introduces a world built from older influences and a format that asks some viewers to see it on a large screen.

There is still a long wait before audiences can judge whether the finished movie delivers on that world. For now, Ray Gunn has moved from a long-running idea to a dated release, with its first trailer finally showing what Bird has been imagining.

Sources

• Netflix Tudum: Ray Gunn trailer and Brad Bird interview
• Netflix Tudum: Ray Gunn first look and project background
• About Netflix: 2026 animation slate and Ray Gunn release details
• Ars Technica: Ray Gunn trailer, cast and release window
• TheWrap: Brad Bird on Ray Gunn and its theatrical presentation

The post Brad Bird’s 30-Year Passion Project Ray Gunn Finally Has a Trailer appeared first on The FreshWire.

That does not mean the cloud is going away, or that every feature works without a server. It does suggest that AI is becoming less like a destination and more like a capability built into the computer itself.

The model is starting to fit on the machine

Apple’s M5 Ultra Mac Studio began arriving on September 22. Apple says the desktop can be configured with up to 512GB of unified memory and 1.2TB per second of memory bandwidth, enough for researchers to run enormous language models entirely on the device. The 512GB configuration is scheduled for late October, and the M5 Ultra model starts at $5,499.

There is independent evidence behind the basic idea, even if the top configuration has not shipped yet. Tom’s Hardware tested an M5 Ultra Mac Studio with 256GB of memory and ran local models, including Qwen 3.8-27B. That supports the case for local inference; it does not mean the reviewer tested Apple’s 512GB configuration or that every large model will run well on every desktop.

FreshWire’s earlier Mac Studio report covered Apple’s pitch for local AI. The practical change is straightforward: when a compatible model runs on the computer, that request does not need to be sent to a remote model for processing. That can matter for latency and for keeping some work on a machine the user controls. It does not prove that every assistant feature stays local, or that a powerful local computer is affordable for most people.

Phones are being designed for agents that can act

At the other end of the scale, Qualcomm announced its Snapdragon 8 Elite Extreme Gen 6 and Snapdragon 8 Elite Gen 6 phone platforms on September 22. Qualcomm is marketing them around “agentic AI” and on-device intelligence. TechCrunch separately reported that the Extreme version can run a 30-billion-parameter mixture-of-experts model locally.

That is a more ambitious pitch than putting a chatbot shortcut on the home screen. A phone is already where people manage calendars, photos, messages and apps. An assistant that can work across those tasks could save steps, but only if its access is clear and its actions are easy to review, stop or undo. A mistaken answer is frustrating; a mistaken message or purchase can have consequences.

Qualcomm’s announcement describes platform capabilities and a coming device ecosystem, not a guarantee that every phone will perform every task locally. FreshWire’s Snapdragon coverage has more on the chip launch and the company’s agent-focused claims.

Google is building a laptop around Gemini

Googlebook makes the operating-system shift especially visible. Google has opened preorders for the new laptop line at $899, with U.S. retail availability beginning October 4. The company describes Googlebook as designed for Gemini Intelligence, with features such as Magic Pointer for working with what is on screen, Rambler for turning speech into organized text, and Create My Widget for building small tools from a description.

These are Google’s descriptions of a new product, not proof that every feature will be useful in daily work. TechCrunch’s early analysis noted that parts of the Magic Pointer idea overlap with existing Android tools such as Circle to Search. That raises a real question for Google: can the operating system make the assistance feel coherent enough to justify a new laptop, rather than simply adding another AI button?

Googlebook also mixes local hardware with cloud services. Google says the laptops include a year of Google AI Pro, and some capabilities rely on Google’s online services. FreshWire’s Googlebook report covers the $899 starting price, preorder timing and the platform’s other features.

Meta is splitting the computer between your face and your pocket

Meta’s new VR Glasses take the interface off the desk. Announced September 23, the glasses weigh about 100 grams and put Meta AI into the operating system. Meta says users can speak, use eye movement and make hand gestures to open apps, play video or adjust a workspace.

The light frame depends on a separate puck connected by an optical tether. Meta says the puck handles computing, battery and storage; Axios also tried a prototype and described the puck as a necessary part of the system. The glasses are planned for spring 2027 at $1,299.99.

That design makes the “computer disappearing” idea literal in one way: the display and sensors sit near your eyes, while much of the processing moves into a device in your pocket or bag. The computer has not vanished. It has been divided between a wearable interface and a tethered box. FreshWire’s Meta VR report has the product’s announced features and release window.

The cloud still has a job

None of these launches makes data centers obsolete. Local models can be useful for tasks that benefit from speed, offline access or keeping a particular request on the device. Cloud models can still handle larger jobs, frequently updated services and features that a phone or laptop cannot run by itself. In practice, the work is likely to move between the two.

Cloud AI is also getting cheaper. OpenAI says GPT-6 Sol and Luna cost 50% less per token than its GPT-5.6 promotional pricing. That is OpenAI’s comparison against its own promotional rates, not an industry-wide price measure. FreshWire’s GPT-6 coverage puts that announcement in the broader model-price competition.

The likely arrangement is a division of labor: a device handles some private or time-sensitive work, while a remote model takes on jobs that need more scale. Ideally, the person using the device will not need to think about where each step runs. But companies still need to make that boundary visible when it affects privacy, cost or control.

A built-in assistant needs built-in limits

Moving AI into the operating system changes the questions users should ask. What can the assistant read? Which actions require confirmation? Where does a request go when the local model cannot handle it? Can the user see what happened and reverse it?

On-device processing can reduce the information sent away for a particular task, but it does not answer those questions by itself. Nor does calling a feature an agent explain how much access it receives. A useful system will need permissions people can understand, clear handoffs to cloud services and confirmations before consequential actions.

The announcements do not prove that people want an AI agent in every device. They show where the industry is placing its bets: models closer to the screen, software ready to act across apps, and new hardware built to keep those features nearby. Chatbots may have taught people to speak to computers. The next test is whether computers can help without taking control away from the person using them.

Sources

• Apple: M5 Ultra Mac Studio announcement and on-device AI claims
• Apple: Mac Studio availability, pricing and 512GB timing
• Tom’s Hardware: Independent Mac Studio M5 Ultra testing
• Qualcomm: Snapdragon 8 Elite Gen 6 announcement
• TechCrunch: Qualcomm’s new smartphone chips and AI focus
• Google: Googlebook pricing, preorders and availability
• Google: Gemini features built into Googlebook
• TechCrunch: Googlebook’s Gemini strategy and early assessment
• Meta: VR Glasses weight, design, AI controls and release plans
• Axios: Hands-on with Meta VR Glasses
• OpenAI: GPT-6 Sol and Luna pricing announcement

The post AI Is Disappearing Into the Computer Itself appeared first on The FreshWire.

The opening lineup is being built by Acer, Asus, Dell, HP and Lenovo. Google describes Googlebook as a new laptop category that combines the Android technology stack with desktop foundations from ChromeOS. That places the product between Google’s phone ecosystem and the company’s established Chromebook market, with Gemini positioned as part of the main sales pitch.

A laptop designed to follow your Android phone

Google says Googlebook can bring over an Android phone’s settings, saved passwords, Wi-Fi networks and messages during setup, with that transfer protected by end-to-end encryption. The company is also promising a desktop version of Chrome, support for Android apps and an isolated Linux environment with a full terminal for developer tools and coding agents.

The pitch is continuity: start a task on a phone, then finish it on a larger screen without rebuilding your setup. FreshWire previously covered Google’s Pixel 10 launch, another step in the company’s effort to make its Android devices work together. The new laptop will have to make those connections feel dependable, not merely convenient in a launch demonstration.

Gemini is built into the interface

Google is giving the Gemini assistant several ways into everyday computer tasks. Magic Pointer is designed to appear when a user wiggles the cursor, then help with the text, images or other context on screen. Google says the feature stays inactive until summoned. Its examples include turning a highlighted training plan into calendar entries and asking Gemini to assess a suspicious email.

Rambler is a voice-dictation tool that is meant to clean up rough notes and organize them into text. Create My Widget lets people describe a small utility they want, such as a sports-score tracker, while Google’s Antigravity coding tool and the Linux terminal aim at developers. These are Google’s product descriptions; how well they work in daily use will depend on accuracy, app compatibility and how much access users are comfortable granting.

On hardware, Google says launch models will use Intel or Qualcomm processors, dedicated neural-processing units rated above 45 trillion operations per second, and battery life of up to 14 hours. The company is also promising premium materials, high-resolution OLED display options and a 10-year window of regular updates for Googlebook OS.

The price makes this a bigger PC bet

Every Googlebook starting at $899 includes 12 months of Google AI Pro, which Google says adds 5TB of cloud storage and higher Gemini usage limits, plus three months of YouTube Premium, Adobe Photoshop and other bundled services. The inclusion makes the laptop’s first-year package look broader, but buyers should compare the ongoing costs of subscriptions after those offers end.

Google says U.S. sales begin October 4; models arrive October 5 in Canada, the U.K., Ireland, France, Germany and Australia. TechCrunch described the product as a move up from traditional Chromebooks, and Axios also framed it as a bid for the premium laptop market. At $899, Googlebook is asking customers to consider a new operating-system mix at a price where established Windows and Mac laptops are familiar alternatives.

The bet is that Android phone owners will value the handoff between devices, while Gemini can make a laptop feel more helpful than a conventional desktop interface. Google has laid out the software pieces and recruited major hardware partners. The next test is whether the first devices deliver enough everyday value to justify buying into a new kind of Google computer.

Sources

• Google: Googlebook preorders, partners, pricing and availability
• Google: First look at the Googlebook laptop line
• Google: Gemini features built into Googlebook
• Google: Introducing Googlebook
• TechCrunch: Googlebook is a bet that buyers will buy a laptop for Gemini
• Axios: Google takes aim at premium laptops with Googlebook

The post Googlebook Starts at $899 as Google Bets Android Can Power a New Kind of Laptop appeared first on The FreshWire.

A viewer could ask for a feed of long documentaries for an evening wind-down, new do-it-yourself creators or podcasts suited to a commute. YouTube says people will be able to save several feeds as dedicated tabs on the homepage, with recommendations refreshed automatically. The feature is coming soon to users in the United States on the web, mobile devices and TVs; the company has not given a more specific launch date.

A more direct way to steer recommendations

YouTube’s usual Home page builds its recommendations from signals such as viewing and subscription activity. Custom Feeds add a clearer instruction: describe an interest, then let YouTube assemble a dedicated stream around it. The company says viewers can save multiple feeds and switch among them from the homepage.

That makes the change less like replacing the recommendation system and more like giving viewers another way to guide it. The ranking still depends on YouTube’s systems deciding which videos fit a prompt. But a direct request could be easier to manage than trying to retrain a single feed after a few searches or a binge on one topic.

The scale helps explain why YouTube is experimenting with discovery controls. The company says more than 20 million videos are uploaded to the platform every day. A larger library gives people more choice, but also makes it harder to find something beyond what the main feed predicts they will watch next.

AI is moving deeper into the creator workflow

Custom Feeds were one part of the Made on YouTube announcements. The company also described a conversational editing assistant for Shorts and the YouTube Create app. Creators will be able to start with a prompt, then ask the assistant to trim pauses, reorder footage or sync edits to music. YouTube says creators can still adjust the timeline manually.

YouTube Studio is also getting personalized feedback on draft videos, channel-matched thumbnail generation and a new video test that will let creators compare up to three cuts. The cut-testing feature is coming soon. YouTube says creators have run more than 40 million title and thumbnail experiments since that A/B testing feature launched in 2024; the new tool extends that approach from packaging into the video itself.

The company is also bringing its likeness-detection tools to mobile, so creators can review matches and take action from a phone. Later this year, YouTube says it will combine speaking-voice detection with facial detection to improve match accuracy. The announcements point to a platform automating more of video production while trying to help creators identify unauthorized uses of their likeness.

What viewers should watch for

Custom Feeds are the most immediately visible change for viewers, but they are not a promise of a fully user-controlled algorithm. YouTube still chooses the videos and ranks them. The useful test will be whether a prompt produces a feed that stays focused over time, especially as new videos are added and interests change.

The idea also reflects a wider shift in platform design: recommendation systems are becoming more conversational. TechCrunch’s report framed the feature as a way for viewers to build a more personal algorithm. YouTube has confirmed the prompts and saved feeds, but it has not yet said how much control people will have over tuning or correcting each feed once it is live.

Sources

• YouTube: Made on YouTube 2026 announcements
• YouTube: Custom Feeds and viewer features
• YouTube: New creator tools and conversational editing
• TechCrunch: YouTube will let viewers build their own algorithm with AI

The post YouTube Will Let You Build Your Own Homepage With Prompts appeared first on The FreshWire.

Discord CTO Stanislav Vishnevskiy announced the change on September 22. The company says the rollout starts the following day and will reach accounts over the course of the week. The automated model is not currently available to users in the U.K. or Australia, where Discord has used region-specific age-assurance measures. Other teen-safety details may also vary under local rules.

What Discord’s model looks at

Discord’s model estimates whether an account most likely belongs to an adult, a teen or an unconfirmed group. The company says it uses signals such as account age, the communities and games an account is connected to, general activity levels and device information. Its technical explanation also describes aggregated activity counts, including how many messages an account sends.

That distinction matters: Discord says it does not read the words in messages, inspect voice calls or use uploaded media to make the estimate. It does analyze account-level and behavioral patterns. The company says no single server or friend determines an account’s age group, and that the result is not displayed on a user’s profile or to other people in a server.

Accounts categorized as adults should see little change. Teen accounts, which Discord defines as ages 13 to 17, receive added protections around age-restricted content, servers and settings. They can still message friends, join voice calls and use non-age-restricted servers. Discord says accounts the model cannot confidently classify receive teen-style safeguards until they confirm their age group.

Confirmation is for a smaller group

Discord says people do not need to confirm their age just to keep using core features. An adult whose account remains unconfirmed may need to verify adulthood to open age-restricted spaces or settings. The company says users can check their status in User Settings under Account Status.

Available options vary by account and region. Discord lists age-range sharing from the Apple App Store or Google Play, a supported ID pass in Google Wallet, a credit-card check and reusable AgeKey credentials among options that do not require an ID scan or selfie. Video-selfie age estimation runs on the user’s device, according to Discord. For an ID scan, its age-assurance partner k-ID routes the document to Veratad; Discord says the vendors delete the selfie and extracted document data after the check, and Discord receives only an age result.

A history that makes privacy central

The new model follows backlash to Discord’s earlier age-verification plan, which the company delayed after users worried that ordinary access could require biometric scans or government documents. TechCrunch reported that the revised rollout uses account signals for most people and reserves direct confirmation for cases where it is needed.

The concern is not abstract. Discord said roughly 70,000 users may have had government-ID photos exposed in a 2025 breach of 5CA, a third-party customer-service provider that handled age-related appeals. The company says it has retired that process and no longer works with that provider. The FreshWire previously noted the incident in its FCC and NextGen TV roundup.

Discord’s approach avoids asking nearly every user for a sensitive document, but it creates a different trust question: how comfortable are people with a platform inferring their age from account history and participation patterns? The company acknowledges the model can make mistakes. How it handles false classifications and appeals will matter as much as the privacy design it describes.

Sources

• Discord: Safer for Teens, Same Discord for Adults
• Discord: How it estimates age without collecting identity
• Discord: Age-assurance methods, vendors and data handling
• Discord: 2025 third-party customer-service security incident
• TechCrunch: Discord’s age-verification rollout follows user backlash

The post Discord Begins Age Assurance, Says Most Users Won’t Need an ID or Selfie appeared first on The FreshWire.

The new Mac Studio and Mac mini went on sale September 22. The 512GB Mac Studio configuration is due in late October, so it is not available at launch.

A desktop with room for larger models

Apple’s M5 Ultra Mac Studio can be configured with an up-to-36-core CPU, an up-to-80-core GPU and 1.2 terabytes per second of memory bandwidth. The M5 Ultra model starts at $5,499; Apple has not listed the final price of the 512GB configuration in its availability announcement.

That memory capacity matters because local AI models must fit their weights and working context in memory. Exactly how much a system can run depends on the model, software, precision and quantization. A 512GB machine does not make every model fast or practical, but it raises the ceiling for researchers and studios experimenting with larger systems.

Apple is also pitching clustered Mac Studios for distributed inference. The company says Thunderbolt 5 with Remote Direct Memory Access can connect multiple systems and deliver up to three times the performance of one Mac Studio in its own testing. That is a vendor claim, and the result will depend on the models and workflows being run.

Local AI comes with a large entry price

The Mac Studio is a professional workstation, not a low-cost computer for trying a chatbot. For a smaller desktop, the updated Mac mini with M6 starts at $899 with 16GB of unified memory and can be configured to 32GB. Mac mini with M5 Pro starts at $1,699 and supports up to 64GB.

Local inference can give a team more control over where a model and its data run, depending on the software being used. It can also reduce dependence on an outside API for some tasks. The tradeoff is the upfront hardware cost, and cloud data centers still have access to far more combined computing power.

Apple’s pitch fits a broader shift in its Mac strategy. The company has been emphasizing custom silicon and device-level features, from its Liquid Glass redesign at WWDC to this push for larger local AI workloads. The Mac Studio makes that direction unusually concrete: nearly half a terabyte of memory is meant to be useful for the work people run, not just a headline specification.

Sources

• Apple Newsroom: The new Mac mini and Mac Studio are available today
• Tom’s Hardware: Mac Studio M5 Ultra review and local AI testing
• The FreshWire: Apple’s Liquid Glass and iOS 26 coverage

The post Apple’s New Mac Studio Can Run Large AI Models Locally With 512GB of Memory appeared first on The FreshWire.

Announced September 22, the Snapdragon 8 Elite Extreme Gen 6 and Snapdragon 8 Elite Gen 6 give phone makers two tiers for their next flagship devices. Both use a 2-nanometer process and Qualcomm’s custom Oryon CPU, Adreno graphics and Hexagon neural-processing hardware.

More AI, with more of it on the phone

Qualcomm describes the chips as a platform for “agentic AI”: assistants that can take a goal, plan several steps and use apps or tools to complete it. The company says a new sensing hub can run models with up to 200 million parameters for low-power, context-aware tasks. TechCrunch reports the higher-end Extreme model can also run a mixture-of-experts model with 30 billion total parameters locally.

Those figures describe what the silicon may support, not a ready-made phone assistant. Whether a device can use an agent to search across apps, organize a trip or carry out a multi-step request will depend on the phone maker’s software, Android integrations, app permissions and the model itself. The chip can provide capacity; it cannot make every app safe or compatible on its own.

Local processing can reduce the need to send some information to a remote service and may make supported features respond faster. It also has limits: larger models take memory and power, and a phone maker decides which capabilities reach customers. “Agentic” is Qualcomm’s product framing, not a guarantee that launch phones will act autonomously across the system.

The upgrade case goes beyond assistants

Qualcomm is also bringing machine-learning features into graphics and camera processing. The Extreme model adds Adreno Neural Fusion, which Qualcomm says can improve game visuals and efficiency; its camera platform includes new video and image-processing features. Those smaller, task-specific tools may be easier to notice than a phone-wide agent.

Qualcomm lists HONOR, iQOO, Motorola, OnePlus, OPPO, REDMI, RedMagic, vivo and Xiaomi among the manufacturers planning devices based on the new platforms. It has not named individual phones or retail dates in this announcement, so the first real test will come when manufacturers show how they use the chips.

Android phones have had on-device AI features for years, from image processing to speech recognition. The new Snapdragon generation raises the ceiling for those workloads, but the specifications alone do not settle whether an AI agent will be useful enough to shape anyone’s next upgrade.

Sources

• Qualcomm: Snapdragon leads the agentic AI age with two new flagship platforms
• TechCrunch: Qualcomm launches two new smartphone chips with an emphasis on AI

The post Qualcomm’s New Snapdragon Chips Bet on AI Agents in Android Phones appeared first on The FreshWire.

On September 22, OpenAI introduced GPT-6 Sol and GPT-6 Luna, while Anthropic released Claude Opus 5.5. The companies are still making capability claims, but both announcements put efficiency and long-running agent work front and center.

Lower prices for work that runs longer

OpenAI lists GPT-6 Sol at $2 per million input tokens and $10 per million output tokens for standard prompts up to 272,000 input tokens. GPT-6 Luna is $0.10 per million input tokens and $0.50 per million output tokens. Those figures compare with $4 and $20 for GPT-5.6 Sol and $0.20 and $1.20 for GPT-5.6 Luna. OpenAI says the new models are 50% cheaper than the promotional pricing of their predecessors.

The company says the cost question matters more as coding agents take on larger jobs. OpenAI put its own internal use in striking terms: valued at API prices, daily token usage exceeded $600 for the median researcher and $7,000 for researchers at the 90th percentile. That is a company-reported estimate for intensive internal work, not a measure of ordinary chatbot use.

GPT-6 Sol and Luna launched in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users, while Free and Go users could access Luna through the desktop app. At launch, OpenAI said the models were not yet available in ordinary Chat. Both are available through the API.

Anthropic is cutting the bill too

Claude Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens, down from $5 and $25 for Opus 5. Anthropic also cut cached reads from $0.50 to $0.20 per million tokens. The company says Opus 5.5 costs about 40% less on typical workloads and generates output more than 30% faster than Opus 5.

Those comparisons come from Anthropic’s testing and should be read as vendor claims. The release still points to a real change in the market: providers are competing over the price of completed work, not only benchmark scores or the cost of a single prompt.

The agent economics test

An agent can search files, call tools, make a change, check the result and try again. Each step adds time and compute. A cheaper model can make more experiments practical, but a low token price is not enough if a system needs many more attempts to finish the job.

That makes cost per successful task a more useful question than cost per token alone. OpenAI and Anthropic are both asking developers to let their systems tackle longer, messier work. Whether the savings hold up will depend on how reliably each model completes that work.

The AI race is still about capability. These launches show that it is also becoming a contest over how much useful work a dollar can buy.

Sources

• OpenAI: Introducing GPT-6 Sol and Luna
• OpenAI API pricing
• Anthropic: Claude Opus 5.5
• TechCrunch: OpenAI launches GPT-6 Sol and Luna

The post OpenAI and Anthropic Just Made High-End AI Much Cheaper appeared first on The FreshWire.

The layoffs arrive alongside a major reshuffle of who runs some of Xbox’s best-known franchises. Most notably, Activision will develop the next Halo game.

Activision gets a broader role

Xbox says Activision will add Rare and World’s Edge to its remit. A new, purpose-built Activision team will develop the next Halo title, separate from the ongoing work and plans for Call of Duty. A smaller Halo Studios team will remain to support the existing games and community.

The arrangement would have been difficult to imagine before Microsoft completed its roughly $69 billion Activision Blizzard acquisition in 2023. The FreshWire covered the deal when the UK regulator cleared it. The new structure turns Activision into a much larger organizing center for Xbox’s game development.

Studios move under new umbrellas

Obsidian Entertainment is moving under Bethesda while continuing its existing projects, including Grounded, and a new Fallout project being developed with Bethesda Game Studios. Playground Games and Turn 10 Studios are being combined into one studio focused on Forza and Fable. King will take responsibility for Microsoft Casual Games.

Other changes move in the opposite direction. Compulsion Games and Double Fine returned to their own management in August with their intellectual property, back catalogs and runway funding. Undead Labs has now made a similar transition with a new publisher; Xbox says State of Decay 3 is still planned for a day-one release on Game Pass.

Ninja Theory faces a proposed closure

Xbox says two separate agreements involving Ninja Theory fell through. The company will begin employee consultation on a proposed closure while continuing to explore other options. That does not mean the studio has already closed: the announcement describes a proposal and a consultation process.

For players, the changes raise questions beyond the number of teams on Microsoft’s organization chart. Consolidating studios could simplify management, but the company is also reducing jobs and changing which groups control long-running series. Xbox’s wide-ranging lineup looked very different in The FreshWire’s 2025 Games Showcase roundup. The next releases will show whether this reset helps teams ship games consistently while preserving the different voices that made those studios valuable.

Sources

• Xbox Wire: Continuing Our Reset
• GeekWire: Microsoft cuts jobs and shifts next Halo to Activision
• The FreshWire: UK regulator clears Microsoft’s Activision Blizzard deal

The post Xbox Cuts 268 Jobs as Activision Takes Over the Next Halo appeared first on The FreshWire.