Connect with us

Tech

Four AI agents coordinating in real time outperformed Claude Opus 4.8 on enterprise coding tasks

Published

on

As enterprise codebases grow, AI agents tasked with analyzing them are buckling under the weight of long-horizon tasks that require multiple interactions and tool calls. Dividing the work among a team of agents seems like the obvious fix, but it introduces a fatal flaw: most multi-agent systems are not designed for agents to coordinate among themselves mid-task and in real time.

To solve this, researchers at Coral AI Labs and multiple universities introduced AgentRadio, an asynchronous message-passing layer that allows agents to communicate between their execution steps without interrupting their main work. In real-world enterprise applications where subtasks are highly interdependent, this architecture enables agents to make mid-course corrections rather than continue on dead-end paths until a formal review phase.

On a benchmark of long-horizon questions over production repositories, a team of agents powered by AgentRadio nearly doubled task accuracy for four Claude Code agents working independently. It also outmatched single agents running on more advanced models. For AI practitioners, AgentRadio shows that the right coordination structure can outmatch raw compute and model scale.

The challenge of codebase understanding

LLM-based agents are increasingly capable of handling long-horizon tasks that require interacting with different tools and environments. Codebase understanding represents an extreme version of this challenge. It requires an AI agent to build the software, execute it, trace execution paths across multiple files, and synthesize evidence over extended periods.

Advertisement

Under these conditions, single-agent systems usually break down because of a “coverage problem.” 

"A single agent follows one serial path through the repository," Xinxing Ren, Caelum Forder, and Peter Carroll, co-authors of the AgentRadio paper, explained to VentureBeat. As its context grows, "the initial plan becomes harder to revise and discoveries made late in the investigation do not always propagate." The model can usually execute individual steps, but "the hard part is keeping every obligation, dependency, and piece of contradictory evidence active across a long investigation."

One benchmark that helps measure AI performance on large codebases is SWE-Atlas QnA. This benchmark consists of long-horizon, natural-language questions over live production repositories. The tasks can’t be solved by just exploring the code. AI agents must run the software and execute multiple commands to find the answers.

According to the research team’s experiments, a single Claude Code instance running on Opus 4.6 resolves just 32.3% of these tasks. Upgrading to a newer, more advanced model like Opus 4.8 only yields a 57.2% success rate.

Advertisement

A natural remedy is to distribute the workload across multiple agents, allowing each to work with a smaller, cleaner context. Multi-agent solutions can provide substantial performance gains when tasks are cleanly decomposable, meaning they can be solved separately and merged at the end.

Codebase understanding, however, is rarely cleanly decomposable. The subtasks are highly interdependent. A critical configuration file or a bug uncovered by one agent can completely rewrite or redirect the entire exploration path of another agent. Because of these dependencies, agents must coordinate, negotiate, and share intermediate discoveries in real time.

Despite this need, asynchronous multi-agent communication is rare. The researchers point out that existing multi-agent systems generally fall into three flawed patterns:

  • Parallel but isolated: Agents operate simultaneously but do not communicate at all.

  • Parallel but round-synchronized: Agents can communicate, but only at strict, synchronized round boundaries. This forces agents to stop and wait for one another to finish a round before they can debate or exchange intermediate findings. Round-based systems assume that important discoveries can wait until the next communication phase, which is an expensive assumption when agents are working on interdependent parts of a live system. For example, an agent investigating an API symptom might uncover evidence that invalidates the storage agent's current hypothesis. "If that information waits until both agents finish, the storage investigation may complete along the wrong path," the researchers said.

  • Asynchrony in adjacent forms: These systems offer limited asynchronous features, such as top-down task dispatching. They don’t have peer-to-peer lateral channels between agents or shared memories that require an agent to actively pause its work to read updates.

In their paper, the researchers point out that the main bottleneck hindering current multi-agent systems is that “an agent that is working cannot also be listening.”

Advertisement

“To our knowledge, no existing system gives concurrently working agents passive awareness of one another over a lateral, natural-language channel,” the researchers write.

How AgentRadio works

To dissolve the mutual exclusion between working and listening, the researchers developed AgentRadio, an asynchronous message-passing layer designed to plug directly into existing coding-agent harnesses.

AgentRadio equips agents with three primitives:

  • The create_thread primitive opens a conversation between participating agents.

  • The send_message primitive appends a message to a thread and returns without blocking the sending agent.

  • The wait_for_mention primitive blocks the process until a message mentioning the caller arrives. It delivers the message along with a full snapshot of all threads so the agent has instant context. 

This trio enables agents to have a state of “passive awareness,” where they can continue their primary tasks while passing messages and updating their knowledge in the background.

Advertisement

AgentRadio's code is available under the Apache 2.0 license on GitHub. It is designed to be lightweight, requiring no direct modifications to the underlying agent harnesses like Claude Code or Codex CLI. 

The architecture consists of two main parts:

  • The message server: A standalone process that acts as the central hub, storing all active threads, messages, and mentions for the group of agents.

  • Harness-side integration: Agents interact with the server using three simple shell scripts, one corresponding to each primitive.

The only strict requirement for the system to work is that the agent harness must be able to run a shell command as a background task. The agents are instructed in their system prompts to keep one watcher running and to send messages through the provided scripts. Running the wait_for_mention script in the background allows the agent to continue its work and receive notifications asynchronously.

To integrate this into an existing stack, a team still needs a "thin adapter that starts the workers, assigns identities, connects them to the shared server, and manages final synthesis," the researchers said. That work sits around the coding agent rather than requiring changes to the underlying model.

Advertisement

AgentRadio in action

To validate the real-world utility of AgentRadio, the researchers tested the framework on 124 tasks from the SWE-Atlas QnA benchmark. The tests covered domains including system design, root-cause analysis, security, and API integration.

The researchers used Claude Opus 4.6 and DeepSeek V4 Pro as the backbone models. For the harness, they evaluated configurations ranging from a single Claude Code agent (B0) to a team of agents with classic division of labor (L1), up to a team of agents using AgentRadio to coordinate asynchronously (L3).

The experimental results showed that the AgentRadio communication architecture outperforms both naive multi-agent setups and raw compute scaling.

While a single Claude Code agent with Opus 4.6 resolved only 32.3% of the tasks, the full AgentRadio setup nearly doubled that metric, resolving 62.1% of the tasks, and surpassed the single agent running on Opus 4.8, which hit 57.2%. It also boosted the DeepSeek V4 Pro results from 29.0% to 50.8%. 

Advertisement

To understand how this practically impacts enterprise AI, the paper highlights a real-world task involving a MinIO system. Solving the task required checking per-request server logs, a requirement the agents did not anticipate during their initial planning phase.

In the L2 setting, where agents collaborate but lack asynchronous communications, two agents independently realized they needed these logs while executing commands. Because they could not share this finding mid-execution, one agent gave up privately and the other failed to propose it to the team. During the review phase, the team unanimously agreed on the wrong answer, missing five rubrics.

With AgentRadio activated, the agents made the same mid-execution discovery, but one agent instantly broadcasted the required server-side log evidence to the shared worklog. Because the other agents were passively listening, they absorbed this new evidence immediately. This real-time coordination transformed a failing score into a perfect 16 out of 16.

"The useful distinction is timing," the researchers said. "The team did not need another agent or another review round. It needed one agent's discovery to reach the right peers before its operational value expired."

Advertisement

The researchers note that the same pattern appears in enterprise incident work. For example, an agent investigating an API symptom might uncover evidence that invalidates the storage agent's current hypothesis. If that information waits until both agents finish, the storage investigation may complete along the wrong path. “Passive awareness lets the second agent incorporate the contradiction at its next work step without interrupting a command already in progress,” they said.

The cost and complexity of coordination

AgentRadio requires a fixed multi-agent team budget, which inherently multiplies the token cost. The researchers acknowledge that the "tax is real," noting that average API spend rose from $2.96 per task for one Opus agent to $19.45 for the full AgentRadio stack.

However, raw scale does not equal performance. When researchers compute-matched the test by spending $17.76 on six independent Opus runs, the models only resolved 37.9% of tasks, compared with 62.1% for AgentRadio. This suggests that AgentRadio's architecture is a structural win, not just a brute-force scale win. Teams should still be aware of inter-agent churn. "Communication can redirect an agent toward better evidence, and it can also distract an agent from a valid path," the researchers warned.

A fixed multi-agent team should not become the default response to every engineering task. The more useful test to determine if a multi-agent setup is required is whether the task contains "responsibility breakpoints," the researchers said. These are places "where a competent engineer would involve another person because the work crosses an ownership boundary, needs an independent hypothesis, or carries enough risk to justify separate verification."

Advertisement

“Coordination is a strong fit when the task can be decomposed, the resulting parts remain interdependent, the single-agent success rate is unreliable, and an incomplete answer has a meaningful downstream cost,” the researchers said. Examples include repository-wide architecture questions, unfamiliar legacy systems, cross-service incident investigation, security analysis, dependency migrations, and multi-module refactors.

Conversely, a single agent remains the cleaner choice for “bounded, local, and reversible work,” such as a known one-file change or boilerplate generation. 

“Use one agent while one context can still own the problem honestly,” the researchers said. “Introduce another responsibility when the existing agent would otherwise need to compress away evidence, cross an independent ownership boundary, or verify its own high-impact conclusion.”

From research to commercialization: Coral Code

While AgentRadio serves as a controlled research implementation using a fixed four-agent team and a five-phase protocol, the underlying principles are being adapted into a commercial product called Coral Code.

Advertisement

Instead of a rigid, multi-agent protocol applied to every ticket, Coral Code works from the bottom up. An engineer begins with their existing coding agent, and Coral introduces repository-scoped investigation, specialist responsibility, and communication only when the emerging evidence justifies it. "Coral packages the operational concerns around the tools engineers already use, providing the repository context, scoped specialists, communication, and evidence layer around the harness rather than inside it," the researchers said.

This dynamic approach optimizes costs by targeting the relevant unit: the cost of a completed, reviewable outcome.

The future of autonomous software engineering

While AgentRadio provides a major upgrade to agent orchestration, there are still hurdles to overcome. One major bottleneck that the researchers pointed out to is “attention governance and verification.”

“Passive awareness makes communication available during execution. It does not decide which agents should exist, which discovery deserves an interruption, who should receive it, or when the evidence is strong enough to revise the plan,” the researchers said. If every agent receives every update, the communication layer becomes noise. If several agents share the same bad assumption, faster communication can spread the error.

Advertisement

For example, in one of the case studies in the paper that involved the Grafana platform, four of nine rubrics required negative conclusions, such as observing that a datasource picker did not select automatically. The agents ran the relevant tests, yet none formed the missing negative hypothesis. Both configurations failed the four rubrics. 

“Passive awareness can distribute an idea that somebody develops. It cannot supply a conception that never appears anywhere in the team,” the researchers said.

As task durations stretch longer, communication and coordination become critical. "The next generation of systems… needs adaptive responsibility assignment, evidence-aware routing, conflict resolution, explicit cost limits, permissions, recovery, and clear human escalation points," the researchers note. Most importantly, it requires durable provenance so engineering leads can inspect which agent made a claim and why an action was accepted.

"Longer-running agents make communication more important. They also make accountability much harder to fake," they said.

Advertisement

Source link

Continue Reading
Click to comment

You must be logged in to post a comment Login

Leave a Reply

Tech

How To Hide And Unhide An App On Your iPhone Or iPad (And Why You May Want To)

Published

on

There are a few different ways to keep apps under the radar.

Sometimes, you don’t want your iPhone or iPad to display all installed apps right on the Home screen. The simplest method is long-pressing an icon, tapping the – (minus) icon next to it and choosing Remove from Home Screen. This banishes it to the App Library, but you can go further. 

Maybe you want to declutter your iPhone and keep the list tidy, keep getting distracted by social media or don’t want others seeing what you have installed. In these cases, iOS and iPadOS give you several ways to hide apps to keep them out of your (and others’) view. While these methods aren’t secret agent-level tricks that hide in plain sight, they’re suitable for keeping icons out of view and locking anything that needs extra security. In case a full lock and hide isn’t right for you, you can still remove apps from your screen or suppress their content around your phone. 

We focus on iPhone and Face ID for brevity, but this works the same way on iPad and models that have Touch ID.

Advertisement

The primary method for hiding iPhone apps

Prior to iOS 18, there was no official way to hide apps on iPhone or iPad; you had to stick one inside a folder to keep it out of view. Now, there’s a better method. To hide an app, long-press it on your device’s Home screen and choose Require Face ID. Doing so shows a new prompt that explains what happens upon locking. You have to authenticate with Face ID or your passcode to open it, notifications won’t contain any details and its content won’t show up in places like Spotlight searches or CarPlay.

It’s a smart idea to lock financial tools for a second layer of security. You’ll find lots of stories online of people being scammed by strangers who asked to use their phone in an “emergency.” Once they have your phone in hand, it only takes seconds to transfer money to themselves using Venmo or similar. Even though you shouldn’t hand your phone to a stranger in the first place, requiring re-authentication for the app prevents schemes like this.

Choosing Require Face ID keeps the icon where it is while requiring authentication every time you open it. Alternatively, pick Hide and Require Face ID to remove it from your Home screen and searches. After hiding an app, you’ll get no notifications from it. You can’t lock most default apps; this is limited to ones you’ve installed. And the status of what you’ve hidden doesn’t sync to your other Apple devices.

Advertisement

Hidden apps live in a special folder, aptly called Hidden, at the bottom of the App Library. Swipe all the way to the rightmost page to see this. That Hidden folder appears blank no matter how many apps you have hidden (including zero). Tapping it requires you to authenticate before you can see what apps are inside, and then you’ll need to verify again when opening one. To unhide something later, long-press on it in the Hidden folder, choose Don’t Require Face ID, and the app will return to the top of the list after it verifies you. You can then search for it using Spotlight and drag it back to wherever you’d like.

Hidden apps aren’t totally invisible

While this process greatly reduces an app’s visual presence on your device, it doesn’t erase all traces. Anyone with your passcode can browse the Hidden folder, as well as the same list under Settings > Apps > Hidden apps.

But there are other places where “hidden” apps can appear: Settings > Screen Time > See All App & Website Activity and Settings > Battery > View All Battery Usage. These show how long you’ve used each and how much battery it’s consumed on your phone, respectively. Even if it’s hidden, it will still show up in these spots if there’s data to report.

Advertisement

Apps you download also show up in your App Store purchase history (even free ones, despite the name). To hide these, open the App Store, then tap your profile icon at the top right and choose Apps & Purchase History. If you’re in a family group, tap Your Apps. Everything you’ve downloaded, including any not on your iPhone anymore, appears here. Swipe from right to left and choose Hide to remove an item from this list. If you change your mind later, go back to the top-level App Store menu, touch your name at the top, and choose Hidden Purchases. There, you can Unhide any to put them back in the list.

This method is more for decluttering than any form of privacy. You don’t need to authenticate to view the Hidden Purchases list, plus your transaction history is also kept under Purchase History in the Apps & Purchase History menu. Hiding purchases prevents anyone in your family group from redownloading them, but it won’t hide the apps themselves.

Advertisement

Other methods to keep apps under control

Hiding apps isn’t the only method of controlling what appears on your phone. As we’ve seen, it’s a good way to spend less time on addictive software, since it adds additional steps for access and hides all notifications. But you can add more restrictions for these.

Head to Settings > Screen Time to set a daily time limit for chosen apps (App Limits) or set hours when you can’t access them (Downtime). Also in this menu under Content & Privacy Restrictions > Allowed Apps & Features, you can disable access to some built-in tools (like FaceTime and Mail), which removes them from the Home screen. While many of these options are great for controlling a child’s iPad, anyone can take advantage of them.

If an app is bothering you too much, go to Settings > Notifications, select the offender, and fine-tune its notifications so you aren’t nagged. To make your iPhone stop suggesting an app, visit Settings > Apple Intelligence & Siri > Apps, choose one, then disable the sliders to stop your phone from suggesting it. And to remove an app from search, visit Settings > Search, select it and disable both sliders. Hiding an app does all of these, but you can take these actions without hiding if you want to keep it visible.

Hiding apps isn’t 100 percent foolproof, but it adds more steps for people to see what you’d rather them not. Maybe you don’t want others snooping around your phone when you hand it to them, or you want to add additional protection against someone trying to use your phone for nefarious purposes. For more protection, consider enabling Stolen Device Protection on your iPhone.

Advertisement

Source link

Continue Reading

Tech

Apple Watch faces big changes as new designs are considered

Published

on

The Apple Watch will see a revamp in the coming years, as Apple is reportedly in the middle of a shake up its wearables formula to meet an ever-changing market.

The Apple Watch has retained the same core design of a square display on a watch band, but it has evolved over time. While it has become thinner, larger, and more rugged in the Apple Watch Ultra, it can potentially do more.

In Sunday’s “Power On” newsletter for Bloomberg, Mark Gurman writes that the Apple Watch is ready to get overhauled into something new. The company’s industrial design team has been rethinking smartwatches in general over the last year, and it’s considering new directions to go in.

The team hasn’t quite decided what to actually do with the Apple Watch yet, but it is mulling the problem over.

Advertisement

These changes apparently include switching the square display for a circular version of various sizes. There’s also the possibility of removing the display altogether, to match the growing market for wearables without screens.

Smart rings are also seeing a rise in adoption, which gives Apple another wearable avenue to explore. It has repeatedly filed patents in the field, but an Apple Ring release continues to elude the company.

Sooner changes

Whatever new direction the Apple Watch takes, it won’t be anytime soon. The expected Apple Watch Series 12 and Apple Watch Ultra 4 will be largely performance-based updates.

Health and fitness changes are said to be included. Despite previous rumors, there are no major design changes in 2026 for the main Apple Watch body.

Advertisement

There is, however, a rumor that Apple will be introducing a new band attachment method for the 2026 model. If correct, it’s a change that will probably make earlier Apple Watch bands incompatible with the new model.

That said, it’s a rumor that we have heard on and off since 2022. It may happen, but it could be better suited as part of a larger overhaul.

Source link

Advertisement
Continue Reading

Tech

I pitted my dream PC from 2017 against an RTX 5090 monster in 2026 in a fight to the death – here’s what happened

Published

on

I was fourteen or so when I went to the Gadget Show Live (once a big tech event in the UK which showed off the latest technology) in 2016 as a punter and overheard someone on a stand I’d visited mention they’d spent £1,000 on a custom gaming PC.

A thousand pounds. On a computer. I remember turning that figure over in my head (it’s roughly $1,400) and not quite being able to make sense of it, not necessarily because I couldn’t fathom how much money that was, but because at the time, it seemed like a lot of money to spend on a PC without any frame of reference as to what that price got you.

Source link

Advertisement
Continue Reading

Tech

The Complicated Case of Passing On Your Digital Estate

Published

on

When a loved one dies, who downloads their important files from their cloud storage account? Who monitors their email inbox? Who decides what happens to the photos and videos on their social media accounts? And what if those tasks fall to you?

Everyone will die, but not everyone has planned what they want to happen with their digital assets after they’re gone. Even when someone makes a plan, survivors might still be limited in what they can do.

Tying up loose ends can become a nightmare for the living, especially when the volume of digital assets is enormous. Still, the more you know, the better you can plan for your own digital estate, and the easier it will be to manage someone else’s.

Take Inventory

The biggest determining factor in how much work it’s going to be to manage the online accounts and digital assets of someone who is incapacitated or deceased is whether they did any estate planning. If a person doesn’t write down what digital assets they have and what they want done with them, it’s impossible for anyone to know.

Advertisement

It’s not always a simple matter of memorializing a Facebook account or downloading photos from iCloud either. Digital assets can have as much monetary value as sentimental value. Say a person’s social media accounts earn dividends. How will a beneficiary collect future proceeds? And should they keep the account alive?

What about cryptocurrency? If it’s stored in a private wallet and no one has the key, the money is lost forever. It’s a different story, however, if a third party, like Coinbase or PayPal, holds the crypto. At present, bitcoin and other cryptocurrencies are considered “digital assets” and thus need to be treated that way when doing any estate planning.

Navigating the Law

In the US, digital inheritance is overseen by state law, the same as traditional probate and estate matters, according to Benjamin Orzeske, chief counsel at the Uniform Law Commission. He and his organization developed a state law known as the Revised Uniform Fiduciary Access to Digital Assets Act (RUFADAA), which has been enacted in 48 states, Washington, DC, and the US Virgin Islands. The missing two states are Massachusetts, where RUFADAA has been adopted but not yet enacted as of this writing, and Louisiana, which went its own way with a similar but different law.

“At the heart of RUFADAA is this recognition that digital property is in some ways different from traditional, tangible property,” Orzeske says. He gives the example of mail versus email. When a person dies, their mail gets forwarded to a dedicated person, the fiduciary, who then receives incoming communication, bills, and payments. If they get a bill in the mail for a magazine subscription, they know to cancel it. Receiving the mail effectively gives the person appropriate information and access to manage the deceased’s accounts and estate going forward. Email is different. The fiduciary doesn’t just get new incoming mail. They might also have access to a searchable history of communication, which the deceased person might have expected to be kept private.

Advertisement

The real point of conflict, according to Orzeske, lies in the Stored Communications Act, a federal law that says companies that handle our online assets can’t release them without our permission. So RUFADAA gives survivors some rights while retaining the original asset holder’s privacy.

Under RUFADAA, a named trusted person can close accounts, but they can only get the contents—meaning the bodies of emails, private messages, videos, photos, attachments, and so forth—if the decedent specifically “grants the authority to the personal representative fiduciary,” according to Catherine Hodder, a senior attorney editor at FindLaw. FindLaw is an informational website that breaks down legal issues for a general audience.

Source link

Advertisement
Continue Reading

Tech

Transmitting Analog Video Via Frikkin’ Laser Beams

Published

on

Transmitting analog video via photons is old hat: that’s how everything started, after all, back in the day with over-the-air TV. Up the frequency of those photons from radio to visible light, though? Well, now that’s rather interesting. [Daniel] aka [milar111]’s LYME 101– which doesn’t seem to stand for anything–laser-video transmitter/receiver pair was a strong contender in the recently-completed Frikken’ Laser Beams challenge, but somehow we missed putting it up on the blog.

The project is documented quite well on GitHub as linked above, as well as on Instructables, and Hackaday.io, and in a YouTube video we’ve embedded below so you can see it in action. In principle it’s pretty simple: a Raspberry Pi is used to generate the composite video signal, which modulates a red laser diode through a 2N2222 NPN transistor and some passives. The reciever is a BPW34 photodiode wired with reverse bias for speed and fed through a LM318N op-amp. To get +9V and -9V for this circuit, [Daniel] makes the easy hack of using a pair of 9V batteries for a noiseless dual supply. It hooks up to a CRT just fine, but a little finessing in the form of a terminator resistor and a DC bias pot on the transmitter were needed to get his USB capture card working with the signal.

It’s not the weirdest way we’ve seen people hack analog video signals– there’s no audio cassettes to be seen,  and the signal isn’t even SECAM, the oddest encoding— but that’s not a slight. Transmitting video with higher-than-normal-frequency photons might not be that weird, but it looks like a lot of fun.

Advertisement

Source link

Advertisement
Continue Reading

Tech

Suno Will Press Your AI Songs on “Vinyl” for $45. Who Exactly Asked for This?

Published

on

Vinyl has survived cassettes, CDs, Napster, the iPod, Spotify and several decades of people confidently predicting its imminent death because records continue to provide something streaming cannot: ownership, permanence, large artwork, ritual and a physical connection to music that someone presumably cared enough to write, perform, record and master. Artificial intelligence has apparently looked at all of that and decided it needed a piece of the action.

Suno is preparing to launch Suno Vinyl, a service that will turn music generated through its AI platform into personalized 12 inch records. Users will be able to select up to 46 minutes of music, create custom artwork and labels, and have the finished record pressed and shipped to their door for an estimated $45 plus shipping. The service is not yet live, although Suno is accepting users for its waitlist, which raises the fairly obvious question of who exactly has been sitting around waiting for this particular collision between generative AI and physical media.

That does not make Suno Vinyl an irrelevant story. Quite the opposite. Suno has become one of the largest generative music platforms in the world, raised hundreds of millions of dollars and is now operating at a scale where its decisions can influence how AI generated music is created, distributed and monetized. What makes this announcement interesting is not that somebody can turn an AI song into a record; it is that AI generated music is rapidly acquiring the same commercial infrastructure as conventional recorded music, including licensing, distribution, royalty policies, copyright disputes and now physical products.

Whether it deserves all of that is another question.

Advertisement

Calling It “Vinyl” Requires an Asterisk

There is one detail getting somewhat lost in the excitement: Suno calls the service Suno Vinyl, but the actual 12 inch record is manufactured from black PETG rather than conventional PVC vinyl. Suno describes the material as environmentally friendly and says the records are genuinely pressed rather than novelty lathe cuts, with full color printed sleeves, custom labels and audio taken from the user’s final mix.

suno-vinyl-pitch
Why Suno Vinyl exists (from Suno Vinyl website)

That is fine as far as it goes, but if you are charging roughly $45 for something intended to spend its life on a turntable, we would like considerably more information. Suno currently says nothing meaningful about the pressing plant, mastering process, record weight, manufacturing tolerances or whether the supplied file receives any additional preparation before being committed to the physical format. The company promises “high quality audio,” but without any technical context that phrase carries about as much useful information as “premium sound” on the side of a Bluetooth speaker box.

Those omissions would be troubling if Suno were positioning this as an audiophile product, although there is little indication that it is. The target customer is not likely to be comparing cutter heads, lacquer sources or different pressings of A Love Supreme before deciding whether a PETG copy of an AI generated ska song about the family Labradoodle deserves shelf space.

$45 Buys a Lot of Actual Music

The pricing becomes more amusing when placed beside what traditional record buyers can purchase for the same money. We recently reviewed Rhino High Fidelity’s new edition of Curtis Mayfield’s Super Fly, an all analog release cut by Kevin Gray and produced specifically for listeners who care about sound quality, and that record sells for $39.98.

Suno estimates about $45 plus shipping for a personalized PETG pressing of whatever its software generated after somebody typed a sufficiently elaborate prompt. The comparison is not entirely fair because Rhino can manufacture an established title in quantity while Suno is offering personalized, low volume production with custom artwork and packaging, but consumers do not make purchasing decisions inside an accounting seminar. They have $45, and that amount buys excellent new releases, several used records, a stack of inexpensive CDs or, depending on your level of restraint inside a record shop, the first twelve minutes of what eventually becomes a $140 purchase.

Advertisement

That is why Suno Vinyl makes almost no sense as a value proposition for conventional music collectors. It is not really competing with Rhino, Blue Note, Craft Recordings, Analogue Productions or any of the labels serving the vinyl enthusiast market. It is competing with personalized gifts, novelty merchandise and the emotional satisfaction of taking something that existed only on a screen and turning it into an object.

What Problem Is Suno Vinyl Solving?

Generative AI has removed almost every traditional obstacle involved in creating something resembling a finished song. You do not necessarily need musicians, a recording studio, microphones, instruments, engineers or enough musical ability to survive “Hot Cross Buns” on a recorder because software can produce something in seconds that has vocals, instrumentation, structure and a finished mix.

Advertisement. Scroll to continue reading.

If the result is terrible, you generate another one. If that is terrible, you generate ten more. Deezer recently reported receiving tens of thousands of fully AI generated tracks every day, which underscores one of the central contradictions of this entire market: we can now create music much faster than anybody can realistically consume it.

Advertisement

Against that backdrop, Suno’s decision to turn one of those infinitely reproducible digital files into a scarce physical object is almost comically logical. The technology creates abundance, while the record creates the appearance of permanence and scarcity. You have to admire the capitalism involved because Suno has effectively identified a way to assign physical and emotional weight to a product whose defining advantage is that it can be generated endlessly and almost instantly.

That may ultimately be the real appeal.

Vinyl Means Something AI Music Does Not

The resilience of physical music has very little to do with consumers suddenly forgetting how convenient streaming is. Hundreds of millions of people pay for streaming subscriptions precisely because instant access to almost everything is useful, yet vinyl and CDs continue selling because ownership still means something when the rest of our media lives increasingly exist behind subscriptions, licensing agreements and corporate servers.

People stream because it is convenient; they buy records because some music matters enough to own. A copy of Kind of Blue, London CallingSuper Fly or Songs in the Key of Life does not become meaningful because somebody pressed it onto plastic. The record exists because the music already demonstrated enough cultural or personal value that listeners wanted to keep it, display it and return to it.

Advertisement

Suno reverses that relationship. Pressing the music onto a record can become part of the process of convincing ourselves that the thing mattered in the first place, which is a considerably stranger proposition. Someone may create a song about a deceased parent, a wedding, a child’s birth or another deeply personal experience and treasure the resulting object, and there is nothing artificial about that emotional reaction. But Suno also promotes birthday songs, anniversary tracks and inside jokes as likely applications, which tells us far more about the real market than any grand claim about reshaping music creation.

So Who Is This Actually For?

Suno Vinyl is primarily personalized merchandise, and viewed through that lens the concept begins to make considerably more sense. It is for the person who generates a birthday song for Dad and thinks presenting him with an actual record will be funnier and more memorable than texting him a link. It is for couples who want their AI generated anniversary anthem sitting beside the wedding photographs, and for people creating fictional bands, joke albums and personal projects who simply want to hold the finished result in their hands.

There is also a more legitimate creative use case for musicians who employ AI as one tool inside a larger process involving their own lyrics, vocals, instrumentation or production. AI assisted music and fully AI generated music are not the same thing, something we have argued repeatedly in our recent coverage, and a physical pressing of a genuinely collaborative project may have real personal or promotional value.

But none of that turns Suno Vinyl into the next meaningful development in analog playback. For most customers, it is closer to a custom photo mug that happens to spin at 33⅓ rpm, which is not necessarily an insult because personalized products can be fun. It merely means the thing belongs in a very different conversation from an AAA Blue Note reissue, and nobody should expect Kevin Gray to arrive with tasting notes.

Advertisement

The Copyright Mess Now Comes With a Record Sleeve

The physical format also makes the unresolved legal and ethical issues surrounding AI music feel considerably more tangible. Suno says users who create songs while subscribed to its paid plans are considered the owners of those songs and retain commercial usage rights, while material created on its free tier is intended for noncommercial use. The company also acknowledges that ownership of a generated recording does not necessarily mean the work qualifies for copyright protection, particularly when the human contribution is limited.

That distinction becomes harder to ignore when the thing arrives in the mail with your name printed on the jacket.

Suno’s relationship with the traditional music business remains equally complicated. Warner Music Group settled its dispute with Suno and entered into a partnership intended to support future licensed models and new AI experiences involving participating artists, yet litigation involving AI training and copyrighted material continues elsewhere. A Munich court recently ruled against Suno in a case brought by German rights organization GEMA, illustrating just how unsettled the legal framework remains even as companies race ahead with new commercial products.

Advertisement. Scroll to continue reading.
Advertisement

That is why Suno Vinyl deserves more attention than a goofy product announcement normally would. AI music now has creation platforms, label partnerships, licensing agreements, streaming distribution, royalty policies, chart debates, copyright litigation and physical records. It has managed to recreate almost the entire music industry with remarkable speed, although nobody has yet announced an AI tour manager capable of losing the band’s cash somewhere outside Cleveland.

Physical Media Deserves Better Than More Content

We have spent much of 2026 covering the continued strength of physical music because records and CDs provide something increasingly valuable in an era of rented digital access. They offer ownership, preserve particular masterings and editions, support artists and labels, and encourage listeners to engage with albums rather than allowing an algorithm to pour background audio through the room until everyone goes to bed.

That is what makes Suno Vinyl simultaneously fascinating and faintly ridiculous. The vinyl revival has been driven partly by listeners deciding that certain music deserves more attention, more permanence and more emotional investment than another disposable stream, while Suno’s technology is built around making the creation of additional music almost frictionless. Those ideas are fundamentally at odds, and putting one inside the packaging of the other does not solve the contradiction so much as give it a rather handsome record sleeve.

The Bottom Line

Suno Vinyl is an important story attached to a profoundly unnecessary product. The company plans to let users take up to 46 minutes of music from their Suno libraries, add custom artwork and receive a genuinely pressed 12 inch PETG record for approximately $45 plus shipping, but there is almost no audiophile reason to care. Suno has disclosed very little about the mastering or manufacturing process, the record is PETG rather than conventional PVC vinyl, and the same amount of money can buy exceptionally well produced recordings containing music written, performed and engineered by human beings.

Advertisement

That comparison is ultimately beside the point because audiophiles are not really the customer. Suno Vinyl is a keepsake, novelty item, personalized gift and piece of creator merchandise designed for people who want something generated on a screen to feel more permanent, and that is precisely why the product matters even if most serious record collectors would never consider buying one.

AI companies have become remarkably good at generating limitless quantities of content. Suno is now testing whether putting some of that content into a physical object can make consumers assign it greater emotional and monetary value, and perhaps it can. If the next phase involves AI generated yacht rock pressed onto PETG, numbered 1 of 1 and listed on Discogs for $175, we may finally have reached the runout groove.

For more information: vinyl.suno.com

Advertisement

Source link

Continue Reading

Tech

Trump administration has spent nearly $4B to cancel offshore wind farms

Published

on

The Trump administration is paying another $1.2 billion to cancel an offshore wind lease, this time to German utility RWE. 

The energy company said in an announcement that the wind farms would have been built off the coasts of California, Louisiana, and New York. Instead, RWE will spend $900 million to buy a small stake in a Louisiana liquid natural gas export terminal.

The New York wind farm would have generated more than 3 gigawatts, according to Heatmap News.

The remaining $300 million will go toward buying natural gas turbines to power 15 peaking power plants around the country. Peaking power plants are among the most expensive and most polluting natural gas power plants to operate. It’s unclear when those will be completed — there’s currently a backlog for new turbines stretching into the early 2030s.

Advertisement

RWE isn’t backing off its offshore wind investments elsewhere, though. The company said it bought 6.9 gigawatts of capacity in the U.K.’s recent auction.

The Trump administration has paid $3.93 billion for the 12 leases it has coaxed developers into abandoning.

Source link

Advertisement
Continue Reading

Tech

This ‘adversarial’ pattern can prevent surveillance cameras from detecting you

Published

on

Bill Swearingen has spent the past year running largely the same test, over and over again. The goal was to produce a computer-generated pattern that could block the surveillance cameras lining America’s streets from detecting it.

Some 31 million tests later, Swearingen says he can now produce patterns on-demand that, when applied to clothing and objects, prevent some of the most commonly deployed license plate readers and surveillance cameras from detecting whatever the pattern covers, from people to vehicles.

His project, which he calls noRecognition, allows people to escape the automatic detection and algorithmic surveillance used across the U.S. and beyond.

In recent years, surveillance cameras have been supercharged with the ability to detect what is happening in the footage being recorded, from tracking the license plates of speeding vehicles to using facial recognition to identify suspected criminals, albeit with mixed success and sometimes terrifying results. The detection algorithms that power most surveillance cameras today can sift through vast amounts of footage, allowing law enforcement to pick out activity of interest, akin to pulling a needle out of a haystack.

Advertisement

Swearingen’s computer-generated patterns do not block surveillance cameras from recording video footage. Instead, they scramble the camera’s ability to identify objects, people, or faces, so that the cameras do not trigger any detection alerts. By blocking the camera’s ability to detect what the pattern covers, the person becomes a needle in a haystack again — until someone knows where to look. 

“Privacy is a fundamental right,” Swearingen told TechCrunch in a call this week. He described his patterns as a way to allow people to “opt-out of being tracked.”

In its first public test Friday at the Def Con cybersecurity conference in Las Vegas, Swearingen successfully demonstrated the pattern printed on a vehicle, proving that these patterns can be effective at defeating surveillance detection in the real world.

Teaching a model how to paint

In a call from his home in Kansas City, where he co-founded cybersecurity meet-up SecKC, Swearingen told TechCrunch that as a cyber professional he is acutely aware of the privacy and security risks of surveillance.

Advertisement

He described how his town is swamped with surveillance cameras, sometimes located just a few feet from each other. He said that he and others never opted in to being watched, just like he never opted-in to having the government use his driver’s license for facial recognition.

Swearingen described himself as a middle-aged white guy who lives in the center of the United States, and acknowledged that as a result he has not faced hardship or discrimination for being who he is or what he looks like. Swearingen recounted how last year he wanted to attend a protest, but felt uncomfortable and concerned that the vast number of cameras could track people who were exercising their constitutional rights to free expression.

If he felt this way, undoubtedly others would as well, including those who wanted to exercise their rights but may not feel safe or comfortable doing so themselves. Swearingen got to work.

a screenshot showing the counter-detection capabilities of Swearingen's patterns.
Image Credits:Bill Swearingen

For as long as there have been cameras capable of detecting things, there have been efforts to counter the technology. Several art projects and clothing brands have introduced apparel that aims to help people defeat facial recognition. Some eyeglass makers are jumping on the trend, albeit not with much efficacy. 

Swearingen said his research builds on some of this earlier work, which showed that it was possible to block camera detections. 

Advertisement

He started out last year with a proof-of-concept test lab that began by incrementally defeating one open-source video camera detection algorithm after another. Over the course of the year, he refined the patterns by scaling up his tests with additional computer processing power. He thanked the wider community who showed up with hardware to help further the project along. 

His proof-of-concept evolved over time into a reinforcement learning model, essentially a self-contained system that could train itself on which patterns work and which do not against the specific camera algorithms he is testing. In simple terms, Swearingen told TechCrunch that he essentially taught his model “how to paint.”

Each time a pattern failed and an algorithm detected it, the model would try again, over and over, until it eventually defeated multiple algorithms at once.

His model soon began to find perfect recipes for patterns that were able to defeat all of the 11 open-source detection algorithms he tested, including the software that powers Flock license plate readers, Axon body-worn cameras, and cameras running Clearview AI.

Advertisement

Now the model creates new patterns every minute, each batch mathematically better than the last, he said.

On Friday at the Def Con cybersecurity conference in Las Vegas, Swearingen ran his first real-world test. With help from Donut Media, the test involved covering a 2009 Toyota Yaris with one of Swearingen’s newest patterns to see if the car would be invisible to detection by a Flock camera.

“We proved it was effective;” said Swearingen; though, the wheels were a challenge, he said. The video of the demo will be out in the next few weeks, said Donut Media.

With a public demo in Las Vegas now under his belt, the project is early proof that it is possible to avoid algorithmic detection in public spaces. The next step is getting the patterns into the hands of people who want them, he said.

Advertisement

The noRecognition project also has a crowdsourcing campaign to help fund the sale of early merchandise featuring the patterns, from T-shirts to hoodies, with the potential for pattern-printed skins for vehicles down the line. Swearingen said the aim is for the patterns to be high quality and resolution good enough to work from a distance, while also looking aesthetically fashionable.

He said he is keeping his strongest patterns off the internet to prevent the camera makers from defeating them, but that the work is not yet done. His models are continuing to grind out new patterns.

“Every failure improves my model, and so [the patterns] keep getting better and better,” he said.

a photo of the toyota yaris covered in a pattern made by Bill Swearingen, as part of a test to see if it can defeat surveillance camera detection.
A photo of a 2009 Toyota Yaris at the Def Con conference in Las Vegas, covered in a pattern made by Bill Swearingen, as part of a test to see if it can defeat surveillance camera detection.Image Credits:Bill Swearingen / Donut Media (used with permission)

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

Source link

Advertisement
Continue Reading

Tech

Quake Celebrates 30th Anniversary: New Official Episode With New Maps and Mechanics

Published

on

To celebrate Quake’s 30th anniversary, Bethesda released “a brutal new episode” — a new campaign chapter titled Dawn of the Machine “delivering ferocious combat, labyrinthine level design, and nightmarish realms that twist reality beyond recognition…”

Face deadly new twists on familiar foes, including the Rocket Ogre, Demo Dog, Blood Shambler, and more. Each variant introduces lethal new behaviours — from overwhelming firepower to explosive death traps — forcing you to rethink every encounter. Expand your arsenal with brutal variants inspired by classic Quake expansions, including the Super Axe — unleashing lightning on successive strikes — and the Laser Cannon, firing ricocheting projectiles that tear through enemies from every angle… Reality is never stable. Shift between dimensions to solve puzzles and navigate the world, or find hidden secrets scattered throughout the realms. Face enemies that rise again or transform after death, and experience encounters where the rules change without warning — forcing you to adapt or be overrun.

More details from Ars Technica:

This is part of a series of new campaign chapters, all developed by MachineGames “in collaboration with id Software.” MachineGames is best known for making the games in the relatively recent Wolfenstein reboot series, as well as Indiana Jones and the Great Circle. The Quake maps are made by a team dubbed “Quake Club,” which includes several MachineGames employees who work on Quake maps in their spare time.

Advertisement

The ambition and scope of some of these maps is beyond what was typical for Quake maps back in the late ’90s, thanks in part to the team’s use of TrenchBroom, a newer map editor that works well on modern systems and is easier to work with and more capable than what folks were using in those days. There are 19 new levels, and they have several new gameplay gimmicks, like a time loop mechanic and a disorienting-but-cool teleportation system that makes it look like the levels are changing on the fly — kind of similar to the multi-time-period levels we saw in Titanfall 2 or Dishonored 2, though in these Quake maps the time shifts happen without player input. There’s also new music… The new maps have been added to the campaign list in the remastered version of Quake, and like that remaster, they’re available free to anyone who already owns Quake on Steam, as well as Quake on current and last-gen PlayStation, Xbox, and Nintendo consoles.

Read more of this story at Slashdot.

Source link

Advertisement
Continue Reading

Tech

iPhone Fold may ship in a choice of Silver or Dark Blue

Published

on

The iPhone Fold’s two colorways may have been revealed, with the foldable smartphone potentially shipping in dark blue and silver colorways.

The iPhone Fold is expected to arrive this fall alongside the iPhone 18 Pro. However, while the general design and specifications have pretty much been settled on, rumors over the summer couldn’t decide what colors it will ship with.

If images shared by Sonny Dickson are accurate, it will ship in at least two fairly pedestrian colorways. That would be Silver and Dark Blue.

The post to X on August 8 shows a pair of third-party accessories for the inbound model. The covers for the camera bump are claimed to be color-matched to the iPhone Fold.

Advertisement

This does depend on the manufacturer of the camera bump protector to be correct with their information. It’s possible that the two images are correct, but they can always be an educated guess by the accessory producer.

Sonny Dickson is a fairly reliable leaker, which means we can consider the accessories as being genuinely produced by someone. Whether that manufacturer is right is another matter entirely.

This is not the only color-related post Dickson has made for the inbound iPhones. In May, he posted photos of dummy units of the iPhone 18, including options in Silver, Black, Light Blue, and Dark Cherry.

Speculative colors

Back in February, one rumor about the specifications of the iPhone Fold said that two colors were on the way. It was claimed that one color was white and the other was unknown, but that the camera plateau would be all-black either way.

Advertisement

The same month, another report insisted that one of the two colors would be a black or dark gray, while the other was white or light silver.

By June, it was alleged that Apple still hadn’t nailed down the colors for the model, and that a black option probably wasn’t on the way.

Based on the iPhone Fold being a premium model that Apple’s supply chain is only getting to grips with producing for the first time, the limited color choice seems very plausible. That said, it doesn’t stop Apple from introducing other colors down the road.

Advertisement

Source link

Continue Reading

Trending

Copyright © 2025