Google Research and NASA JPL have released a deep learning model that maps methane plumes globally at 60 metres, finding about 50% more than human analysts and 23,000 additional plumes, including at 24 of the 25 largest-emitting landfills. Europe’s Methane Regulation ordered a satellite monitoring tool and a super-emitter alert system, but it covers only oil, gas and coal.

Google Research and NASA’s Jet Propulsion Laboratory have released a deep learning model that maps methane plumes worldwide from space, Google Research said. It flags around 50% more plumes than human analysts, and 23,000 more than the existing product.

The model is a vision transformer trained on 3.6 million simulated plumes. It reads raw spectra from EMIT, a NASA instrument on the International Space Station, at 60 metres per pixel.

Europe already runs a version of this. The Copernicus atmosphere service switched on a methane hotspot explorer in February last year, built by Dutch researchers on Sentinel-5P data.

The two are not doing the same job. Sentinel-5P covers the planet daily at roughly 5.5 by 3.5 kilometres per pixel, a region rather than a site.

Sixty metres resolves a landfill cell. The paper reports plumes at 24 of the world’s 25 largest-emitting landfills, against 17 for the NASA product it replaces. UCLA built last year’s list of the worst sites from Carbon Mapper analysis of EMIT and Planet Labs data.

Which is where Europe’s own rules stop short.

The Methane Regulation applies to oil, gas and coal. It obliges the Commission to build a global monitoring tool on satellite data, and Article 31 alerts member states to super-emitting events.

Waste is outside it. Prognos and the ifeu Institute estimated in January that municipal waste landfilled in Europe between 2022 and 2050 will release roughly 1.5 billion tonnes of carbon dioxide equivalent as methane.

Europe’s high-resolution answer is still being built. AIRMO raised EUR 5M in March to launch its first satellite in early 2027, and the Copernicus CO2M mission is due in November of that year.

What is available now costs nothing. The plume database sits on Earth Engine under a Creative Commons licence, the model is on Kaggle and the code on GitHub.

The limit is where the instrument looks. EMIT was built to map mineral dust in arid regions, and it flies on a space station orbit inclined at 51.6 degrees. Europe was never the target.

So the sharpest methane map in existence is a by-product of an American dust mission, given away.

European Earth observation has the opposite trouble, as TNW has reported: plenty of data and too few buyers. A regulator wanting facility-level methane attribution tomorrow would download Google’s file, then find the emitters it resolves best are the ones the regulation does not reach.

The company opened a Dublin office earlier this year with plans to hire 40 people.

Legal AI start-up Harvey has raised $550m at a $15.5bn valuation in a round co-led by Diffusion and Lightspeed Venture Partners.

The round marks an almost $5bn valuation growth for the start-up, coming less than six months after it announced a $200m raise at an $11bn valuation.

Other investors, including Sequoia, Kleiner Perkins, a16z, Coatue, Conviction, Elad Gil, Evantic, GIC, Goldman Sachs Alternatives, Verified Capital, Sapphire Ventures, Whale Rock and WNDR, participated in the latest round.

Alongside the raise, Harvey also announced its acquisition of San Francisco-based Guardrails AI, a security platform for AI agents. Guardrails’ co-founders Shreya Rajpal and Zayd Simjee and their team will join Harvey’s product and engineering organisation, the legal-tech company said in an announcement today (9 September).

“The opportunity for companies to accelerate their competitive advantage with AI has never been higher,” Harvey co-founders Winston Weinberg and Gabe Pereyra said in a statement.

“Harvey wants to be the global partner legal teams turn to for this work, and we plan to hire and develop the best team in the space to support our clients through the next wave of AI transformation.”

The San Francisco-based Harvey provides AI for legal and professional services. Similar to many domain-specific large language models, its platform leverages data tuned using legal, tax and business knowledge, utilising a multi-model approach powered by leading AI models.

The company’s platform is used by a majority of the world’s largest law firms and by in-house legal teams at five of the Fortune 10, it said.

The latest funding round comes just after Harvey unveiled its first post-trained open-weight model, Harvey Tenet. The model has a base built from Chinese AI start-up Moonshot’s Kimi K3 and is trained specifically for legal tasks.

Earlier this year, the company opened an office in Dublin with plans to hire more than 40 employees over the next two years. The office was initially announced in January with early hiring plans eyeing 20 roles in the first year.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Apple surely took its time to launch its first foldable iPhone. However, the device it has revealed comes with an IP68 rating for dust and water resistance. For a folding phone, the 6 in that rating is the part worth paying attention to.

IP ratings use two numbers. The first measures protection against solid particles such as dust, while the second covers water resistance. A rating of IP68 means the iPhone Duo gets the highest level of dust protection under the standard and strong protection against water. Apple says the phone was tested at depths of up to 6 meters for 30 minutes.

Dust protection is where the iPhone Duo stands out

Water resistance is already common on expensive foldables. Dust is a much tougher problem because hinges, moving parts, and tiny gaps give sand, lint, and other debris more places to get inside.

Samsung’s Galaxy Z Fold 8 carries an IP48 rating. Its 8 still represents strong water resistance, but the 4 does not certify the phone as dust-tight. The iPhone Duo’s 6 does. That gives Apple’s first foldable the highest dust-protection level under the IP standard.

Google has already reached the same level

Apple is not the first foldable maker to get there. Google’s Pixel 11 Pro Fold also carries an IP68 rating, so it matches the iPhone Duo for certified dust and water protection.

Still, IP68 is reassurance, not a free pass to abuse the hardware. Apple notes that dust and water resistance can decrease through normal wear. In other words, your $1,999 foldable may be dust-tight, but giving it a dust bath is still a terrible idea.

Apple is trying to get consumers to accept, and even welcome, a new reality: Your tech is always listening.

At Wednesday’s “Surprise and Shine” iPhone event, where the company introduced its new hardware lineup, the real surprise turned out not to be the foldable iPhone, but how the famously privacy-sensitive tech giant has decided its watches will now be able to process live audio and transcribe what was said.

Does this mean you’re always being recorded if the person you’re interacting with is wearing an Apple Watch? Not exactly — though it may make some people uncomfortable regardless.

It also raises the question of whether consumers actually want this technology, or if Apple forged ahead in this area because it was afraid that another company, such as OpenAI, would soon introduce a competitive AI device of its own.

Today, a number of startups are acting on the belief that AI-powered transcription is a key consumer use case for AI. So far, this more often involves transcribing meeting notes or recording lectures or interviews. Some companies, like the AI pendant Friend and Amazon’s Bee, ventured into other areas, too, with always-available AI companions that can help you remember your day by passively taking notes.

Apple decided to go a slightly different route. With its new watches’ “audio intelligence,” “Live Rewind,” and “Siri Recap” features, the company is straddling the line between accessibility and utility on one side, and what starts to feel like invasive technology on the other.

On its own, Audio Intelligence seems to be a useful feature for the deaf or hard of hearing, as it can alert Apple Watch owners to sounds like sirens, alarms, doorbells, babies crying, and more, using on-device AI. This can work even when the wearer’s iPhone is not with them and, in some cases, could provide life-saving assistance, such as when a fire alarm or carbon monoxide detector goes off. It’s doubtful anyone would take issue with this type of audio AI.

However, Live Rewind and Siri Recap move out of the assistive tech space and into a murkier area.

While Apple’s data protections may make these features more palatable to some, there’s still the possibility that normalizing listening technology may not be the best move for a privacy-conscious company. This is especially true given the broader consumer backlash against surveillance technology like Flock cameras, AI data centers, and AI apps in general.

For starters, the Live Rewind AI feature lets you go back in time 15 seconds and transcribe what was just said by double-pressing the watch’s digital crown. This transcription can then be saved to Apple’s new standalone Siri app so you can revisit the information later.

Apple suggests that this could be used to remember things like a “book recommendation, a colleague’s brilliant idea, or something you didn’t hear the first time,” but it also means that you can essentially catch anyone saying something they didn’t necessarily mean to have recorded and make a note of it.

This has implications around consent, not only from an ethical standpoint but from a legal one as well. AI device-maker Plaud, for instance, advises customers to obtain legally required consent before recording. Amazon’s Bee‘s terms of service remind users that they’re solely responsible for complying with the law around minors’ data privacy, not Amazon.

Since Apple isn’t saving audio, it’s unclear how the court system might handle these new, text-only transcripts. It’s possible the notes could be introduced as evidence in court, but their lack of audio could make them harder to authenticate or verify. Plus, the evidentiary rules around their use will also likely vary by state.

Beyond the legal issues, the idea that someone can always record from their watch has cultural implications.

Already, recording devices have impacted the way people behave in real life, studies have shown, as has social media and data surveillance culture. Knowing you could be recorded at any time has a dampening effect on how you interact with others and engage in self-expression.

However, recording a photo or video still requires a fairly obvious move on the recorder’s part: you have to hold up your phone and face it toward the action. Snapping up a bit of audio to turn into a transcript just by pressing the watch’s digital crown is arguably a much subtler gesture, even though it comes with an audible chime and full-screen animation of a microphone, as Apple says.

Another new Apple Watch feature, “Siri Recap,” uses ambient listening to “take high-level notes” of your conversations, which can be reviewed in the Siri iOS app. These notes will automatically generate a title, summary, and key points based on what was recorded and summarized by AI. They will not be exact transcripts.

Apple suggests this could make sense for recording meeting takeaways or key info from a parent-teacher conference, for instance.

The company also tried to assure users that it has approached these listening features with respect for user privacy.

Apple notes that Siri Recap doesn’t default to always being on like some other AI wearables are. Instead, you can choose when it’s enabled, like only at work or only during the day. Plus, you can turn the feature off and on at any time from Control Center.

Notably, Apple points out that it does not create or store audio recordings, and that the raw audio isn’t accessible, even to Apple. It also doesn’t identify or attribute the speakers, and it protects the generated transcripts and recaps with end-to-end encryption.

Whether that’s enough for consumers not to feel like the feature is creepy is another matter. Either Apple has found the right balance between utility and privacy, or it’s about to see some consumer backlash.

RAMageddon 2026 spares no one.

Healthcare technology company Veradigm disclosed a data breach after a cybersecurity incident at one of its third-party vendors exposed patients’ personal data.

The company says the incident did not cause operational disruptions but affected a small number of customers.

Formerly known as Allscripts Healthcare Solutions, Veradigm is a Chicago-based healthcare technology company that supplies medical practices with electronic health records, e-prescribing, patient-engagement, practice-management, and revenue-cycle software.

Thousands of hospitals, clinics, and biopharmaceutical firms across the United States use its solutions.

The company says in a filing with the U.S. Securities and Exchange Commission (SEC) that an attacker obtained credentials from a vendor’s environment for a Veradigm API reserved for customer services. The threat actor then used their access to copy patient data.

Veradigm’s disclosure notes that the stolen data includes personal details and Social Security numbers (SSNs) for some of the patients. Clinical or medical information remained safe.

“The vendor’s compromised credentials provided access only through that limited interface and did not provide access to any other part of the Company’s environment, including the Company’s broader network, servers, databases, or other systems,” the company says in the SEC filing.

After discovering the breach, Veradigm initiated its incident-response procedures, notified law enforcement, and is currently investigating to determine the scope.

Affected customers and individuals are being notified, with credit-monitoring services offered where applicable.

The investigation is ongoing, but based on current information, Veradigm does not believe the incident is reasonably likely to materially affect its business, operations, financial condition, or results.

The Gentlemen ransomware claims the attack

Although Veradigm’s disclosure did not identify the attacker, The Gentlemen ransomware group has claimed the intrusion on September 5 and listed the company on its data leak site.

The threat actor alleges to be holding 3.5 million patient records that include full names, home addresses, SSNs, email addresses, phone numbers, and personally identifiable information or guarantors.

The ransomware actor threatens to leak the stolen data by Friday, September 11, if the company doesn’t engage in a ransom payment negotiation.

The Gentlemen threat actor emerged around mid-2025 and operates as a double-extortion group, combining data theft with data encryption on Windows, Linux, NAS, BSD, and ESXi systems.

On its data leak site, the gang listed more than 800 victims from 86 countries and various sectors, including manufacturing, technology, healthcare, transportation, and financial services, indicating opportunistic attacks that rely only on access availability.

In April 2026, Check Point reported that it discovered a SystemBC proxy malware botnet with more than 1,500 hosts and linked it to an affiliate of The Gentlemen ransomware gang.

In June 2026, ESET said that The Gentlemen was employing a new endpoint detection and response (EDR) killer called GentleKiller.

Overall prevention scores can hide what happens after initial access. Once attackers are using valid credentials, prevention drops sharply.

The Blue Report 2026 measures defenses technique by technique across 338 million simulations run in customer production environments.

Get the report

Class D amplification has spent the better part of two decades trying to convince audiophiles that small, efficient and cool-running are not synonyms for cheap, sterile and disposable.

We have covered that evolution rather extensively at eCoustics, including the rise of Hypex, Purifi and GaN-based designs that have made some of the old complaints about Class D sound increasingly like arguments over whether CDs can ever replace cassettes.

Orchard Audio has been part of that conversation for years with its Starkrimson amplifiers. Now the New Jersey manufacturer is moving the technology into home theater with the Starkrimson HT Ultra Series, available in three, four and five-channel configurations.

The last part is relevant because this is not our first encounter with Orchard Audio in a theater system. At Capital Audiofest 2025, we heard an SVS 5.2.4 system powered by Orchard amplification and came away impressed by its dynamics, control and value. The HT Ultra simply turns that concept into a purpose-built product rather than an audiophile science project involving a small mountain of separate amplifiers.

Related: Fear And Loathing In Class D Amplifiers: What $130 To $50K Actually Buys

What Exactly Is GaN?

GaN stands for gallium nitride, a wide-bandgap semiconductor material used here in place of the silicon MOSFETs found in most Class D amplifiers.

GaN transistors can switch faster with lower switching losses than conventional silicon devices. That does not automatically make every GaN amplifier sound better; circuit topology, feedback, power supply, PCB layout and output filtering still matter enormously, but it gives designers considerably more room to work at higher switching frequencies while controlling heat and efficiency.

Orchard’s Starkrimson Ultra circuitry switches at approximately 800kHz, which the company says is three to four times faster than traditional silicon-based Class D designs. Its modulation is performed entirely in the analog domain, using Orchard founder Leo Ayzenshtat’s proprietary DC-coupled, fully balanced dual-feedback modulator and bridged GaN output stages.

And no, the “D” in Class D still does not mean digital. We have been over this.

The Power Supplies Are the More Interesting Part

The HT Ultra uses Orchard’s Starkrimson Ultra 500W amplifier modules and 1200W power supplies, but buyers can determine how those resources are distributed.

The five-channel model can be configured with one, three or five 1200W supplies. The four-channel version offers one, three or four, while the three-channel model uses one shared supply. Higher-performance configurations add 43,600µF capacitor boards to channels that receive dedicated supplies.

That allows someone building a hybrid music/home-theater system, for example, to give the front left and right channels dedicated power supplies and capacitor reserves while the center and surrounds share another supply.

That is considerably more interesting than simply putting five identical amplifier modules in a box and hoping nobody asks what happens when all of them get hungry at once.

Starkrimson HT Ultra Specifications:

• Channels: 3, 4 or 5
• Output: 250W into 8 ohms / 500W into 4 ohms*
• Gain: 19dB
• Amplifier topology: GaN Class D
• Switching frequency: approximately 800kHz
• Signal path: fully balanced
• Power supplies: 1200W each, configuration dependent
• Capacitor boards: 43,600µF each on selected configurations
• Load capability: 2-ohm capable; Orchard specifies stability down to 1 ohm
• Inputs: Balanced XLR
• Power-on: Front-panel control or trigger input

*The 250W/500W rating applies when a channel has its own power supply or when two channels share one. That qualification matters rather more than the number printed in 72-point type on an amplifier brochure.

One installation detail also deserves attention. Orchard specifies that the five-channel/five-power-supply configuration requires a 20A circuit in 110/120V countries. Your electrician may therefore become part of the upgrade path.

Price & Availability

The Orchard Audio Starkrimson HT Ultra Series is available for preorder now, with current discounted pricing ranging from $3,199.95 to $6,399.95, depending on configuration. Orchard says the preorder pricing represents 20% off MSRP, with initial shipments expected in December 2026 or January 2027.

The Bottom Line

GaN itself is no longer enough to make an amplifier unusual. What makes the Starkrimson HT Ultra interesting is Orchard Audio taking the same amplifier architecture it developed for serious two-channel systems and allowing home-theater buyers to decide where the expensive power-supply hardware actually goes.

A five-channel amplifier with five independent 1200W supplies is very different from one asking all five channels to drink from the same trough.

Class D already won the argument over whether it belongs in high-end audio. Orchard Audio now wants to know whether your home theater has been paying attention.

For more information: orchardaudio.com

Related Reading:

from the phony-business dept

In addition to the $111 billion Paramount/Warner Bros merger just being foundationally bad for creatives, labor, and consumers, the Ellison family-owned company’s executives just keep demonstrating why they’re not really the kind of folks you want dominating the lion’s share of U.S. media.

That’s included calling deal critics “antisemitic,” cozying up to foreign authoritarians, constantly threatening to take Paramount out of California, or, more recently, using a fake consumer group (dubbed the “Neighbors for Strong Communities“) to try and create bogus public support for what’s a pretty broadly unpopular deal.

Neighbors for Strong Communities doesn’t disclose its funding donors and insists it’s not being “directly funded” by Paramount, but it has all the telltale signs of “astroturf,” or fake public interest advocacy. One of the things the group does is generate comments to California AG Rob Bonta urging him to drop the 12-state antitrust lawsuit against the merger. You know… for the children.

Rachel Antell at The Contrarian took a closer look at the Neighbors for Strong Communities website, and tried to fill out a letter to Rob Bonta opposing the merger, only to see something curious happen: The website’s AI-assisted letter system simply changed her writing to not only express support for the merger, but make up an entire fake backstory about her. This is not what she told it:

Antell notes the astroturf group’s letters leverage Influent, an AI vendor founded by a guy named Tanner Kelly, who registered as a California lobbyist last year. Antell herself is an archival producer who founded the Archival Producers Alliance and has crafted guidelines for the use of generative AI aimed at more transparently disclosing when it’s being used.

U.S. corporations have a very long history of flooding regulatory comment proceedings (or the press) with completely artificial support for what’s almost always bad policy few people actually support. You may recall how the telecom industry was caught using fake and dead people to support the Trump administration’s destruction of net neutrality rules and other consumer protections.

This stuff was bad before AI, now everything’s been taken to the next level. There’s no meaningful effort afoot to stop or identify it anywhere in the U.S. government. It exists as an effective way to drum out genuine public interest opinion and replace it with lazy cack that supports whatever shitty policy companies drum up (gaining merger approvals, gutting oversight).

In this case, that means creating fake workers who are super excited about a merger that’s going to result in thousands of layoffs as the bigger company’s executives offload their massive new debt obligations to labor and consumers. How do we know this? It’s what happens literally every time regulators rubber stamp more media consolidation — particularly when Warner Brothers is involved.

Filed Under: ai, antitrust, astroturf, consumers, david ellison, lawsuit, media consolidation, mergers

Companies: neighbors for strong communities, paramount, warner bros.