More Information

As enterprise data volumes grow, replication pipelines built for smaller loads often stop scaling cleanly. Jobs that once finished overnight begin to run into business hours, freshness gaps widen, and compute costs climb. At that point the constraint shifts from the size of the data to the efficiency of the architecture that reads and writes it. Two techniques address this directly. Parallel partitioned reads divide a large source table into row-range partitions and read them at the same time across CPU threads, which reduces read time on large datasets. Write-path optimizations lower the cost of processing result metadata and writing files on the destination side. Wide tables with hundreds of columns benefit most, since per-column work repeats across every file operation. Both techniques build on cloud-native bulk loading, which stages data as optimized files and loads it through a warehouse’s native ingestion interface for higher throughput than row-by-row writes. This White Paper explains the benchmark methodology, reports measured results across common cloud destinations, and outlines a practical configuration for applying these techniques to large-scale replication.

Shortwave radios offer a way to connect one location on Earth to practically anywhere else with minimal infrastructure. But these radios come with some drawbacks—a significant one being that, unlike satellite communications, their transmission rates for digital data are typically measured in just hundreds of bits per second.

Peter Bloom is the founder of Rhizomatica, a Philadelphia-based nonprofit that has open-sourced a digital shortwave-radio set called the High-frequency Emergency and Rural Multimedia Exchange System, or HERMES. The set operates in the high-frequency (HF) band from 3 to 30 megahertz, as does Mercury, its digital modem. Rhizomatica staff travel around the globe to remote locations in countries like Bangladesh, Brazil, and Ecuador. Wherever they go, they use HERMES to help connect locals to the rest of the world.

Bloom spoke with IEEE Spectrum about how HERMES brings better data rates and encryption to shortwave radios.

How does HERMES connect remote locations?

Peter Bloom: We use the ionosphere as our satellite—or mirror—which helps us move information, voice, and data over really long distances. We’re using small radios that put out about 20 watts of power, and we can pretty reliably do 400- to 600-kilometer links between two radios. We’re talking about places that are not easy to reach, where it’s not simple to put terrestrial infrastructure.

What can HERMES send that a basic voice radio can’t?

Bloom: HERMES is a software stack—it’s a set of different programs that all work together in order to be able to send data over HF. HERMES allows you to send pretty much any file. Depending on what the file is, whether it’s a photo or an email or a voice memo, it just sends it as a file. It’s like a data pipeline over HF.

Why does sending files and data matter more than just voice?

Bloom: In emergency situations, people send their latitude and longitude over HF to say, “Hey, I’m here at this place.” People need to be able to send data over HF if there’s a manifest, a parts list, telemedicine—here’s what we have, here’s what we need. Instead of trying to read that out over the air, it’s much easier to just send the file. Same with a photo—if we need evidence that an area was logged illegally, we can just have someone send that over HF, rather than spending days getting down the river to get the photo where it needs to go.

Why did you build in encryption that amateur-radio regulations in many countries don’t allow?

Bloom: Encryption [regulations] for ham radio operators are different in each country. So it’s all optional—you turn it on, you turn it off. The reason we built the encryption is that some of the partners we work with are in very sensitive areas and don’t want to be sending out information that can be easily captured and used against them.

How has HERMES made an impact?

Bloom: We’ve been working with artisanal fishers in Bangladesh on a pilot project. There’s 10 or 11 boats that have HERMES systems on them. Pretty soon after we installed those, one of the boats had a mechanical issue in the Bay of Bengal, 100 or 200 kilometers offshore. They were able to send their GPS position and an SOS that they were having trouble. They were able to coordinate the rescue of the crew and the boat. So that was a really cool moment of HERMES in action that we’re super happy about.

This article appears in the October 2026 print issue as “Peter Bloom.”

Harvey’s gross margin fell from about 50% at the start of this year to minus 50% by June as customer usage of its AI agents spiked, and turned positive again only after it released a model of its own built on Moonshot’s Kimi K3. Abridge, Decagon, Ramp and Rogo are making similar moves, and Sequoia Capital and General Catalyst are funding the shift.

Harvey’s gross margin fell from about 50% at the start of this year to minus 50% by June, Bloomberg reported, citing a person familiar with the matter. The legal startup is valued at $15.6B. Customer usage had spiked after a March update to its AI agents.

The cost was the models it rented.

Harvey has seen a twentyfold rise in token usage this year, according to the company. Both OpenAI and Anthropic now charge enterprises for model usage on top of base subscription fees. Its founders knew the spike would not survive contact with those rates.

So it stopped renting.

In August Harvey released its first in-house model, post-trained on Kimi K3 from the Chinese lab Moonshot. Gross margins turned positive again after that launch and other changes to how it uses AI, people familiar with the work told Bloomberg. Harvey declined to comment on specific financials.

It is not alone.

Abridge is building a clinical model on Nvidia’s open weights, and Decagon now routes 80% of customer queries through models of its own. Ramp raised $750M in June and is weighing training for the first time. “It made absolutely no sense a year ago,” co-chief executive Karim Atiyeh said. “It’s starting to make a lot more sense now.”

Anthropic has noticed.

At an investor forum in its offices it showed a slide on startups building their own models, with a reminder that Harvey still needs Opus for its hardest tasks. Mistral’s Arthur Mensch spent July arguing that closed models give providers immense leverage, and named Anthropic cutting off Windsurf while building Claude Code.

September made his case for him.

OpenAI moved to end its Cursor contract after SpaceX bought the company, citing past terms of service breaches by Musk businesses. European buyers have switching rights under the Data Act, but those cover a customer leaving a supplier, not a supplier leaving a customer.

Europe has its own Harvey.

Legora, in Stockholm, reached $100M in revenue inside 18 months and now serves more than 1,200 firms, and it has never said whose models it runs on. Not everyone thinks the answer matters. “Having your own model or not is such the wrong question,” said Menlo Ventures partner Matt Kraning. “In most cases, it tends to be a lot of cosplay.“

Internet Explorer Matters: Internet Explorer is now an obsolete and retired browser, but IE Mode is still hanging around in the latest editions of Windows. Microsoft recently reminded everyone that the company plans to stop supporting this compatibility solution sooner rather than later. Migration efforts should begin soon, although some companies might need more time to replace the ancient technology.

Microsoft retired support for Internet Explorer 11 years ago, but the ancient browser is still lingering in modern Windows releases through the IE Mode feature. By the end of 2029, even IE Mode will finally be gone, along with IE’s ancient web engine technology. Enterprise organizations should prepare for the switch-off now by upgrading or replacing applications that still require IE to run on recent operating systems.

In a recent message pushed to the Microsoft 365 admin center (MC1473151), Redmond reminded organizations about the planned retirement of IE Mode. As the company first announced in 2021, Internet Explorer (IE) mode in Microsoft Edge is going to be supported “at least” until the end of 2029.

Microsoft will share a new end-of-support notice for IE Mode at least one year in advance. However, IT admins must properly plan the migration of their company’s legacy applets, apps, and applications that still require Internet Explorer/IE Mode. Some modernization efforts might take years to complete, which is why Microsoft is routinely sending out deprecation reminders such as MC1473151.

IE Mode is a special compatibility mode available in modern Edge releases. It is based on Google’s Chromium browser engine, and can display web pages using Internet Explorer 11’s Trident engine. Also known as MSHTML, the proprietary layout engine has gone through many changes since its introduction with Internet Explorer 4.0.

Internet Explorer 11 uses MSHTML 7.0, which supports WebGL applets and the now-retired SPDY web communication protocol. If an organization does not use any IE-based applets, no action is required after Microsoft publishes the MC1473151 alert.

IE Mode will likely continue working with few issues even after 2029. However, organizations are asked to properly assess their IT infrastructure to identify the business- and mission-critical elements that still require IE Mode. These applications and websites should be updated, rehosted, replaced, or rebuilt using modern web technologies, Microsoft warned.

The company is even offering to lend a hand through its App Assure service, which should help eligible customers check and improve compatibility of custom applications with modern Windows platforms.

Either way, Microsoft warns that “planning early can help your organization identify migration risks and reduce the likelihood of disruption as you move away from IE mode dependencies.”

OPPO has teased the K14 Plus 5G for an upcoming launch in India. The phone will be the latest addition to the company’s K14 series, following the K14 Lite, K14, and K14x. So far, OPPO has shared only a teaser image, with no word on the phone’s processor, display, cameras, or battery. The teaser does point to smooth performance and everyday use, though. Here’s what we know about the K14 Plus and what to expect from it.

What the Teaser Reveals

OPPO has shown a teaser of its K14 Plus 5G handset for the Indian market. So far, OPPO has released only side-view images of the new phone. These include images of the power and volume buttons on the side. The teaser also hints at the colors the phone will be available in: orange and dark grey or black. OPPO has said the new phone will focus on performance and usability. However, the teaser does not reveal the hardware that will power the phone. The company has not confirmed its processor, display, or camera setup.

OPPO has not shared details on the processor, screen, cameras, RAM, storage, or battery. The “Plus” in the phone’s name might mean a bigger screen or battery, but it is still unclear.

OPPO K14 Plus 5G Price and Position in the Lineup

It is also unclear where the OPPO K14 Plus sits in the K14 series. As of now, OPPO K14 Lite and K14x are positioned in the below-Rs 20,000 range. Meanwhile, the base variant of the OPPO K14 starts at Rs 25,999 in India. Based on this, the K14 Plus may cost more. It may land in the Rs 30,000 range. This is only an estimate, not the final price. The K14 Plus could help OPPO expand its options in the budget-to-mid-range segment. Its final position will depend on the specifications and features OPPO announces. Buyers should wait for the official price and full specifications before choosing the K14 Plus.

“I was hearing from people in the campaign world that there’s this new strategy of paying political influencers to post content,” said Marc Berman, the California assemblyman. “Some were good about disclosure and others were a bit fast and loose.”

His bill is now law. Gavin Newsom signed AB 1130 on Saturday. In California, paid political posts that hide the payment now carry a $5,000 fine each.

The state has required those disclosures since 2023. Until now it could not punish anyone for skipping them. AB 1130 changes that. It hands the Fair Political Practices Commission the power to impose administrative, civil or criminal penalties. The commission can also refer a violation to law enforcement as a potential misdemeanour.

Ken Bensinger reported the law for the New York Times. He wrote on X that a wilful violation could carry up to a year in prison.

What the law actually covers

The scope is narrower than “influencers must disclose”. The Governor’s office describes AB 1130 as covering a person paid by a campaign committee to post online political content. The content must support or oppose a candidate or a measure. The penalty applies when that person leaves out the required disclaimer saying they were paid. Payment from a campaign committee is the trigger. An unpaid creator posting their own opinions sits outside it.

That distinction matters because of how the money now moves. Politicians and advocacy groups are paying creators to reach younger audiences, the Times reported. No federal rule requires anyone to disclose those payments.

The campaign that prompted it

Berman introduced the bill after the Times reported in May on money reaching influencers during California’s contested primary for governor. Politicians were funnelling it quietly, the paper found.

Tom Steyer, the billionaire Democrat, paid dozens of creators to post favourably about his campaign, Bensinger reported. Those posts ran on TikTok, YouTube, Instagram and Reddit. Many carried no written disclosure at first.

Multiple complaints followed to the Fair Political Practices Commission against the Steyer campaign. Somebody also filed at least one against Xavier Becerra, his main rival, who went on to win the primary.

The commission opened an investigation into Steyer’s campaign in May and closed it this week. Steyer had properly notified the influencers of their obligations, it found, and “then reminded the influencers when posts were discovered without the disclaimer”. A second investigation into the campaign is still open.

Becerra faces Steve Hilton, a Republican, in November.

One of thirteen

AB 1130 arrived inside a package of 13 bills on elections and social media. Newsom signed them at the Democracy Center in Los Angeles, and framed the package as a defence against federal interference.

“Donald Trump won’t stop until he can exert dictatorial control over your free vote and disenfranchise millions of people this November,” he said in a statement. “California will stop him at every opportunity.”

Two other bills in the package matter for anyone building or hosting synthetic media.

AB 686, also Berman’s, extends the sunset on California’s deepfake rule. The old date was 1 January 2027, and the new one is 1 January 2031. That rule bars distributing materially deceptive audio or visual media with actual malice, where the intent is to injure a candidate’s reputation or deceive a voter. Media carrying a disclosure that it has been manipulated is exempt. The rule was due to lapse before the 2028 presidential cycle. It now runs past it.

AB 502 amends the rules on knowingly distributing election communications that contain materially deceptive content.

A third, AB 2281, requires the Secretary of State to assess whether California needs to replace election cybersecurity resources the federal government previously supplied.

The rest of the package is ballot security rather than platform policy. Several bills create felonies for seizing ballots, election records or certified voting technology before results are certified.

California keeps legislating at the platform layer

The state has built a stack of these rules over several years, and the enforcement body is usually the same one.

In 2024 Newsom signed AB 2655. It requires large online platforms to remove or label deceptive digitally altered election content during set periods, and to provide reporting mechanisms. He signed AB 2355 the same year. That one requires electoral advertisements built with AI-generated or substantially altered content to carry a disclosure, and the Fair Political Practices Commission enforces it.

Eleven days before this package, he signed a separate set of child safety bills covering algorithmic feeds for under-16s.

The disclosure question is not confined to campaigns either. A watchdog reported last month that Meta had used paid influencers to push back on teen social media bans. YouTube has separately been carrying AI slop channels that produce political misinformation.

Who else is doing this

Texas already mandates disclosure of paid political content online. Lawmakers in Congress and in other states, including New York, are considering similar rules, according to the Times.

Elsewhere the pattern runs further along. Brazil’s electoral court set rules on AI in elections last year. Meta’s own Oversight Board has been ruling on election deepfakes in Europe.

California’s version lands six weeks before Americans vote. The Fair Political Practices Commission now has a penalty it can apply, and the primary that produced the complaints has already been decided.

Meta’s latest AI app, Muse, could be the social giant’s next major hit, new data suggests. According to just- released estimates from market intelligence provider Apptopia, Muse’s app on mobile devices has been downloaded more times during its first 12 days on the market than ChatGPT was in the 12 days after its debut. (This data compares the U.S. and Canada markets only.)

Muse also has more daily active users than ChatGPT had at the time, the firm found.

Until now, it had been difficult to make an exact apples-to-apples comparison between these two apps, given that they had pursued different launch strategies. When ChatGPT arrived on mobile, it was made available globally, but only on iOS. Muse, meanwhile, is available across both Apple’s App Store and Google Play, but only in the U.S. and Canada.

To better align the numbers for comparison, Apptopia looked only at the iOS data for the U.S. and Canada for both apps during the first 12 days of their respective launches. In this subset of the data, Muse has now seen 1.8 million downloads to ChatGPT’s 1.3 million.

Overall, Muse has seen 2.8 million total installs globally in its first 12 days, the firm also said. Its growth hasn’t yet stagnated, either; Muse has moved up from its original position as No. 2 overall on the U.S. App Store immediately after its launch to now No. 1, as Business Insider reported on Friday. That jump put the app higher than ChatGPT, the outlet noted. Another firm, Appfigures, said at the time that Muse had then crossed 1 million downloads.

In addition, Apptopia’s data indicates that Muse’s U.S. daily active users are now higher than they were for ChatGPT at the same point after its launch. When comparing just the U.S. mobile app daily active users, Muse comes in higher with 642,000 daily users compared with 231,000 for ChatGPT at the time.

To be fair to the fact that Muse is available across both iOS and Android, while ChatGPT launched on iOS only, Apptopia narrowed the comparison to iOS alone. Yet, even here, Muse is coming in higher, with 359,000 daily active users on iOS, still above ChatGPT’s figures from that time period.

Apptopia can only provide third-party estimates about an app’s downloads and active users; it doesn’t have direct access to Meta’s internal figures. But even if these numbers are only correct from a general “ballpark” perspective, they indicate that Muse could have a shot at becoming Meta’s newest top app.

Meta already perfected its cross-promotion strategy when it launched Instagram’s Threads, which now has 500+ million users, thanks to heavy marketing and integrations in Meta’s largest apps, including Instagram and Facebook. Muse is likely to get a similar push, given that the AI app can connect with both of those platforms. It also works inside WhatsApp, so that could give it another boost.

While Apptopia doesn’t have visibility into Meta’s cross-promotion efforts or ads directly, it did note that over 95% of Muse’s users are also Facebook users and 63% are Instagram users.

Meta, which has been asked for comment, has not yet shared public figures related to Muse’s early adoption.

Correction: An earlier version of this article referenced the analytics firm Appfigures as a source for data from Apptopia. This has been fixed.

Nvidia is buying a further $1.5B of SB Energy shares at 90% of the price of the data centre developer’s coming US listing, taking its total backing to $3B. SB Energy has 8.8 gigawatts in contract or under construction in Texas and Ohio, while its parent’s European capacity is not due before 2031.

Nvidia is buying a further $1.5B of shares in SB Energy at 90% of the price of the SoftBank-owned developer’s coming US listing, Bloomberg reported, citing a regulatory filing on Monday. That takes its total backing to $3B.

The discount is the interesting part.

The chipmaker is taking new N class nonvoting shares in a private placement, which buys the economics and not the votes. It is also the supplier whose processors create the demand SB Energy is listing on. Nvidia bought into three land and power companies in August alone, as its liquid-cooled Rubin chips push rack density from roughly 250kW to 600kW.

SB Energy has the sites.

It holds 8.8 gigawatts of data centre capacity in contract or under construction, including projects in Texas and Ohio. OpenAI, which is planning a listing of its own, is also an investor. SB Energy is targeting $5B to $7B of proceeds.

It paid to get that tenant.

SB Energy gave OpenAI warrants worth $5.5B to anchor a 20-year lease on a 10 gigawatt campus in southern Ohio, due to run in 2028. The warrants were issued in January at $3.6B and had gained about $1.9B by June.

Europe gets the other half of the same group.

SoftBank has promised up to EUR 75B for 5 gigawatts of AI data centres in France, starting in Hauts-de-France at Dunkirk, Bosquel and Bouchain. The first phase is 3.1 gigawatts by 2031, and it sits in a separate vehicle from the one listing in New York. Schneider Electric is the strategic partner at Dunkirk, on a grid that is roughly 70% nuclear.

That is three years behind Ohio.

No European power developer is being financed this way. Nvidia is not buying discounted pre-listing stock in one, and no campus on this continent has a tenant paying its rent in warrants.

These economics have already failed here once.

OpenAI paused Stargate UK in April, a project it had announced with Nvidia and Nscale for northeast England, citing British industrial electricity prices more than four times those in the United States and uncertainty over copyright law. It said it would proceed when regulation and the cost of energy allow long-term infrastructure investment. Neither has changed since.

In a nutshell: AMD has become only the fourth US chipmaker to reach a $1 trillion valuation, following in the footsteps of Broadcom, Micron, and Nvidia. AMD’s stock entered 2026 at a price of $223.47 and would dip below the $200 mark a few times before climbing in April. As of writing, shares are trading at $607.24 – a healthy return for anyone that bought in early this year.

Nvidia eclipsed the $1 trillion valuation mark in May 2023, and has not looked back since. The company has continued to tack on value driven largely by the AI boom. Currently, the chipmaker has an incredible market cap of roughly $5.48 trillion, making it the most valuable company in the world by a sizable margin.

Apple, for comparison, is worth around $4.94 trillion. Google ranks third with a market cap of $4.29 trillion. Microsoft, another major player in the tech space, is worth $3.69 trillion as of today.

Credit: App Economy Insights

AMD is not the only chipmaker to start the week strong. Shares in Intel are trading up more than 12% on the day, and Qualcomm’s stock is up nearly 8.3%. Broadcom is having a slower day, up just 1.22% as of this writing.

Intel, often considered AMD’s primary rival due to their long history in the CPU industry, is also having a stellar 2026. Share values are up 208% year to date, but the company’s market cap sits at just $643.5 billion – far behind AMD’s.

Last month, AMD CEO Lisa Su said the company expects to double data center sales in 2027. That was welcome news considering the stock’s dip in July and August.

Like others, AMD is benefiting from the ongoing AI boom. At some point, however, the demand for AI hardware will cool. Fortunately for AMD, the company has other divisions to lean on when the time comes. For others that are putting all of their eggs in the AI basket, the outcome could be far more catastrophic.