Japan’s SoftBank will be the first to deploy the new SN50 chips, while Intel is partnering with SambaNova to roll out its Intel-powered AI cloud.

Intel-backed Nvidia rival SambaNova has raised $350m in a Series E round led by Vista Equity Partners and Cambium Capital, with strong participation from Intel Capital.

Reuters was first to report the planned raise earlier this month. While details of the investment were not disclosed at the time, sources had told the publication that Intel Capital’s contribution would be around $100m, with potential commitments of up to $150m.

Other participants in the round included Gulf Development, Assam Ventures, Battery Ventures, Atlantic Bridge, GV and BlackRock.

According to SambaNova, proceeds from the raise will be used to expand the production of the company’s newly introduced SN50 chip – touted to deliver “the best tokens per watt” – as well as to scale ‘SambaCloud’ and deepen enterprise software integrations.

SoftBank will be the first to deploy SN50 within its AI data centres in Japan, powering inference services for sovereign and enterprise customers across the Asia-Pacific.

Intel has close ties with the 2017-founded SambaNova, with CEO Lip-Bu Tan serving as chairperson on SambaNova’s board. Alongside the raise, the two companies have also jointly announced a multi-year collaboration to deliver cost-efficient AI inference solutions for AI companies, model providers, enterprises and governments worldwide.

As part of the collaboration, Intel is making a strategic investment in SambaNova to accelerate the rollout of an Intel-powered AI cloud.

“AI is no longer a contest to build the biggest model,” said Rodrigo Liang, co‑founder and CEO of SambaNova.

“With the SN50 and our deep collaboration with Intel, the real race is about who can light up entire data centres with AI agents that answer instantly, never stall, and do it at a cost that turns AI from an experiment into the most profitable engine in the cloud.”

The company positions itself as a rising competitor looking to take some of Nvidia’s gigantic share in the AI chips market. Liang, who previously worked as an executive at cloud provider Oracle, said in 2024 that Nvidia had “lost some of its sheen” and that “rivals are biting at its heels”.

Kevork Kechichian, the executive vice-president and general manager for Intel’s data centre group, said: “Customers are asking for more choice and more efficient ways to scale AI.

“By combining Intel’s leadership in compute, networking and memory with SambaNova’s full-stack AI systems and inference cloud platform, we are delivering a compelling option for organisations looking for GPU alternatives to deploy advanced AI at scale.”

Other companies are also looking for alternatives to Nvidia. Meta yesterday (24 February) said that it would buy billions of dollars’ worth of AMD’s chips to develop AI tech and power new data centres. The deal could see Meta taking a stake of up to 10pc in AMD.

Earlier this month, Cerebras Systems, which also positions itself as a rival to Nvidia, raised $1bn in a Series H round led by Tiger Global with participation from AMD. In January, Cerebras and its early backer OpenAI announced a partnership to deploy 750MW of Cerebras’s wafer-scale systems to make OpenAI’s chatbots faster.

Positron, another Nvidia competitor that offers energy-efficient AI chips for inference, raised $230m from Arm Holdings and the Qatar Investment Authority in recent weeks, taking its valuation above $1bn.

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.

Wearables startup CUDIS is launching its newest series of health rings this week. The updated ring comes equipped with a number of features, including an AI “agent coach” designed to keep users on track to attain their fitness goals.

CUDIS says it differentiates itself from other wearables by not just delivering health metrics but also incentivizing healthy behavior through a points system. Users garner digital “health points” for healthy behaviors — things like daily sleep, 10,000 steps every day, sports activities, and conversations with the ring’s AI coach — which can then be redeemed through an integrated marketplace for discounts on health supplements and other products.

The ring’s AI Agent Coach, meanwhile, is designed to leverage generative AI to aid with healthy programs for exercise and daily health. The company says that its agent generates tailored programs including “daily tasks, recovery protocols, supplement recommendations, and direct referrals to licensed medical professionals.”

The ring also tracks a host of body metrics and daily behaviors, such as sleep quality, stress management, movement, and recovery. This helps them see how these metrics affect their Pace of Aging (PoA), showing whether their body is aging faster or slower than their chronological age, the company explains.

CUDIS CEO and co-founder Edison Chen told TechCrunch that since his company’s first wearable was launched in 2024, the company has sold over 30,000 units across its first two models. The app’s user base has also grown to 250,000 users across 103 countries, he added.

“Our strongest markets so far have been North America, Europe, and Asia,” Chen said. “What we’re good at is pattern recognition for healthy people trying to optimize,” Chen told TechCrunch.

“The AI spots when you’re trending in the wrong direction, such as chronic poor sleep, declining HRV, elevated resting heart rate, and either suggests lifestyle changes or connects you to a professional. The control is in the escalation pathway to the right care access,” he said.

The company claims that it keeps user data encrypted and secure via the Solana blockchain. It has previously been described as a “web3 AI wellness company.” (TechCrunch was not able to test the smart ring directly to verify its security claims.)

CUDIS announced $5 million of seed funding in 2024. The round was led by Draper Associates and included a number of other investors, including a number of blockchain-associated investor groups like Skybridge, DraperDragon, Monke Ventures, and Foresight Ventures, among others. The company also plans to launch a Kickstarter soon.

In San Francisco, it feels like OpenClaw is everywhere. Even, potentially, some places it’s not designed to be. According to posts on social media, people appear to be using the viral AI tool to scrape websites and access information, even when those sites have taken explicit anti-bot measures.

One of the ways they are allegedly doing this is through an open source tool called Scrapling, which is designed to bypass anti-bot systems like Cloudflare Turnstile. While Scrapling, which was built with Python, works with multiple types of AI agents, OpenClaw users appear to be particularly fond of the software. On Monday, viral posts promoting Scrapling as a tool for OpenClaw users started to spread on X. Since its release, Scrapling has been downloaded over 200,000 times.

“No bot detection. No selector maintenance. No Cloudflare nightmares,” reads one viral post this week about the open source tool. “OpenClaw tells Scrapling what to extract. Scrapling handles the stealth.”

Cloudflare is not enthused. The company already blocked previous versions of Scrapling, since users of the open source software kept trying to get around anti-scraping protections. This week, the company was working on a patch for Scrapling’s most recent iteration. “We make changes, and then they make changes,” says Dane Knecht, chief technology officer at Cloudflare. He says the company’s trove of website data and its ability to track trends has given it the upper hand.

“We already had a signal that they’re starting to get a higher ability to get around us,” says Knecht. “The team of security operations engineers had already been working on a new set of mediations.”

Large language models were trained on the corpus of the internet—and the process involved a lot of scraping. In some sense, Scrapling users are following in the footsteps of the original model builders, but on a more individualized scale.

Over the past few years, website owners have attempted to put up additional anti-bot protections, either to block software like Scrapling or to find a way to make money off of the bots trying to access their sites. In turn, Cloudflare has been working overtime to keep blocking increasingly powerful bots attempting to get around these protections.

In July 2024, Cloudflare started to offer its customers additional tools that block AI crawlers, unless the bots pay for access. In less than the span of a year, the company claims to have blocked 416 billion unsolicited scraping attempts.

“I Didn’t Know What I was Getting Into”

As Scrapling gained traction in recent days, crypto enthusiasts capitalized on the attention by launching a $Scrapling memecoin. Karim Shoair, who claims to be the sole developer of Scrapling, posted about the memecoin on X (those posts have since been deleted). After the price skyrocketed for around five hours, $Scrapling quickly fell off a cliff as users sold off their stakes. “Bunch of fucking scammers,” reads one comment on the Pump.Fun site that hosts the coin.

“I didn’t know what I was getting into when people made that coin and I endorsed it,” says Shoair, in a direct message with WIRED. “But once I knew, I didn’t want any association with it and the money I withdrew before will go to charity, I won’t benefit from it in anyway. Or maybe just leave it to be wasted.”

In the fallout of this event, the unofficial GitHub Projects Community account, which has over 300,000 followers on X, deleted its posts from this week highlighting Scrapling’s open source software, and appeared to distance itself from the project. “We do not support, promote, or engage in crypto assets, token offerings, trading activity, or crypto-based fundraising,” it said in a post late Monday night.

Putting the crypto forays aside, most software leaders continue to see agents and autonomous AI tools as the future of the web. Even Knecht from Cloudflare, whose work includes blocking bots from nonconsensual scraping, wants to build toward a world where humans and agents benefit from online data and the wishes of website owners are respected. “I see a path forward for an internet that is both friendly to agents and humans,” he says.

This is an edition of Will Knight’s AI Lab newsletter. Read previous newsletters here.

If you want to excel in the robotics field, these programming languages can get you there.

Coding skills are a valuable and often crucial skill for professionals in the STEM space, even in the wake of ‘vibe-coding’ and AI advancements. For those who want to work in the area of robotics, an in-depth knowledge of the uses of programming and an education in more than one language can give a professional an edge.

But more often than not, there are far too many to get through before you find the ones that best suit your ambitions. With that in mind, what are some of the programming languages that tend to go hand-in-hand with a career in robotics.

For beginners

We all start somewhere, so whether you are an enthusiast, a beginner or an established professional, you should have a knowledge of a standard or starter language that can help you get to the next phase.

Python is one such language. Considered accessible, easy to learn, versatile and with a wealth of online resources available, pretty much anyone who wants to learn this language can – if they put in the time. Education platform Coursera notes that in robotics, Python is particularly useful in scripting robot behaviours, quickly building prototypes and in elements requiring artificial intelligence, such as systems that integrate predictive analytics or use machine learning algorithms.

There is no reason that your chosen programming language, to aid a robotics career, can’t be funky as well as functional. If you want a language a little off the beaten track that offers a more unique learning experience, consider Scratch. It is aimed at younger learners and complete beginners, so if you are only at the start of your educational journey, it can be of real help in developing early and foundational skills. This graphical programming language developed by the MIT Media Lab, offers a simple interface, where students can create digital stories, games and animations, improve their conceptual and computational thinking and develop their problem-solving abilities.

High performance

For the professionals or enthusiasts looking for a coding language that packs some punch, there are a number of programs considered to be high performers.

C++ is in this category. An object-oriented language, C++ is regarded as a strong foundation for robotics engineers and those who work in the hardware ecosystem and can be useful to those interested in competitive robotics or embedded systems. Advantages include strong cross-platform support, high execution efficiency, and concise and flexible low-level control.

Another high performer is the less commonly-used Rust. This systems-level language emphasises memory safety and performance, two highly critical aspects of working with robotics software. It has been compared to C++ in CPU-intensive tasks and is an ideal language for those wanting a bit of a challenge in the name of progress and reliability. 

Another high-level language to consider is MATLAB, which is typically used for numerical computation, for example, in work involving linear algebra, data analysis and algorithm development. MATLAB is commonly used by academics, researchers and developers and can be used to create detailed robotic models. Anyone who wants to learn MATLAB can, but it is primarily suited to academic researchers, control systems engineers and those working in labs or companies using MATLAB for rapid iteration. 

Features

So, what makes up a ‘good robotics language’? In truth, that is like asking, ‘how long is a piece of string’? The answer really depends on your own professional needs and ambitions. That being said, there are a number of features and characteristics that often jump out as being important for a programming language used primarily in the robotics space. 

Performance is one such element that is critical in robotics. Experts tend to use a high volume of data and have to make quick decisions, so high performance is important. Ease of use, especially for those just starting off, is also important, as students and professionals often depend on a simple interface and a large community of peers. The robotics space is ever-evolving so flexibility is a key feature, as is easy access to support and resources when needed. 

Anyone looking to boost their coding skills should make sure to check out local groups and organisations, attend industry events, engage with online learning opportunities and if a more in-depth approach is required, consider in-person courses at relevant third-level institutions. 

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.

The global TV business just tilted again. Weeks after Sony sent shockwaves through the industry by negotiating a manufacturing partnership with TCL, Panasonic has made its own decisive move. The company announced a “strategic partnership” with China-based Skyworth aimed at strengthening and accelerating sustainable growth in its U.S. TV business. The agreement takes effect April 1, 2026.

Let’s not pretend these are routine supply-chain tweaks. When two of Japan’s most recognizable TV brands shift production relationships toward Chinese manufacturing giants within weeks of each other, that signals something bigger than cost optimization. It reflects a structural reset in how premium TV brands compete in 2026 — where scale, panel access, pricing pressure, and speed to market matter as much as brand heritage.

Sony’s move was significant. Panasonic’s is equally telling. The balance of power in the TV industry continues to migrate east, and legacy brands are adapting in real time.

“North America remains a key strategic region for Panasonic, with consumers consistently recognizing the exceptional quality and value of our products,” said Akira Toyoshima, CEO of Panasonic Entertainment & Communications Company (PEAC). “The new business model change will leverage the power of Panasonic’s core technical excellence in AV processing, quality, and service standards with the global scale economy of Skyworth’s manufacturing volume and speed to provide a winning formula for the customer value proposition.”

What It Means for the U.S. TV Market in 2026

Here’s what we know so far.

Skyworth USA Corporation will serve as the principal operating partner in the U.S., handling sales, marketing, and logistics. Panasonic, meanwhile, retains responsibility for development expertise and quality assurance to ensure its established performance standards are maintained.

In plain terms: Skyworth will manufacture Panasonic-branded TVs for the U.S. market.

Panasonic has also confirmed it will continue to support all TVs sold through April 2026, as well as models produced under the new agreement. Customer service and warranty coverage are not being handed off and forgotten.

The real question that TV reviewers and industry analysts are already circling, is what this means for Panasonic’s product development going forward, particularly its highly regarded OLED lineup. Over the past several years, Panasonic has delivered OLED models with some of the strongest video processing and image accuracy in the category.

If manufacturing shifts, does the performance DNA remain intact? That’s the part that will determine whether this is simply a business realignment or something more consequential.

Skyworth is hardly new to OLED. The company already produces OLED TVs under its own brand using LG Display panels; the same panel supplier Panasonic relies on. On paper, that creates technical overlap. In practice, the U.S. market is where things get complicated.

Panasonic has struggled to regain meaningful traction in the United States. Once the undisputed king of plasma, the company exited that business in 2014, which led to a 10-year absence from the U.S. TV market. Although Panasonic did continue TV operations internationally, their return to the crowded U.S. TV market in 2024 failed to capture shoppers attention, despite generally high praise among reviewers.

So what happens now?

The likely framework is straightforward: Skyworth manufactures Panasonic-branded TVs at a cost structure designed to win back market share. That requires profitability for both companies. This isn’t charity. It’s math.

Panasonic will define the product, whether OLED or LCD based on its performance standards, processing expertise, and brand expectations. Skyworth’s role will be to engineer and source the components at scale to meet those targets.

The pressure point comes if Panasonic’s traditional quality benchmarks push costs beyond what the U.S. market will bear. At that stage, both sides face a decision: preserve every performance advantage and accept limited volume, or adjust specifications to hit competitive price tiers.

That balance between maintaining Panasonic’s performance DNA and achieving aggressive pricing, will determine whether this partnership strengthens the brand or quietly reshapes it.

In other words, if you’re shopping for a high-end Panasonic TV built entirely under the company’s current structure, now might be the moment to act. Once the Skyworth agreement takes effect and newly manufactured models begin arriving in the U.S., the formula, even if only slightly, could change.

That’s not a knock on Skyworth. It’s a reality of scale manufacturing in a brutally competitive market. If a TV can’t be produced at a price that resonates with a broad enough audience, it won’t survive long no matter how good it looks in a calibration lab. Performance definitely matters, but sales volume keeps the lights on.

Who Is Skyworth? 

Skyworth may not have the brand recognition in the U.S. of fellow Chinese heavyweights Hisense and TCL, but globally, it is a major force. The company ships roughly 36 million TVs annually and ranks among the top five worldwide in TV revenue, reportedly moving ahead of Sony in overall TV sales revenue. That’s not a minor player, which is why this move by Panasonic is rather significant.

Under the broader Skyworth Group umbrella, the company manufactures far more than televisions. Its portfolio includes consumer electronics, display devices, digital set-top boxes, security systems, networking and communications equipment, semiconductors, refrigerators, washing machines, smartphones, and LED lighting. Skyworth sells products under its own brand while also operating as an OEM for other companies — a role that makes this Panasonic agreement less surprising.

In the U.S., recent Skyworth-branded introductions include the Canvas Elite Art TV and the Clarus S1 Outdoor TV — niche-focused models aimed at lifestyle and specialty segments rather than direct mainstream domination.

In addition to the Panasonic-Skyworth agreement, the company has also outlined a broader corporate shift. Effective April 1, 2026, Panasonic confirmed that PEAC (Panasonic Entertainment & Communication) will be reintegrated into the main Panasonic Corporation structure. The stated objective is to strengthen the long-term positioning of Panasonic’s consumer business globally including in the United States.

The Bottom Line 

The TV business isn’t just competitive in 2026. What we are witnessing is a monumental shift in power that is going to change what we buy and from brands that many of us might not have considered a few years ago. But is that bad news for the consumer?

On one side, display technology keeps advancing at a relentless pace: brighter OLED panels, better processing, smarter platforms. On the other, brand stability is anything but secure. Former giants are fighting to stay relevant, and consolidation across both TV and audio continues to thin the herd. Legacy names aren’t disappearing quietly; they’re restructuring, and partnering with rivals in China to make themselves more competitive. That tension was on full display at CES 2026.

While walking the halls of the Venetian during the show, I noticed a suite marked Skyworth. Inside were several televisions clearly wearing the Panasonic badge. Not concept sketches. Not mockups. Finished sets. I took a few photos. That didn’t last long.

I was quickly informed that photography wasn’t permitted and that what I was seeing was not something I could report on. At the time, it was obvious something was brewing. The question wasn’t whether Skyworth was involved with Panasonic, it was how deep the relationship would go.

Now we know.

With Panasonic’s official announcement confirming a strategic partnership and manufacturing shift, what felt like a quiet industry rumor on a CES show floor has become a formal restructuring of one of Japan’s most storied TV brands.

In the case of both Sony and Panasonic, it will likely take another year before the real impact of their manufacturing partnerships with major China-based TV companies becomes clear, as new product cycles roll out and revised models reach the U.S. market. 2026 will be a transition year, and 2027 should provide a more accurate picture of how these strategic shifts affect performance, pricing, and brand positioning.

Find out more about skyworthusa.com.

Related Reading:

In an era where “new model” usually means “time to replace the old one,” Cambridge Audio just did something refreshingly unglamorous and important. The company has rolled out Qobuz Connect to its legacy StreamMagic Gen 2 and Gen 3 platforms, extending modern streaming control to products that have been sitting in racks and on shelves for years.

That means owners of the StreamMagic 6 V2, CXN, CXR, Azur 851N, CXN V2, and Edge NQ can now control playback directly from the Qobuz app via an OTA update; no new hardware, no forced upgrade cycle. Qobuz still works through the StreamMagic app as before, but Connect adds a more direct, native control path that many subscribers prefer.

It’s good to see a brand remember that its customer base didn’t magically appear in the last 24 months. Supporting Gen 2 and Gen 3 after rolling out Qobuz Connect to Gen 4 isn’t flashy, but it sends a clear message: what you bought still matters. Your streamer didn’t just age out because a new box showed up with a higher number on it.

It also proves something else; customers aren’t shouting into the void. The questions, feature requests, and yes, the complaints that show up on Instagram, Facebook, Audiogon, and the forums? They do get seen. Sometimes they even lead to action. That kind of accountability shouldn’t be rare in the hi-fi category, but too often it is.

There are plenty of alternatives out there. Some cost less. Some arguably do more. In a crowded field, long-term support isn’t just goodwill on the part of the manufacturer but smart business strategy. It’s also hard not to wonder whether Cambridge’s recent decision to appoint Fidelity Imports as its new U.S. distributor played a role here. Distribution shifts aren’t just about logistics and dealer networks, they’re about market responsiveness. If you’re trying to rebuild momentum and reinforce trust in a crowded category, showing existing owners that their gear still has a future is a smart place to start.

This update follows Cambridge’s recent additions of Spotify Lossless, Amazon Music, and QPlay to the StreamMagic platform. Rather than introducing new hardware, the company has continued expanding service support across existing models.

What Is Qobuz Connect and Why Should Cambridge Audio Owners Care?

Qobuz Connect lets subscribers stream directly from the Qobuz app to compatible devices over Wi-Fi or Ethernet, turning the app into the primary control interface. Your Cambridge streamer (or other supported hardware) becomes the playback endpoint, while your phone or tablet functions as the remote. The result is native control with full lossless and high-resolution streaming up to 24-bit/192 kHz. So no Bluetooth, no compressed handoffs, and no need to route playback through a secondary app.

Launched on May 15, 2025, Qobuz Connect has expanded quickly. In less than a year, support has grown to more than 100 hardware partners, covering A/V receivers, network streamers, powered speakers, and integrated amplifiers. That broader compatibility keeps control centralized inside the Qobuz app while giving subscribers more flexibility across systems and brands.

The Bottom Line

The biggest winner here is the existing Cambridge Audio owner. Adding Qobuz Connect to StreamMagic Gen 2 and Gen 3 extends the useful life of hardware that, in many cases, is still performing at a very high level. It removes friction, adds native app control, and keeps those products aligned with how people actually stream music in 2026. No new box required. No forced upgrade cycle.

In 2026, StreamMagic isn’t just a control app, it’s Cambridge’s full streaming ecosystem. It supports major services including Qobuz, TIDAL, Spotify (including Lossless where available), Amazon Music, QPlay, AirPlay 2, Chromecast built-in, Roon Ready integration, and local UPnP playback.

With Qobuz Connect now folded into that framework across multiple generations, Cambridge has strengthened the platform’s long-term relevance. In a crowded streaming market where alternatives are plentiful and often cheaper, sustained software support is what separates a product you replace from one you keep.

Related Reading:

A former candidate in the 2026 race for governor of California and a popular YouTuber have been kicked off Kalshi’s platform for alleged violations related to insider trading, the popular prediction market revealed Wednesday.

In a blog post detailing the cases, Kalshi’s head of enforcement, Robert DeNault, noted that the company’s surveillance system had flagged suspicious behavior in both instances.

In the case of the political candidate, Kalshi cited a video posted online “that appeared to show him trading on his own candidacy.” Kalshi froze the candidate’s accounts and reported the activity to the Commodity Futures Trading Commission, the government agency that oversees prediction markets. It instituted a five-year ban and is fining the account a penalty 10 times the size of the initial trade, which Kalshi says it intends to donate to charity.

While Kalshi does not mention the candidate by name, DeNault notes that they have recently dropped out of the race and shifted to running for Congress. This fits the description of only one person involved in the race: Kyle Langford, a far-right Republican candidate best known for inflammatory antisemitic comments praising Nazis. Langford has since dropped out and launched a campaign styled as a progressive Democrat in California’s 26th district.

In May 2025, Langford posted a video to X showing a screen recording of a trade order placed on Kalshi for an event in the governor’s race. Shortly after the incident, Kalshi confirmed that it was looking into Langford’s activities. In his description of the investigation into the candidate, DeNault notes that Kalshi began its investigation that May.

“Tensions between the USA and Iran are at an all-time high, and the media has chosen to cover a $200 campaign gimmick (aka betting on I, myself) from last year,” Langford said in a statement to WIRED. “Is this really the state of our political discourse?”

This is not the only suspension tied to the California governor race; as Politico reported on Tuesday, Democratic megadonor Stephen Cloobeck, who was briefly a candidate before dropping out to endorse representative Eric Swalwell, has also been blocked by the platform for attempting to trade on the race. “In the event we ever see a candidate trading on their own candidacy, it triggers disciplinary action on the exchange. Their trade gets frozen. They may be permanently suspended from having an account,” DeNault said in a statement. Cloobeck confirmed the bets to Politico and said he is still able to trade on other events.

This crackdown underscores how expansive definitions of “insider trading” can be on prediction markets; while traditional insider trading cases hinge on subjects profiting from “material nonpublic information”—confidential data or intelligence that can impact stock prices—here, even placing bets while working on an election or running for office can violate the rules.

In the case of the YouTube streamer, Kalshi reports that its surveillance systems flagged the account based on “statistically anomalous” trading success. It also received concurrent tips from Kalshi users, who had flagged the unusual activities. “We investigated and found that the trader was employed as an editor for the streamer’s show and likely had access to material nonpublic information connected to his trading,” DeNault writes in the report. The account was frozen before it could withdraw funds; it is now suspended for two years, and it has also received a financial penalty. Kalshi did not disclose the identity of the YouTube streamer.

As prediction markets have exploded in popularity, there has been a series of high-profile incidents of suspected insider trading, including major trades made just prior to geopolitical events like the US capture of Venezuelan leader Nicolás Maduro. In Israel, two Polymarket traders were recently arrested for leaking classified information in connection to trades made on military activity. Following the Maduro incident, Congress introduced a bill to ban government officials from insider trading on prediction markets—but there have not been any enforcement actions made public.

Google is making it very easy to feel like a music producer, whether you have the skills or not. The company has announced that ProducerAI, an AI-powered music-making platform, is joining Google Labs, bringing together sound generation, visuals, and video into one experimental creative tool.

ProducerAI is powered by a preview version of Lyria 3, Google’s newest music-generation model. You can describe what you want, and the AI helps you build it, whether that means crafting beats, shaping melodies, or experimenting with entirely new sounds.

ProducerAI first launched in July 2025 to let users collaborate with an AI agent to generate music, workshop lyrics, and remix tracks from text prompts. Until now, it relied on its own underlying models. Joining Google Labs gives it access to a much larger AI toolkit.

Inside Google’s AI-powered music studio

As a part of Google Labs, ProducerAI will pull from several of Google’s models. Lyria 3 handles music generation, while Gemini powers the conversational interface that guides users through ideas and edits.

Nano Banana will generate album art, and Veo will be used to create AI-generated music videos, turning a song idea into a complete audio and visual project. Google says it will also embed SynthID watermarks into ProducerAI outputs.

This flags AI-generated audio, images, video, and text, adding transparency as AI music becomes harder to distinguish from human work. But companies like Sony have already developed a tool to detect original songs used in AI-generated tracks.

The ProducerAI team has already worked with artists like The Chainsmokers, Lecrae, and Anjulie to shape the platform. Google positions ProducerAI as an experiment, not a replacement for musicians. However, this collaboration arrives at a time when AI-generated songs are topping Billboard charts, drawing scrutiny from artists and listeners alike.

For years, the “last mile” of digital transformation has been littered with forgotten PDFs and ignored training manuals. 

Organizations spend millions on sophisticated software like SAP or Salesforce, only for employees to struggle with basic navigation. Now, as the era of agentic AI arrives, companies face a double-edged sword: they must teach human employees to collaborate with AI, while simultaneously teaching AI agents to navigate the labyrinthine interfaces of the modern enterprise.

One idea that seems to be gaining momentum among AI-forward businesses: using screen recordings and tutorials/walkthroughs of someone performing an enterprise task — be it creating a new ticket or processing an invoice — and training AI to replicate the flow based on the screen capture. Just this week, a startup called Standard Intelligence went viral on X showing an early demo of open-ended version of this for the physical and digital world.

But the truth is, there are already players tackling this problem for the enterprise itself square-on: case-in-point, Guidde, an Israel startup born during the video-centric years of the COVID-19 pandemic, today announced an oversubscribed $50 million Series B funding round led by PSG Equity to address this exact knowledge infrastructure crisis. 

Instead of feeding an agent a static PDF manual, Guidde provides high-fidelity “Video Ground Truth”—a rich stream of data captured from real human experts as they navigate complex software.

The investment signals a shift in how the tech industry views documentation—not as a static byproduct of work, but as the critical telemetry needed to train the next generation of autonomous digital agents.

Technology: from video capture to world models

At its core, Guidde is an AI Digital Adoption Platform (ADAP). However, its technological breakthrough lies in what happens behind the scenes during a recording.

Guidde isn’t just recording pixels; it is capturing every click, scroll, and latent interaction with the HTML page—the subtle pauses, the specific scroll depths, and the corrections a human makes when a system lags. This telemetry transforms raw video into a Vision-Language-Action (VLA) training set.

Meanwhile, the platform’s Magic Redaction automatically obscures sensitive data like passwords or credit card numbers during capture, ensuring materials remain secure and HIPAA-aligned.

“Every time you click a button, you drag-and-drop, you scroll, you type, we gather the interaction… all of it, we do cleanse it—there’s no private information,” explained Guidde co-founder and CEO Yoav Einav in an exclusive interview with VentureBeat.

Under the hood, the platform captures the underlying metadata and DOM (Document Object Model) changes synchronized with the video frames. The differentiator is the telemetry hidden beneath the surface.

This rich metadata creates a “digital world model” of enterprise software. And because each enterprise uses its own unique mix of apps and processes, Guidde is creating a data moat that allows enterprise agents to reason through legacy UIs with the same spatial awareness as a human, ensuring that automation actually works in a production environment rather than just a lab demo.

For a human, it’s a tutorial. For an AI agent, it is a high-fidelity map of the interface. This allows agents to “see” and reason through complex UIs the way humans do, solving the “last mile” of automation where agents previously failed due to lack of specific enterprise and in-situ usage context.

In a sense, Guidde is building a “self-driving car” like a Waymo for computer usage.

Product: three pillars of Guidd-ance

The platform has evolved into three distinct products designed to scale with an organization’s maturity:

1. Guidde Create: The engine for subject matter experts to turn workflows into documentation in minutes.

2. Guidde Broadcast: A personalized recommendation engine—often compared to Netflix—that delivers answers inside the tools people actually use. It knows who the user is and what department they are in to surface relevant content exactly when needed.

3. Guidde Discover: The newly launched “agentic” pillar. Like Waze mapping roads by observing drivers, Discover maps software routes by tracking how employees work. It understands the workflow, creates the content, and updates it automatically when the UI changes.

Training humans how to use AI — and AI using humans

The most non-obvious aspect of Guidde’s growth is its dual-purpose mission. “We’re the only platform that trains both humans and agents,” Einav stated.

As companies roll out AI tools like Microsoft 365 Copilot or ServiceNow agents, they hit a proficiency gap. One of Guidde’s largest customers revealed they were paying over $1 million a year for a sophisticated AI tool, yet “nobody knows how to use them because they did like a 30-minute training session, and then that’s it.” Guidde closes this gap by providing “bite-sized” video tutorials in the flow of work.

Simultaneously, these videos train the AI agents themselves. Foundation models like Gemini or GPT-4 often hallucinate when tasked with specific enterprise workflows because they weren’t trained on the highly specific, internal “vanilla workflows” found in private enterprise systems. Guidde provides the “starting point,” the “metadata,” and the “x, y coordinates of the button” that an agent needs to complete an action without getting stuck.

The multimodal advantage

To maintain this level of accuracy, Guidde employs a multimodal infrastructure. The system doesn’t rely on a single model; instead, it uses a “fleet” of models that evaluate one another.

• Google Gemini: Generally used for visual tasks like analyzing PDFs or PowerPoints.

• Anthropic Claude: Leveraged for writing the storyline and narrative scripts.

• Feedback Loops: When a user edits a video, that data is fed back into the model to prevent the same mistakes from occurring in future captures.

This approach allows Guidde to replace a legacy stack of six or seven disconnected tools—Loom for capture, Adobe Premiere for editing, 11Labs for text-to-speech, and Synthesia for avatars—with a single, AI-native platform. “We basically pack everything for you,” Einav says, “and automate the entire process based on your brand guidelines.”

Video-first origin story

The genesis of Guidde lies in a frustration familiar to any product leader. Before founding the company, Einav and co-founder Dan Sahar spent years mastering video traffic at Qwilt, a company they started in 2010 to analyze how people watched Netflix and Disney+. 

When COVID-19 hit, they saw a massive opportunity to apply that video expertise to the workplace. They observed that short video explainers could increase free-to-paid account conversions by 30%, but the friction of creating them was unsustainable.

In an interview, Einav recalled the “tedious work” of the old world: “My team in Israel were creating the content, someone in the US with a US accent was doing the narration, someone in the marketing team would write the script… and someone in the enablement team would do the edit.” This fragmented workflow meant a single video took two to three weeks to produce. “And then two weeks later, the product changes, and you need to redo it from scratch,” Einav added.

Guidde was built to collapse this cycle into seconds. By automating the “Magic Capture” of a workflow, the platform generates a structured narrative script and professional AI voiceover instantly. This removes the editing bottleneck, transforming subject matter experts into “training powerhouses.”

Licensing and market impact

Guidde’s pricing structure reflects its transition from a utility to a core piece of enterprise infrastructure:

• Free: $0 (Up to 25 videos, web-app support).

• Pro: $18/creator/month (Unlimited videos, brand kits).

• Business: $39/creator/month (Unlimited text-to-voice, analytics).

• Enterprise: Custom pricing (Multi-language translation, SSO, Magic Redaction).

The platform’s impact is already visible in the numbers: a 41% reduction in video creation time and 34% fewer inbound support tickets. 

For customers like Emerson, this translates to 40–60% quicker guide creation. Support teams, in particular, are finding they can offload 80% of their ticket volume with agents—but only if those agents have the content to be useful. 

“The agent without the content is useless,” Einav warns, noting that most enterprise documentation is either years out of date or entirely undocumented.

Community and industry early reception

Guidde already claims 4,500 enterprise customers and seeks to expand this number with its new round of funding. Support and operations leaders have been vocal about the platform’s ease of use. Christopher Cummings, VP of Client Experience at DocNetwork, highlighted its ability to provide “quick, personalized video responses to customer questions.”

 Meanwhile, Wren Cotrone, a Director of Customer Support, noted that “Once you set the branding the way you want, you can really zoom through this stuff.”

Ronen Nir, Managing Director at PSG, summarized the investment thesis: “Guidde is solving one of the biggest blockers to successful AI adoption: the knowledge infrastructure.”

Why this matters now

The paradigm shift from text-only LLMs to agentic video intelligence is the defining trend of 2026. Guidde’s Series B signals that the “ground truth” for enterprise agents will come from raw video observation, not static documentation.

By capturing how work gets done across 10s of millions of workflows, Guidde is building a dataset that few others possess. 

As Einav put it: “It starts with humans in the loop, and over time moves toward full autonomy.” For the modern enterprise, the map is no longer a static document—it’s a living, breathing video intelligence layer that guides both the workforce and the agents that support them.