Last week, Taylor Swift filed a trio of trademark applications to protect her image and voice. One is meant to cover a well-known photograph of the pop singer holding a pink guitar during a concert on her record-breaking Eras tour, while the two sound trademarks are for simple identifying phrases: “Hey, it’s Taylor Swift” and “Hey, it’s Taylor.”

The move comes as AI deepfakes continue to proliferate across social media. Any individual stands to have their likeness exploited in the creation of nonconsensual AI-generated material; earlier this month, an Ohio man was the first person convicted under a new federal law criminalizing “intimate” visual deceptions of this sort. Celebrities, meanwhile, find themselves at risk of both explicit deepfakes and false endorsements.

A new report from the AI detection company Copyleaks shows that Swift and other stars have recently had their likenesses used in scammy advertisements. Researchers identified a cluster of sponsored videos on TikTok that appeared to show Swift, Kim Kardashian, Rihanna, and others promoting “potentially fraudulent or malicious services,” with the clips making use of what the researchers call “realistic-sounding voices” as well as “textured filters meant to mask some of the flaws in the AI-generated visuals.”

The fake ads show Swift et al. in what seem to be common interview settings—red carpet events or talk show sets. Rather than answering questions, however, the AI-generated celebrities talk up supposed rewards programs in which TikTok users are paid for offering feedback on content served to them.

“I was reading about digital behavior this week and came across a testing feature called TikTok Pay,” says a deepfaked Swift in an ad that uses manipulated footage from an appearance the real Swift made on The Tonight Show Starring Jimmy Fallon in October. “Certain users are being invited to watch videos and submit opinions.” The deepfaked Swift goes on to say that the program is in “limited rollout” for the moment but encourages viewers to see if they qualify for it, adding: “If the page opens for you, don’t overthink it.”

Naturally, anyone who clicks is accepted. These ads eventually lead the user to a third-party service that, despite the TikTok name and logo, has evidently been vibe coded using the AI platform Lovable, whose own branding appears on the page and in the URL. At this point, the researchers say, the user is prompted to begin entering their name and personal information.

While it’s not clear what the advertisers intend to with all the data mined through their celebrity deepfake promotion, scam ads with similar objectives are exceedingly common. Last week, the nonprofit Consumer Federation of America sued Meta, alleging that the tech giant misled Facebook and Instagram users about its efforts to crack down on scam ads—and profited by allowing them to proliferate. On Monday, the US Federal Trade Commission reported that social media scams have surged overall, with Facebook scams accounting for the highest total of financial losses.

It’s no surprise that Swift and her peers are taking legal steps to distance themselves from this fraudulent economy. While Swift hasn’t publicly commented on the reasoning behind her trademark filings, the reputational damage that deceitful deepfakes pose to her billion-dollar brand can hardly be overlooked. The trouble is, they grow more sophisticated by the day.

Strands is the NYT’s latest word game after the likes of Wordle, Spelling Bee and Connections – and it’s great fun. It can be difficult, though, so read on for my Strands hints.

Want more word-based fun? Then check out my NYT Connections today and Quordle today pages for hints and answers for those games, and Marc’s Wordle today page for the original viral word game.

Something shifted in enterprise RAG in Q1 2026. VB Pulse data spanning January through March tells a consistent story: the market stopped adding retrieval layers and started fixing the ones it already has. Call it the retrieval rebuild.

The survey covered three consecutive monthly waves from organizations with 100 or more employees, with between 45 and 58 qualified respondents per month across platform adoption, buyer intent, architecture outlook and evaluation criteria. The data should be treated as directional.

Enterprise intent to adopt hybrid retrieval tripled from 10.3% to 33.3% in a single quarter — even as 22% of qualified enterprise respondents reported having no production RAG systems at all. For data engineers and enterprise architects building agentic AI infrastructure, the data reveals a market in active transition: the RAG architecture most enterprises built to scale is not the one they expect to run by year-end. 

Hybrid retrieval has become the consensus enterprise strategy. Unlike single-method RAG pipelines that rely on vector similarity alone, hybrid retrieval combines dense embeddings with sparse keyword search and reranking layers, trading simplicity for the retrieval accuracy and access control that production agentic workloads require.

The standalone vector database category is under pressure. Weaviate, Milvus, Pinecone and Qdrant each lost adoption share across the quarter in the VB Pulse data. Custom stacks and provider-native retrieval are absorbing their displaced share.

A growing minority of enterprises are stepping back from RAG altogether — a signal that the market’s maturity narrative has meaningful exceptions.

Organizations that went wide on RAG in 2025 are hitting the same failure point: the architecture built for document retrieval does not hold at agentic scale.

Enterprises that scaled RAG fast are now paying to rebuild it

The two largest intent movements in Q1 are directly connected — enterprises confronting retrieval quality problems at scale, and hybrid retrieval emerging as the consensus answer.

Investment priorities shifted in parallel. Evaluation and relevance testing led budget intent in January at 32.8% and fell to 15.6% by March. Retrieval optimization moved in the opposite direction, from 19.0% to 28.9% — overtaking evaluation as the top growth investment area for the first time. 

Steven Dickens, vice president and practice lead at HyperFRAME Research, described the operational burden enterprise data teams are facing in a VentureBeat interview in March on Oracle’s agentic AI data stack. “Data teams are exhausted by fragmentation fatigue,” Dickens said. “Managing a separate vector store, graph database and relational system just to power one agent is a DevOps nightmare.”

That fatigue shows directly in the platform data. The custom stack rise to 35.6% is not a rejection of managed retrieval — many organizations run both. It is a consolidation response from engineering teams that have hit the limits of assembling too many components.

Not every enterprise has made it that far. The VB Pulse data includes a signal that complicates the market’s overall growth narrative: 22.2% of qualified respondents reported no production RAG by March, up from 8.6% in January.  The report attributes this cohort to organizations that have “not yet committed to any retrieval infrastructure, or have paused programs” — concentrated in Healthcare, Education and Government, the same sectors showing the highest rates of flat budgets.

Standalone vector databases are losing the adoption argument but winning the reliability one

Recent reporting by VentureBeat illustrates why the dedicated retrieval layer still matters in production. 

Two enterprises building on Qdrant show why purpose-built vector infrastructure still wins in production.

 &AI builds patent litigation infrastructure and runs semantic search across hundreds of millions of documents. Grounding every result in a real source document is not optional — patent attorneys will not act on AI-generated text. That requirement makes the architectural choice clear.

“The agent is the interface,” Herbie Turner, &AI’s founder and CTO, told VentureBeat in March. “The vector database is the ground truth.”

GlassDollar, a startup that helps Siemens and Mahle evaluate startups, runs an agentic retrieval pattern across a corpus approaching 10 million indexed documents. A single user prompt fans out into multiple parallel queries, each retrieving candidates from a different angle before results are combined and re-ranked. That query volume and precision requirement is what drove the choice of purpose-built vector infrastructure.

“We measure success by recall,” Kamen Kanev, GlassDollar’s head of product, told VentureBeat in March. “If the best companies aren’t in the results, nothing else matters. The user loses trust.”

The VB Pulse data shows that framing — retrieval as ground truth rather than feature — is gaining traction across the broader enterprise market, even as standalone vector database adoption declines. 

Why enterprises say they need a dedicated vector layer shifted significantly across Q1. In January the top reasons were access control complexity (20.7%) and retrieval precision (19.0%). By March, operational reliability at scale had surged to 31.1% — more than doubling and overtaking everything else. Enterprises are no longer keeping vector infrastructure primarily for precision. They are keeping it because it is the part of the stack they can rely on when query volumes scale.

How enterprises are redefining what good retrieval means

How enterprises judge their retrieval systems shifted notably across Q1 — and the direction of that shift points to a market getting more sophisticated about what good retrieval actually means.

In January, response correctness dominated evaluation criteria at 67.2% — far above anything else. By March, response correctness (53.3%), retrieval accuracy (53.3%) and answer relevance (53.3%) had converged exactly. Getting the right answer is no longer enough if it came from the wrong document or missed the context of the question.

Answer relevance was the only criterion that rose across the quarter, gaining five percentage points. It is also the hardest to measure — whether the retrieved context is actually the right context for that specific question requires purpose-built evaluation infrastructure, not just pass-or-fail correctness checks. Its rise signals that a meaningful share of enterprise buyers have moved past basic RAG testing entirely. 

The market’s verdict: RAG isn’t dead. The original architecture is

The “RAG is dead” narrative had real momentum heading into 2026. It rested on two claims. The first: that long-context windows — models capable of processing hundreds of thousands of tokens in a single prompt — would make dedicated retrieval unnecessary. The second: that agentic memory systems, which store what an agent learns across sessions rather than retrieving it fresh each time, would absorb the knowledge access problem entirely.

The VB Pulse data is the enterprise market’s answer to the first claim. The long-context-as-dominant-architecture position collapsed from 15.5% in January to 3.5% in February before partially recovering to 6.7% in March. January’s sample was heavily weighted toward Technology and Software respondents — the segment most exposed to long-context model announcements in late 2025. As the sample diversified, the position evaporated.

On the memory question, Jonathan Frankle, chief AI scientist at Databricks, framed the architecture clearly in a March interview with VentureBeat: a vector database with millions of entries sits at the base of the agentic memory stack, too large to fit in context. The LLM context window sits at the top. Between them, new caching and compression layers are emerging — but none of them replace the retrieval layer at the base. New agentic memory systems like Hindsight, developed by Vectorize, and observational memory approaches like those in the Mastra framework address session continuity and agent context over time — a different problem than high-recall search across millions of changing enterprise documents.

The most consequential signal: the share of respondents not expecting large-scale RAG deployments by year-end grew from 3.4% to 15.6% — nearly 5x. That is not a verdict against retrieval. It is a verdict against the retrieval architecture most enterprises built first.

The retrieval rebuild is not optional

The retrieval rebuild is the cost of scaling RAG without first deciding what architecture could actually support it.

If your organization is among the 43.1% that entered Q1 planning to expand RAG into more workflows, the VB Pulse data suggests that plan has already changed for many of your peers — and may need to change for you. Hybrid retrieval is the consensus destination. Custom stack growth to 35.6% reflects teams building retrieval infrastructure around requirements that off-the-shelf products do not fully address.

RAG is not dead. The architecture most enterprises used to implement it is. The data suggests the rebuild is not a future decision. For 33% of enterprises, the rebuild is already the stated priority.

Beijing’s commerce ministry has formally submitted a 30-page document warning the European Commission that its draft Cybersecurity Act, which would make vendor removal mandatory for the first time, could trigger reciprocal measures against European companies in China.

China has formally threatened the European Union with retaliation if a sweeping new cybersecurity law leads to the exclusion of Chinese firms, including Huawei and ZTE, from European critical infrastructure.

The Chinese Ministry of Commerce submitted a 30-page document to the European Commission, reported earlier by the South China Morning Post, explicitly warning that Beijing is prepared to invoke its Foreign Trade Law and State Council Supply Chain Security Regulations, legal frameworks that allow China to restrict trade, investigate foreign entities, and impose reciprocal bans on European companies, if Chinese firms face what it calls discriminatory treatment.

The document was submitted on April 17 to the Commission. MOFCOM spokesperson He Yongqian confirmed the submission at a press briefing on April 24, framing China’s core objection as the draft law’s use of ‘non-technical risk’ factors, a mechanism Beijing argues is a subjective political tool designed to exclude Chinese companies regardless of the actual security properties of their equipment.

What the EU Cybersecurity Act proposes?

The revised EU Cybersecurity Act, announced by the European Commission on January 20, represents a fundamental shift in how Brussels approaches network security. Since 2020, the EU’s ‘5G toolbox’ has recommended that member states avoid high-risk vendors in 5G networks.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

That recommendation has been implemented unevenly: only 13 of 27 member states had acted on it by the time the new law was announced, and several of the bloc’s most significant economies, including Germany, where Huawei provided equipment across approximately 60% of 5G sites as recently as late 2024, had been slow to act.

The new law changes the legal basis from recommendation to obligation. It would require member states to remove equipment from vendors designated as high-risk suppliers from communications networks within three years of the law entering into force.

It also creates a mechanism under which the Commission can designate an entire country as a ‘cybersecurity threat,’ which would trigger exclusions extending beyond telecoms into 18 critical sectors, including energy, transport, and information technology.

The law does not name Huawei or ZTE explicitly, but the intent is unambiguous: EU Tech Commissioner Henna Virkkunen said it would give the bloc ‘the means to better protect our critical supply chains,’ and Strand Consult data puts Chinese vendors’ share of European 5G infrastructure at between 33% and 40%. A full removal would be the largest forced replacement of telecoms infrastructure in European history.

The precedent that makes Beijing’s threat credible

China’s retaliation threats have a documented track record. When Sweden banned Chinese vendors from its 5G networks in 2020, Ericsson’s revenues in China fell 46% the following year.

The company has never recovered that business. Nokia, which has maintained a small footprint in the Chinese market, has watched its China revenues fall from roughly €2.5 billion in 2018 to approximately €913 million last year.

Nokia executives have told the company internally that it faces a total ban in China for national security reasons, with Nokia’s president of mobile networks, Tommi Uitto, publicly stating that the combined China market share of both Nordic vendors has dropped to 3%.

The asymmetry is pointed. China has already been restricting Nokia and Ericsson, the two European companies that stand to benefit most from a Huawei ban, while simultaneously warning the EU that it will face consequences if it formalises its own exclusions.

That double standard is increasingly being called out. Nokia CEO Justin Hotard has contrasted Europe’s continued openness to Huawei with China’s restrictions on European vendors, and Ericsson’s Börje Ekholm has estimated the EU revenue opportunity from replacing Chinese kit at a ‘sizeable’ number given Huawei and ZTE’s combined European market share.

The Swedish precedent also illustrates the implementation challenge the EU faces independently of Chinese pressure. The UK mandated the removal of Huawei from 5G networks by the end of 2027. BT missed the 2023 deadline for its core network.

Germany ordered Huawei removed from the 5G core by the end of 2026, a deadline that applies to a part of the network Huawei was not even present in when the rules were announced, while allowing retention of Huawei’s radio access network until 2029. The practical reality of a three-year EU-wide rip-and-replace at scale is, as Light Reading noted, ‘ambitious and compliance is not certain.’

What Beijing is threatening and why?

China’s 30-page submission argues on four grounds. First, the ‘non-technical risk’ framework is discriminatory on its face, targeting companies by country of origin rather than by demonstrated security flaw. Second, the law violates WTO principles of non-discrimination and proportionality.

Third, that designating China as a ‘country of cybersecurity concern’ would, if triggered, extend exclusions far beyond telecoms into clean energy, automotive, and industrial supply chains where Chinese companies are deeply embedded in European markets.

Fourth, that European companies operating in China, German automakers with €90 billion in annual exports, Dutch chipmakers, French luxury and aerospace firms, would face reciprocal market access restrictions.

The legal mechanisms cited, China’s Foreign Trade Law and the State Council’s Supply Chain Security Regulations, are the same frameworks Beijing has used in previous technology trade disputes. They permit retaliatory trade restrictions, procurement bans, investigations into foreign entities, and entity list designations that mirror the US model China publicly decries.

The spokesperson’s framing, that China ‘still views cooperative dialogue as the correct path’, is the standard diplomatic hedging that accompanies formal coercive submissions of this kind.

A geopolitically loaded moment

The Trump administration has simultaneously been pressuring the EU to accelerate Huawei removal while threatening tariffs over EU enforcement actions against US tech companies.

The EU is navigating a position in which it faces pressure from Washington to act on Huawei and pressure from Beijing not to, while also trying to maintain economic relationships with both.

Germany, the member state with the most at stake both in terms of Huawei infrastructure and Chinese market exposure for its automotive sector, has been the most cautious about implementation pace.

For Nokia and Ericsson, the stakes are direct. Both were among the companies expected to meet EU leadership precisely around the question of European tech competitiveness and strategic supply chain policy.

A full European Huawei ban would represent the single largest new revenue opportunity the Nordic vendors have had in years. Whether the EU actually follows through, given member state reluctance, the implementation timeline, and Beijing’s explicit threat, is now the central question.

The Cybersecurity Act must still be negotiated with EU governments and the European Parliament before it becomes law. No timeline for that process has been confirmed.

China’s formal submission is designed to influence that negotiation, and the governments most exposed to Chinese trade retaliation, Germany, the Netherlands, and France, are also the ones whose implementation of the existing 5G toolbox has been most limited.

Looking for the most recent Wordle answer? Click here for today’s Wordle hints, as well as our daily answers and hints for The New York Times Mini Crossword, Connections, Connections: Sports Edition and Strands puzzles.

Today’s Wordle puzzle is a bit tricky, thanks to its repeated letter. If you need a new starter word, check out our list of which letters show up the most in English words. If you need hints and the answer, read on.

Read more: New Study Reveals Wordle’s Top 10 Toughest Words of 2025

Today’s Wordle hints

Before we show you today’s Wordle answer, we’ll give you some hints. If you don’t want a spoiler, look away now.

Wordle hint No. 1: Repeats

Today’s Wordle answer has one repeated letter.

Wordle hint No. 2: Vowels

Today’s Wordle answer has one vowel.

Wordle hint No. 3: First letter

Today’s Wordle answer begins with C.

Wordle hint No. 4: Last letter

Today’s Wordle answer ends with K.

Wordle hint No. 5: Meaning

Today’s Wordle answer can refer to a thick earthenware pot or jar.

TODAY’S WORDLE ANSWER

Today’s Wordle answer is CROCK.

Yesterday’s Wordle answer

Yesterday’s Wordle answer, April 29, No. 1775, was RURAL.

Recent Wordle answers

April 25, No. 1771: WOMEN

April 26, No. 1772: GLOSS

April 27, No. 1773: EERIE

April 28, No. 1774: QUACK

Schiit Audio is taking a more practical swing at tubes with the $99 Buf. It’s a compact tube buffer designed to sit in your signal chain, not take it over.

Buf isn’t a preamp and it’s not a DAC. There are no inputs for sources beyond basic line level, no volume control, and no system control duties. You place it between components; typically between a DAC and amplifier, or between a source and powered speakers, and it inserts a tube stage into the chain. If you don’t want that, it can be switched out and used as a straight pass through.

That makes it easy to experiment without committing to a full tube setup. It can be used to slightly reshape a solid state system, take the edge off a brighter chain, or add some variation to a desktop or headphone rig. It’s also flexible enough to move around depending on the system or use case, which fits Schiit’s usual modular approach.

At $99, Buf is less about replacing components and more about giving users a simple way to try a tube stage in different setups and decide if it works for them.

Tubes Anywhere Without Rebuilding Your System

Schiit Audio isn’t pretending the $99 Buf is neutral. In fact, they’re leaning in the opposite direction.

“Buf is a tube buffer. It adds a ton of low-order harmonic distortion, without adding a bunch of noise. It destroys the big perfect of ‘measurement gear.’ At the same time, lots of people, including those who use Audio Precisions all day, think it sounds better. So we figured we’d make this thing and let you find out for yourself,” said Jason Stoddard.

That’s the pitch. Not accuracy in the lab sense, but a different presentation that some listeners may prefer, especially in systems that lean hard into ultra-low distortion solid state performance.

Despite the price, this isn’t a stripped-down implementation. The Schiit Buf runs a 100V plate voltage, uses a linear power supply, and incorporates higher-grade parts including Panasonic film capacitors. Schiit’s Coherence topology is also in play, maintaining absolute phase and offering selectable gain; either 0dB or 12dB from a front panel switch.

Stoddard is also clear that this isn’t a generic design.

“It’s not just another cathode follower,” he explained. “It’s more akin to our Lyr and Vali tube amps, but it’s different than both—a simple Class A stage optimized for line level use, rather than driving headphones.”

That last part matters. Buf isn’t trying to power anything. It’s there to sit in the chain and influence it, subtly or not, depending on the system and gain setting.

Connectivity is simple: RCA in, RCA out. It will work in most two-channel or desktop setups without much thought. And because it can be fully bypassed, it’s easy to evaluate in real time without pulling cables.

Schiit Buf Specifications

• Gain Modes
  • Low Gain (0dB): 20Hz–20kHz ±1dB, THD <0.5%, IMD <0.6%, SNR >106dB, Crosstalk -90dB
  • High Gain (12dB): 20Hz–20kHz ±1dB, THD <0.2%, IMD <0.4%, SNR >97dB, Crosstalk -80dB
• Performance & Design
  • Output Impedance: 75 ohms
  • Input Impedance: 470k ohms
  • Maximum Output: 8.2V RMS
  • Topology: Coherence tube gain with BJT inverter, Class A
  • Protection: Delayed start, fast shutdown, muting relay
• Power & Build
  • Power Supply: External 24VAC and 6VAC wall adapter, linear regulated rails, 6V AC heater
  • Power Consumption: 6W
  • Dimensions: 5 x 3.5 x 2.75 inches
  • Weight: 1 lbs

Schiit Buf Setup, Tube Use, and Basics

Buf can be added to a system in two straightforward ways. You can place it between a preamp and power amplifier, which allows it to affect every source connected to the preamp. Alternatively, you can insert it between a single source such as a DAC and an integrated amplifier, preamp, or headphone amplifier. In that setup, Buf only influences that specific source.

Tube lifespan is typically around 5,000 hours of use. That figure reflects active operating time, so it’s best to turn the unit off when it’s not in use to extend tube life.

Buf is compatible with tubes that use a standard 6922 pinout, with heater current up to 600mA. Supported types include 6N1P, 6922, ECC88, and 6DJ8. For simplicity and consistency, using the included tube is the most straightforward option.

Like the rest of Schiit’s lineup, Buf is built in the USA, with assembly in Texas and chassis work in California. It carries a 3-year warranty, with the tube itself covered for 90 days.

The Bottom Line

The $99 Schiit Buf is a simple way to add a tube stage to almost any system without changing your core components and you can bypass it when you don’t want it. What’s unique is the price, true tube implementation, and flexibility. What’s missing is everything else: no volume control, no inputs beyond basic RCA switching, no DAC, no remote. This is for users who already have a system and want to experiment with tube character without committing to a full tube preamp or amplifier.

Where to buy: $99 at Schiit

Related Reading:

LG has secured the first True Wireless Lossless Vision certification from testing body TÜV Rheinland for its premium wireless TVs, independently verifying that the OLED evo W6 and Mini RGB evo MRGB9M deliver visually lossless 4K picture quality over a wireless connection.

The certification addresses a longstanding concern with wireless TV technology, where the absence of a physical cable connection has introduced compromises with colour accuracy, HDR tone reproduction and image detail that made wireless displays a harder recommendation against wired equivalents at the same price point.

TÜV Rheinland’s evaluation confirmed that both LG models maintain colour reproduction, image detail, and HDR tone performance within defined tolerance levels relative to the input signal, with the OLED evo W6 additionally certified for wireless transmission at refresh rates up to 165Hz, a figure relevant to gaming use cases where motion clarity and input lag are primary concerns.

The OLED evo W6 sits at the centre of LG’s certified lineup as the company’s Wallpaper TV, a 9mm panel that ships with a separate Zero Connect Box housing all physical inputs, which can be positioned up to 10 metres from the screen to eliminate visible cable runs between source devices and the display itself.

Advertisement

Advertisement

Extending the certification beyond OLED, the Mini RGB evo MRGB9M brings TÜV-verified wireless performance to LCD technology. Available in 65-inch, 75-inch, and 86-inch screen sizes, the panel applies image processing techniques more commonly associated with OLED-class displays to provide better colour accuracy and contrast management at larger scales.

The True Wireless Lossless Vision standard itself was developed by TÜV Rheinland to evaluate wireless display products against factors including input lag, colour accuracy and gamma tracking, establishing a repeatable test framework against which future wireless TV products from any manufacturer can be assessed.

Neither pricing nor regional availability for the LG OLED evo W6 or LG Mini RGB evo MRGB9M has been confirmed, with LG stating only that availability may vary by market.

Advertisement

• SpaceX admits orbital AI data centers may never become commercially viable
• SpaceX S-1 filing reveals unproven technologies behind space-based computing infrastructure
• Harsh space conditions threaten the reliability of sensitive AI hardware systems

SpaceX has warned potential investors its ambitious plans to build AI data centers in orbit may never become commercially viable due to unproven technologies and the harsh realities of space.

The company disclosed these risks in its pre-IPO S-1 filing, which US securities law requires to inform investors of potential pitfalls while shielding the company from future legal liability.