Meanwhile, a Los Angeles jury is deliberating a social media addiction lawsuit against Meta and Google.

A New Mexico jury has found that Meta endangered children by misleading users about the safety of its platforms. The decision comes after a nearly seven-week trial, resulting in Meta being told to pay $375m in damages.

“The jury’s verdict is a historic victory for every child and family who has paid the price for Meta’s choice to put profits over kids’ safety,” said New Mexico attorney general Raúl Torrez, who filed the lawsuit against Meta in 2023.

In the suit, he claimed Meta knowingly exposes children to sexual exploitation and mental harm for profit. Meta owns Instagram, Facebook and WhatsApp.

According to the New Mexico Department of Justice, evidence presented at the trial established that Meta’s design features enable bad actors to engage in child sexual exploitation. Evidence also showed platforms are also designed to addict young people, according to the department.

“Meta executives knew their products harmed children, disregarded warnings from their own employees, and lied to the public about what they knew. Today the jury joined families, educators, and child safety experts in saying enough is enough,” Torrez added. The company has been found to have violated parts of the state’s Unfair Practices Act.

Meta plans to appeal the decision. “We respectfully disagree with the verdict and will appeal,” Meta said in statement yesterday (24 March).

“We work hard to keep people safe on our platforms and are clear about the challenges of identifying and removing bad actors or harmful content. We will continue to defend ourselves vigorously, and we remain confident in our record of protecting teens online.”

Meanwhile Torrez will be asking the presiding judge to place additional penalties on Meta during a bench trial scheduled in early May. Torrez will also be asking the court to force Meta to make its apps safer for children.

The New Mexico verdict is a major loss for Meta, which is gearing up for a number of trials set for this year. A jury in Los Angeles is currently deliberating a social media addiction suit against Meta and Google. TikTok and Snapchat were involved in the original suit, but have since settled out of court.

Thousands have filed lawsuits against social media companies over the alleged harm they pose to their users, including more than 40 US state attorney generals.

A coalition suit filed in 2023 accused Meta of designing and deploying “harmful features” on Instagram and Facebook, which get younger people addicted to these platforms.

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.

‘Wispit 2C’ is estimated to be about 5m years old and most likely 10 times the mass of Jupiter.

Galway native Chloe Lawlor has discovered a new planet – the second one to be found forming near an infant star called ‘Wispit 2’, some 437 light years away.

As a child, Lawlor wanted to be an artist, she tells SiliconRepublic.com. However, she changed her mind once she joined university. “I moved into physics because I did like physics in school, so I thought, ‘Oh, maybe I’ll just try this out.’”

The 25-year-old says discoveries such as these feed the innate curiosity humans have in wanting to know how we came to be, how we evolved and why we are here. Lawlor is a PhD student at the University of Galway’s Centre for Astronomy at the School of Natural Sciences and the Ryan Institute.

She is working in collaboration with project lead Richelle van Capelleveen, a PhD student from Leiden Observatory in the Netherlands, and postdoctoral researcher Guillaume Bourdarot from the Max Planck Institute for Extraterrestrial Physics in Germany, to learn more about young planets and how they’re forming.

“Most of the planets that we’ve observed have been much older,” Lawlor says. “We don’t know how they get to those sort of final stages like something like our solar system. This is really key for these formation theories and it’s hopefully going to tell us a lot about these young systems, how they’re forming, and then how they evolve.”

Lawlor’s new discovery, an exoplanet named ‘Wispit 2C’, is thought to be about 5m years old. ‘Wispit 2B’, a nearby planet, was discovered last year by van Capelleveen and Dr Laird Close from the University of Arizona.

Both these exoplanets are at early stages of formation in the disc around Wispit 2, which is located in the Constellation of the Eagle, a prominent equatorial constellation visible in the northern hemisphere summer along the Milky Way.

Lawlor’s discovery makes Wispit 2 the second known young and still forming multi-planet system. The only other system yet discovered with more than one planet developing is PDS 70, some 400 light years away from Earth.

Wispit 2C is a gas giant, likely around 10 times the mass of Jupiter. It is twice as massive as Wispit 2B and orbits four-times closer to its host star, which makes it incredibly difficult to detect with ground-based telescopes.

Wispit 2B and Wispit 2C forming around Wispit 2. Image: ESO/C Lawlor, R F van Capelleveen et al.

The exoplanet was detected using the European Southern Observatory’s Very Large Telescope in Chile’s Atacama desert. By linking several telescopes together to act as one giant instrument, the research team was able to observe regions very close to the star. In their analysis, the team was able to detect carbon monoxide gas, a chemical commonly found in the atmospheres of young giant planets.

Lawlor said earlier this week: “At first, we weren’t sure if it was a planet or a very large dust clump. We very quickly made follow-up observations using the Very Large Telescope Interferometer, an incredible setup where multiple telescopes can be connected to form a large virtual telescope.

“This allowed us to take what we call a spectrum, which is essentially a chemical fingerprint revealing the elements and molecules in an object’s atmosphere.”

Lawlor led the study, which has been published in The Astrophysical Journal Letters.

Prof Frances Fahy, director of the Ryan Institute, said: “The discovery of the planet Wispit 2C is a remarkable achievement and highlights the world-class astrophysics research taking place at University of Galway.”

The team will continue on with their efforts to hopefully find more planets in the system.

Last year, a study from Scotland’s University of St Andrews showed how giant free-floating planets could make their own miniature planetary systems without needing a star to orbit around. In a different study from 2025, scientists – for the first time – observed the very early stages of the creation of a new solar system around a baby star.

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.

Go ahead and get ready to cue your inevitable comparisons: the new trailer for HBO’s Harry Potter series dropped on Wednesday, giving audiences a first look at muggles, magical Hogwarts students and The Boy Who Lived. Due to hit HBO and HBO Max for Christmas 2026, the TV show will be a direct adaptation of the wizarding books, starting off with The Philosopher’s Stone.

To fans familiar with the movie franchise, this may feel like a rediscovery — or reintroduction — to the live-action version of the world of Harry Potter. The trailer shows a young Harry and his signature scar, his tyrant of an aunt and the moment he received his invitation from Hogwarts. Take a look at the first meet with Hagrid, tender moments with Ron and Hermione, and a look at Lucius Malfoy and Snape. 

The series features Dominic McLaughlin as Harry Potter, Arabella Stanton as Hermione Granger and Alastair Stout 
as Ron Weasley, and the expansive cast also includes Nick Frost as Hagrid, John Lithgow as Hogwarts headmaster Dumbledore and Paapa Essiedu as Professor Snape. 

Nintendo might finally be doing something gamers have been asking for… forever. The company has officially confirmed that Switch 2 games will have different pricing for digital and physical versions, with digital copies expected to be cheaper.

The change begins in May 2026, starting with titles like Yoshi and the Mysterious Book. For example, early listings on the eShop show the game priced at $59.99 digitally vs $69.99 physically, marking a clear shift in how Nintendo handles game pricing.

Why is Nintendo doing this?

Let’s be real, physical games are comparatively expensive to make. Nintendo says the change reflects the higher costs of manufacturing and distributing cartridges, compared to digital downloads. This aligns with what the industry has been doing for years, except Nintendo has been one of the few holdouts where digital and physical games often cost the same.

There’s also a bigger strategy at play here. By making digital games cheaper, Nintendo could nudge more players toward digital purchases. That essentially translates to higher margins, fewer logistics headaches, and a tighter grip on its ecosystem. In other words, this isn’t just about fairness in pricing… It’s also about where Nintendo wants its future sales to go.

What does this mean for players?

So, does this mean all games going forward will have different prices? Well, not exactly, and this is where things get a little messy. While Nintendo is setting a lower MSRP for digital games, actual pricing can still vary depending on the title and retailer. Plus, not every game will follow the same pattern, so bigger releases could still carry higher price tags, making the gap between digital and physical a bit inconsistent.

For players, though, this is still a win. Digital games are finally getting a clear pricing advantage after years of being oddly equal (or sometimes pricier) than physical copies. That said, the trade-off remains. Physical games can be resold or shared, while digital ones stay locked to your account.

Standard RAG pipelines break when enterprises try to use them for long-term, multi-session LLM agent deployments. This is a critical limitation as demand for persistent AI assistants grows.

xMemory, a new technique developed by researchers at King’s College London and The Alan Turing Institute, solves this by organizing conversations into a searchable hierarchy of semantic themes.

Experiments show that xMemory improves answer quality and long-range reasoning across various LLMs while cutting inference costs. According to the researchers, it drops token usage from over 9,000 to roughly 4,700 tokens per query compared to existing systems on some tasks.

For real-world enterprise applications like personalized AI assistants and multi-session decision support tools, this means organizations can deploy more reliable, context-aware agents capable of maintaining coherent long-term memory without blowing up computational expenses.

RAG wasn’t built for this

In many enterprise LLM applications, a critical expectation is that these systems will maintain coherence and personalization across long, multi-session interactions. To support this long-term reasoning, one common approach is to use standard RAG: store past dialogues and events, retrieve a fixed number of top matches based on embedding similarity, and concatenate them into a context window to generate answers.

However, traditional RAG is built for large databases where the retrieved documents are highly diverse. The main challenge is filtering out entirely irrelevant information. An AI agent’s memory, by contrast, is a bounded and continuous stream of conversation, meaning the stored data chunks are highly correlated and frequently contain near-duplicates.

To understand why simply increasing the context window doesn’t work, consider how standard RAG handles a concept like citrus fruit.

Imagine a user has had many conversations saying things like “I love oranges,” “I like mandarins,” and separately, other conversations about what counts as a citrus fruit. Traditional RAG may treat all of these as semantically close and keep retrieving similar “citrus-like” snippets. 

“If retrieval collapses onto whichever cluster is densest in embedding space, the agent may get many highly similar passages about preference, while missing the category facts needed to answer the actual query,” Lin Gui, co-author of the paper, told VentureBeat. 

A common fix for engineering teams is to apply post-retrieval pruning or compression to filter out the noise. These methods assume that the retrieved passages are highly diverse and that irrelevant noise patterns can be cleanly separated from useful facts.

This approach falls short in conversational agent memory because human dialogue is “temporally entangled,” the researchers write. Conversational memory relies heavily on co-references, ellipsis, and strict timeline dependencies. Because of this interconnectedness, traditional pruning tools often accidentally delete important bits of a conversation, leaving the AI without vital context needed to reason accurately.

Why the fix most teams reach for makes things worse

To overcome these limitations, the researchers propose a shift in how agent memory is built and searched, which they describe as “decoupling to aggregation.”

Instead of matching user queries directly against raw, overlapping chat logs, the system organizes the conversation into a hierarchical structure. First it decouples the conversation stream into distinct, standalone semantic components. These individual facts are then aggregated into a higher-level structural hierarchy of themes.

When the AI needs to recall information, it searches top-down through the hierarchy, going from themes to semantics and finally to raw snippets. This approach avoids redundancy. If two dialogue snippets have similar embeddings, the system is unlikely to retrieve them together if they have been assigned to different semantic components.

For this architecture to succeed, it must balance two vital structural properties. The semantic components must be sufficiently differentiated to prevent the AI from retrieving redundant data. At the same time, the higher-level aggregations must remain semantically faithful to the original context to ensure the model can craft accurate answers.

A four-level hierarchy that shrinks the context window

The researchers developed xMemory, a framework that combines structured memory management with an adaptive, top-down search strategy.

xMemory continuously organizes the raw stream of conversation into a structured, four-level hierarchy. At the base are the raw messages, which are first summarized into contiguous blocks called “episodes.” From these episodes, the system distills reusable facts as semantics that disentangle the core, long-term knowledge from repetitive chat logs. Finally, related semantics are grouped together into high-level themes to make them easily searchable.

xMemory uses a special objective function to constantly optimize how it groups these items. This prevents categories from becoming too bloated, which slows down search, or too fragmented, which weakens the model’s ability to aggregate evidence and answer questions.

When it receives a prompt, xMemory performs a top-down retrieval across this hierarchy. It starts at the theme and semantic levels, selecting a diverse, compact set of relevant facts. This is crucial for real-world applications where user queries often require gathering descriptions across multiple topics or chaining connected facts together for complex, multi-hop reasoning.

Once it has this high-level skeleton of facts, the system controls redundancy through what the researchers call “Uncertainty Gating.” It only drills down to pull the finer, raw evidence at the episode or message level if that specific detail measurably decreases the model’s uncertainty.

“Semantic similarity is a candidate-generation signal; uncertainty is a decision signal,” Gui said. “Similarity tells you what is nearby. Uncertainty tells you what is actually worth paying for in the prompt budget.” It stops expanding when it detects that adding more detail no longer helps answer the question.

What are the alternatives?

Existing agent memory systems generally fall into two structural categories: flat designs and structured designs. Both suffer from fundamental limitations.

Flat approaches such as MemGPT log raw dialogue or minimally processed traces. This captures the conversation but accumulates massive redundancy and increases retrieval costs as the history grows longer.

Structured systems such as A-MEM and MemoryOS try to solve this by organizing memories into hierarchies or graphs. However, they still rely on raw or minimally processed text as their primary retrieval unit, often pulling in extensive, bloated contexts. These systems also depend heavily on LLM-generated memory records that have strict schema constraints. If the AI deviates slightly in its formatting, it can cause memory failure.

xMemory addresses these limitations through its optimized memory construction scheme, hierarchical retrieval, and dynamic restructuring of its memory as it grows larger.

When to use xMemory

For enterprise architects, knowing when to adopt this architecture over standard RAG is critical. According to Gui, “xMemory is most compelling where the system needs to stay coherent across weeks or months of interaction.”

Customer support agents, for instance, benefit greatly from this approach because they must remember stable user preferences, past incidents, and account-specific context without repeatedly pulling up near-duplicate support tickets. Personalized coaching is another ideal use case, requiring the AI to separate enduring user traits from episodic, day-to-day details.

Conversely, if an enterprise is building an AI to chat with a repository of files, such as policy manuals or technical documentation, “a simpler RAG stack is still the better engineering choice,” Gui said. In those static, document-centric scenarios, the corpus is diverse enough that standard nearest-neighbor retrieval works perfectly well without the operational overhead of hierarchical memory.

The write tax is worth it

xMemory cuts the latency bottleneck associated with the LLM’s final answer generation. In standard RAG systems, the LLM is forced to read and process a bloated context window full of redundant dialogue. Because xMemory’s precise, top-down retrieval builds a much smaller, highly targeted context window, the reader LLM spends far less compute time analyzing the prompt and generating the final output.

In their experiments on long-context tasks, both open and closed models equipped with xMemory outperformed other baselines, using considerably fewer tokens while increasing task accuracy.

However, this efficient retrieval comes with an upfront cost. For an enterprise deployment, the catch with xMemory is that it trades a massive read tax for an upfront write tax. While it ultimately makes answering user queries faster and cheaper, maintaining its sophisticated architecture requires substantial background processing.

Unlike standard RAG pipelines, which cheaply dump raw text embeddings into a database, xMemory must execute multiple auxiliary LLM calls to detect conversation boundaries, summarize episodes, extract long-term semantic facts, and synthesize overarching themes.

Furthermore, xMemory’s restructuring process adds additional computational requirements as the AI must curate, link, and update its own internal filing system. To manage this operational complexity in production, teams can execute this heavy restructuring asynchronously or in micro-batches rather than synchronously blocking the user’s query.

For developers eager to prototype, the xMemory code is publicly available on GitHub under an MIT license, making it viable for commercial uses. If you are trying to implement this in existing orchestration tools like LangChain, Gui advises focusing on the core innovation first: “The most important thing to build first is not a fancier retriever prompt. It is the memory decomposition layer. If you get only one thing right first, make it the indexing and decomposition logic.”

Retrieval isn’t the last bottleneck

While xMemory offers a powerful solution to today’s context-window limitations, it clears the path for the next generation of challenges in agentic workflows. As AI agents collaborate over longer horizons, simply finding the right information won’t be enough.

“Retrieval is a bottleneck, but once retrieval improves, these systems quickly run into lifecycle management and memory governance as the next bottlenecks,” Gui said. Navigating how data should decay, handling user privacy, and maintaining shared memory across multiple agents is exactly “where I expect a lot of the next wave of work to happen,” he said.