Breaking AI & Tech News Analyzed

A team of researchers from Princeton and Oxford has uncovered a decades‑old handwritten note by Richard Feynman that formulates a mathematical solution to the classic “restaurant‑stopping” problem faced by travelers.Decoding Feynman's Hidden Stopping ProblemThe study, published in the Proceedings of the National Academy of Sciences, reconstructs Feynman's original equation, which advises diners to keep trying new venues until a quality threshold is met. That threshold is not static; it declines more rapidly as the remaining nights in a city decrease, reflecting the diminishing value of future visits to a discovered gem.Feynman's notes were handwritten in the 1970s after a lunch with friend Ralph Leighton.The model assumes a fixed range of restaurant quality and equal probability of encountering any quality level.When the distribution of restaurant quality is uneven, the optimal threshold shifts—higher when few gems exist, lower when most venues are above average.Experimental Findings from 2,520 ParticipantsTo test human behaviour, the authors recruited 2,520 volunteers for an online simulation where participants imagined staying in a city for varying lengths of time and chose restaurants from a grid.Participants’ thresholds fell linearly with the proportion of nights remaining, rather than the rapid decline predicted by Feynman's formula.Despite its simplicity, the linear rule performed comparably to the original solution in the simulated environment.Implications for Decision‑Making and Tourism BehaviourThe findings bridge theoretical optimal‑stopping theory with everyday intuition, suggesting that people naturally adopt a decreasing‑threshold strategy when faced with limited opportunities. This insight could inform:Tourism recommendation engines that adapt suggestions as a trip progresses.Behavioral economics models of consumer search in other domains (e.g., housing, job hunting).Design of AI assistants that balance exploration and exploitation in real‑time.Future Directions for Adaptive Choice ModelsThe authors propose extending the model to dynamic environments where restaurant quality distributions change over time, and to incorporate personal preference heterogeneity. Real‑world field trials in travel apps could validate whether a linear decreasing threshold improves user satisfaction and discovery rates.

The Restructuring Effort Developer platform GitLab has laid off about 14% of its workforce, approximately 350 employees, as part of a broader restructuring effort. The company announced in May that it would reduce its workforce as it exited 22 countries, flattened management layers, and invested in infrastructure to scale its platform and serve increased traffic from AI workflows. Scaling for AI Workloads CEO Bill Staples said during a conference call on Tuesday that agentic workloads are stressing developer infrastructure more than it was designed to handle. GitLab's rival GitHub has also struggled to deal with a massive influx of AI-powered submissions that have affected its uptime. GitLab is partnering with an unspecified AI lab to design and rebuild its infrastructure for AI workloads. The company is constructing APIs optimized for agents to store and retrieve context, including code. GitLab is investing in orchestration tools for coordinating software development between AI agents and developers. Financial Impact GitLab reported first-quarter revenue of $264 million, up 23% from a year earlier, and gross margins of 88%. The company expects to incur $30 million to $35 million in restructuring expenses as part of the effort. Industry Trend GitLab joins a number of tech companies such as Intuit, Amazon, Block, Cisco, Cloudflare, Meta, Microsoft, and Oracle that have laid off large numbers of employees, citing a need to make AI a core part of their business. The tech industry has already cut more than 100,000 jobs this year, per Statista. The Future Outlook The tech industry is seeing a familiar pattern: companies reporting record revenues while simultaneously shrinking their workforces, with AI cited as both the reason for growth and the justification for cuts. GitLab's focus on AI workloads and infrastructure is expected to drive future growth, but at the cost of significant restructuring expenses.

The Renaissance of Inglewood: A City on the Global Stage Inglewood, California, is undergoing a metamorphosis that is redefining its identity from a struggling urban center to a premier global sports destination. With the 2026 FIFA World Cup, the Super Bowl returning to the region, and the 2028 Olympics on the horizon, the city is leveraging billions in investment to position itself as Los Angeles's primary sports hub. However, this rapid transformation is creating a complex narrative of progress and displacement, pitting the glitz of international events against the daily realities of its nearly 103,000 residents. Building the Sports Capital of the Future The centerpiece of this renaissance is the construction of world-class infrastructure, most notably SoFi Stadium, home to the NFL's Rams and Chargers, and the adjacent Intuit Dome. These venues, alongside the remodeled Kia Forum, have turned the city into a focal point for global entertainment. The development extends beyond the stadiums; major streets are being freshly paved, digital billboards are lining the corridors, and the surrounding area—formerly known as Hollywood Park—is being redeveloped into a massive entertainment complex. This physical overhaul is designed to accommodate the influx of international visitors and high-profile events that will soon define the city's calendar. Billions in Investment and a Population Under Pressure The economic scale of this transformation is staggering, with billions of dollars flowing into infrastructure, entertainment development, and commercial real estate. While the city markets itself as the future of sports, the data reveals a stark contrast between the booming venues and the local commercial landscape. Despite the investment, vacant storefronts still punctuate commercial corridors, and essential community assets, such as a closed public school, remain shuttered. This disparity highlights a critical challenge: the rapid pace of development is outstripping the ability of the local economy to absorb the changes, creating a tension between high-profile capital projects and the maintenance of existing community infrastructure. The "Old vs. New" Divide: Gentrification and Displacement The impact of this boom is creating a palpable divide between the "Old Inglewood" and the "New Inglewood." While business owners like Christian Martin of Fiesta Martin Mexican Grill embrace the growth and expansion, long-term residents like Melisa Arnold and Tyler Fister express deep concerns about gentrification. Residents report dealing with the staccato beat of jackhammers, constant street closures, and traffic congestion that makes daily life difficult. The sentiment among some working-class residents is that they are being "walked over" by the development, unable to afford the luxury of attending the very events they helped build. This raises the fundamental question of whether the economic windfall will be equitably distributed or if it will lead to the displacement of the community that calls the city home. Will the Boom Translate to Local Prosperity? The future of Inglewood hinges on the sustainability of this development model. While the short-term economic boost from hosting global events is undeniable, the long-term success depends on the city's ability to integrate the local population into the new economy. Without equitable revenue sharing, affordable housing policies, and community investment, the city risks creating a legacy of prosperity for a select few while leaving the original inhabitants behind. The coming years will determine if Inglewood can successfully transition from a construction site to a thriving, inclusive community that benefits from its status as a world-class sports capital.

The Masterpiece of John Woo The title of this 1990 John Woo extravaganza might lead the uninitiated to expect a chillingly focused, targeted assassination. Actually, there are innumerable bullets and innumerable heads in this over-the-top gonzo spectacle. It is a crime thriller, a wartime action film set in Vietnam, but it offers something other than the usual Hollywood perspective; it is a parable of greed comparable to The Treasure of the Sierra Madre, and even a kind of romantic melodrama. The Symbolic Bullet There is, however, one key bullet in a head, a literal bullet lodged in the skull of someone who achieves a macabre zombie-like semi-survival, the bullet being symbolic of the way violence takes root in the brain, dehumanising its victim. The final “boardroom” scene disclosing this image is toweringly mad and strange. Yet in this movie, as in so many other Woo films, we can see how the director counterintuitively uses sad music – harmonica, woodwind – over grisly, brutal action sequences, as if what he wants us to register is not the violence or the shock but just how poignantly futile and pathetic it all is. The Plot Unfolds The setting is – initially – late 60s Hong Kong; Tony Leung plays Ben, a young guy getting married to his sweetheart Jane (Fennie Yuen), and on hand are his buddies Frank (Jacky Cheung) and Paul (Waise Lee). This trio are involved in a gang war with a rival mob who corner Frank when he has gone to get cash from the local moneylender to pay for the wedding. The confrontation ends in violence and, simply to get away and avoid the heat – Woo has a cameo as a police inspector – the three guys accept a crooked job from a local gang boss smuggling contraband pharmaceuticals in Vietnam. The Climax In Saigon, all their plans explode in pure anarchy; they are at first arrested under suspicion of working for the Vietcong, then upgrade their strategy to cynical arms smuggling for the North Vietnamese, and stealing the gold belonging to the local wiseguy who was supposed to be distributing their drugs. They make contact with a worldly fixer called Luke (Simon Yam), who has a Catherine Deneuve poster in his apartment, and also gallantly undertake to rescue a Hong Kong singer Sally (Yolinda Yam), who has been trafficked to Vietnam to sell sex. They end up on the spectacular field of battle itself (with helicopters, explosions, burning villages), where Paul, increasingly obsessed with the gold, finds himself at a key moment at mortal risk of detection if he cannot keep the wounded Frank quiet, whose cries of pain might alert the enemy to their position. The Legacy It’s an extraordinary, uninhibited barnstormer from Woo, who takes us from regular, domestic crime to military chaos with confidence and sweep. One moment, in which someone stands in front of a tank, even appears to echo the Tiananmen Square protests. No one else could have made it. The Release Bullet in the Head is in UK cinemas from 1 June, and on 4K UHD and Blu-ray from 22 June.

The Rise of Recursive Self-Improvement in AIThe word "recursion" is the latest buzzword in AI circles. Two separate startups have taken on the name, and many more have started referencing recursive self-improvement (RSI) in their roadmaps. Like AGI before it, RSI has become a three-letter byword for a cataclysmic AI takeoff – even if there's still a little disagreement about what it exactly means.In basic terms, RSI refers to an AI system that can continuously upgrade itself. Once AI systems can manage the upgrade cycle better than humans, the process can become a closed loop, limited only by the compute power they can access, and humans are no longer necessary or even helpful.Scary or not, that's a vision that a lot of AI labs are eager to chase.Key Players Pursuing Recursive SystemsEarlier this month, well-known AI researcher Richard Socher launched the aptly named Recursive Superintelligence with RSI as an explicit goal. "Our main focus is to build truly recursive, self-improving superintelligence at scale," Socher told TechCrunch at launch, "which means that the entire process of ideation, implementation, and validation of research ideas would be automatic."A number of other prominent researchers are already chasing that same goal, hoping for a breakthrough that will make recursive self-improvement possible.One of the most prominent is Andrej Karpathy, a legendary figure from Tesla and OpenAI, who is using agent swarms to train LLMs on simple tasks for a project he calls Auto-Research. Karpathy has been unusually open about the project, tweeting about milestones regularly and making the building blocks available through a public GitHub repo. So far, the work has mostly been confined to making minor improvements on a GPT-2 scale model — as Karpathy noted in March, "It's not novel, ground-breaking 'research' (yet)" — but it's been enough to convince lots of other researchers to follow the RSI dream. And with Karpathy now working on pre-training at Anthropic, he will have plenty of opportunity to apply the idea at a larger scale.Adaption — founded by Cohere and Google alum Sara Hooker — recently launched a similar tool called AutoScientist in an effort to automate frontier training. Like Karpathy's auto-researchers, the system trains agents to make incremental improvements — but for Adaption, the goal is to make it easier to train a full-scale frontier model. If those same researchers start to push the frontier forward, the system could quickly spiral into something very much like RSI.Disarray founder Doris Xin drew more specific RSI interest when her self-trained machine learning agent took home 28 medals in a recent Kaggle competition, beating out many human-trained agents. As she sees it, the major challenge is reliability."I would argue, given infinite compute and infinite time horizon, we are already there," Xin told me. "I want to make an argument that this is not a creative endeavor, really. It's just a lot of meat-and-potatoes engineering."The Current State of Self-Improving AIThere's also plenty of evidence that the AI industry isn't very close to recursive systems in any meaningful way — and is still grappling with talking to a wary public about its progress. So Google CEO Sundar Pichai basically admitted in a recent podcast interview."It's a continuum, and we are all definitely making progress," Pichai said. "But in the way people describe RSI, that would represent a next level of acceleration and would have a lot of implications, but we aren't quite there yet."But the continuum includes an awful lot of self-improving AI systems.In January, one of Anthropic's lead programmers for Claude Code estimated that "close to 100%" of his team's code was written by the tool — a frank admission that Claude Code was literally writing itself.Just because engineers are using an AI tool doesn't mean the tool can replace them — but Anthropic seems to be getting close to replacing engineers too. In a recent survey tied to the Mythos preview, five out of 18 Anthropic engineers believed that, with harness improvements, this version of Mythos could soon substitute for an L4 engineer — a midlevel programmer who can take on involved projects without supervision.Still, there were some of the same weaknesses you might expect."Some of Claude's major reported weaknesses compared to an L4 include: self-managing week-long ambiguous tasks, understanding org priorities, taste, verification, instruction-following, and epistemics," the report reads.In other words, its weaknesses are everything involved with self-direction, which is the cornerstone for RSI. But sure, for everything else, Claude is ready to step right in.Expert Perspectives on RSI TimelinesJust like the AGI term before it, the AI industry also can't tell us how far away it is from showcasing a meaningful recursive system. When Georgetown's Center for Security and Emerging Technology assembled a group of experts to study RSI last year, the group found a major split in assessments — some expecting an imminent "superintelligence" style explosion while others expected slower progress and an eventual plateau. But all agreed that recursion made the future especially difficult to predict.Helen Toner, director of CSET and a former board member at OpenAI, told TechCrunch that simply using AI tools to do AI research isn't enough to qualify as RSI. "They're just using AI for as much as they can," Toner told TechCrunch. "And I think that is different from the classic definition of RSI, which is really that there are no humans needed."Toner pointed to a recent post by METR's Ajeya Cotra, which distinguishes different milestones on the path to the AI research takeover. One step, which Cotra calls "adequacy," would come when the system can still perform research after all humans are removed — even if the resulting research isn't as valuable or efficient. "Parity" comes when an AI-only system is as good at research as a human-only system. "Supremacy," the final stage, comes when an AI-only system outperforms a collaborative system between humans and AI.Ultimately, Cotra concludes that AI is very close to the adequacy threshold of being able to produce some work on its own — similar to the incremental changes made by Karpathy's Auto-Research system. "I wouldn't be totally shocked if you told me this milestone had already passed, and I expect it to happen in the next couple years," Cotra wrote.She was less clear on when parity will come, but once it does, she thinks it would "massively accelerate the pace of AI progress, leading to AI research supremacy within another year."The Challenges Ahead for Recursive AIWith so much of AI built on scaling laws, there's a strong tendency to think RSI will follow the same curve. Toner thinks that many of those pursuing AI research and development via RSI "think of it as a pretty smooth ladder, where you can just keep scaling up."But even if AI researchers are able to make incremental improvements like Karpathy's auto-researchers, there will be larger challenges in handing off the whole process of research. Toner put it in terms of the history of computing, which has seen human beings handing off more and more of the process while still directing things from the top."We went from machine languages to assembly language and compiled languages; you're getting further and further from the guts of the computer," Toner said. "But the human is still, in some intuitive sense, running the show."Moving beyond that paradigm will take significant challenges, both in engineering and alignment. But even with the massive investments happening, there's no infinite compute available — and the basic trade-off between human labor and machine intelligence will be hard to overcome.The Future of Recursive Self-ImprovementAs for a total recursive AI system of apocalyptic visions? The only thing researchers essentially agree on is that, like AGI, it's not here yet.

The Challenge: Designing Escape RoutesThe puzzle presents a scenario with two snakes of equal width but different lengths trapped in a cage. The objective is to design two distinct escape passages, A and B, that allow one snake to pass while blocking the other.Passage A: Must allow the short snake to escape but block the long snake.Passage B: Must allow the long snake to escape but block the short snake.The Logic of the SolutionThe solution relies on exploiting the physical dimensions of the snakes. For Passage A, the design features a loop that is longer than the short snake but shorter than the long one. When the long snake enters the loop and doubles back, its body blocks the exit point, trapping it. The short snake, being shorter, can navigate the loop without obstruction.Passage B utilizes a floor hole. Assuming the snakes have non-zero rigidity, the short snake cannot stretch far enough to move over the hole without falling in, whereas the long snake can bridge the gap and pass safely.Why Spatial Reasoning MattersThis puzzle underscores the critical role of spatial intelligence in problem-solving. It demonstrates how understanding the relationship between length, width, and path constraints can create solutions that are counter-intuitive yet logically sound.The Future of Logic Puzzles in AIAs AI models continue to advance in spatial reasoning, puzzles like this will likely serve as benchmarks for testing the flexibility of machine intelligence. The future of puzzle design may shift towards scenarios that require not just calculation, but a nuanced understanding of physical constraints.

Rapid Growth Trajectory Database provider ClickHouse has crossed $250 million in annualized revenue run rate, tripling its business from last year, Yury Izrailevsky, co-founder and president of product and technology, told TechCrunch. Izrailevsky expects the revenue figure to reach the high-nine digits by the end of the year. Valuation and Funding ClickHouse was valued at $15 billion in January following a $400 million Series D funding round led by Dragoneer Investment Group. The latest valuation implies a steep multiple of over 60x annualized revenue. IPO Ambitions The fast revenue growth and premium valuation position the less-than-five-year-old company for an IPO within the next few years, according to Izrailevsky. ClickHouse joins a small but growing list of tech startups signaling plans to go public as the IPO window is expected to be flung wide open by SpaceX’s historic June debut, followed by highly anticipated listings from OpenAI and Anthropic later this year. Strategic Moves Last fall, the startup hired Jimmy Sexton, who previously ran investor relations at Snowflake, one of ClickHouse’s main competitors, as chief financial officer. Bringing on a CFO is often viewed as a signal that a company is preparing for public markets. Acquisition Strategy The company has already acquired six startups, including Langfuse, which helps developers track and evaluate AI agent performance. Izrailevsky indicated that ClickHouse plans to remain acquisitive, looking to scoop up “relatively young, but showing very promising technology” startups, typically open source, that complement its core product suite. Product and Customer Base The technology behind ClickHouse was originally developed inside Russian search giant Yandex 17 years ago, but spun out as an independent startup in 2021. ClickHouse has over 4,000 customers, including Anthropic, Meta, Capital One, and Decagon. Business Model The startup’s open source database is designed to process the massive datasets required by AI agents. ClickHouse generates revenue by selling managed cloud services. Izrailevsky claimed that this commercial offering ultimately costs clients less than self-managing the open source version. It “is something that’s a little counterintuitive, but it also has been a big tailwind for us,” he said.

OpenAI has reported a major advance in AI reasoning after its model successfully challenged an 80‑year‑old conjecture in discrete geometry, the planar unit distance problem first posed by Paul Erdős in 1946.OpenAI’s Model Cracks the 80‑Year‑Old Planar Unit Distance ConjectureThe conjecture suggested that the number of equal‑distance dot pairs on a plane grows only slightly faster than the number of dots.OpenAI's reasoning system generated a family of point arrangements that exceed Erdős’s proposed limit.The result was announced on X and confirmed in a companion paper co‑authored by mathematician Thomas Bloom.Quantifying the Breakthrough: No Monetary Figures, but Scientific SignificanceWhile the article provides no financial data, the achievement is described as a “milestone in AI mathematics” by Tim Gowers.The validation by experts underscores the credibility of AI‑generated proofs, contrasting with a prior, unverified claim from last year.Implications for AI‑Driven Mathematical ResearchThe model’s ability to explore unconventional solution paths highlights AI’s potential to augment human intuition.Researchers, including Andrew Rogoyski, note that AI is becoming a fundamental tool for future scientific inquiry.The breakthrough may accelerate AI involvement in other open problems across mathematics.What the Next Steps Could Mean for AI and MathematicsFurther collaboration between AI systems and mathematicians is expected to refine the new constructions and explore their consequences.OpenAI’s upcoming IPO could bring additional resources to expand its reasoning capabilities.The community anticipates more AI‑driven insights that could eventually resolve the broader Erdős problems.

The LeadSpotify has introduced a new AI-powered audiobook creation tool in partnership with ElevenLabs, allowing authors to self-publish audiobooks without exclusivity. The platform is expanding to support 10 more languages and aims to generate $100 million in annualized recurring revenue from its Audiobook+ subscriptions.AI Audiobook Creation Platform LaunchAlongside tools for AI-generated podcasts, Spotify on Thursday introduced a new, ElevenLabs-powered AI tool for self-publishing audiobooks within the Spotify for Authors platform. The company said at its Investor Day event that the feature will launch in beta this June on an invite-only basis, initially with support for the English language only.The AI-powered audiobook generation won't bind authors to an exclusive contract, meaning they are free to publish their generated audiobooks anywhere. This approach contrasts with some other platforms that require exclusivity for audiobook distribution.The news builds on Spotify's previous partnership with ElevenLabs, which allowed writers to submit audiobooks created on the voice AI startup's platform to Spotify. The audio streaming platform also already had a partnership with Google Play Books to allow for digitally narrated content. However, it may have wanted authors to access newer voice models that sound more expressive and human-like, like those offered by ElevenLabs. Notably, ElevenLabs had released its own self-publishing platform for authors in 2025.Financial Performance and Growth MetricsSpotify has increased its focus on audiobooks heavily in the last few years and has managed to build its catalog to 700,000 titles. Through these initiatives, the company has managed to bump up listening hours by 60% year-on-year, the company claims. Spotify also said that more than half of its audiobook listeners started in the last year.To date, Spotify has clocked in over a million Audiobook+ subscriptions, and it is on track to generate $100 million in annualized recurring revenue for the platform. The company will expand its Audiobook+ plans this year to allow for higher listening limits and will add new options for students and families in the future.Industry Transformation and Market ExpansionSpotify is also expanding its "Spotify for Authors" platform to support 10 more languages, including French, Canadian French, German, Dutch, Latin American Spanish, Swedish, Finnish, Icelandic, Danish, and Norwegian. This expansion will significantly broaden the platform's reach and accessibility to authors and listeners worldwide.The company brought the program to international markets, made an investment in non-English titles, enabled in-app purchases, and released audiobook charts. This year, it also started a program for authors to sell physical books in the U.S. and the U.K., creating a comprehensive ecosystem for content creators.Future Outlook and User Experience EnhancementsAt the event, the company introduced a new way for users to ask questions using natural language for audiobook discovery. This summer, Spotify will also expand a feature that allows users to create prompt-based playlists for podcasts and music to include audiobooks, it said.These enhancements reflect Spotify's strategy to leverage AI not just for content creation but also for improving user discovery and engagement. The integration of natural language processing for audiobook discovery could potentially revolutionize how users find and consume audiobooks, making the platform more intuitive and user-friendly.