Breaking AI & Tech News Analyzed

The 24/7 Agentic Assistant Breakthrough Google has introduced Gemini Spark, a 24/7 agentic assistant designed to help users navigate their digital lives autonomously. Unlike traditional chatbots that require local hardware to stay active, Spark runs on virtual machines in the cloud, allowing users to close their laptops while tasks are being completed. The service is deeply integrated into the Google Workspace ecosystem, connecting with Gmail, Calendar, Docs, Sheets, and Slides to handle work-adjacent tasks. Cloud-Native Architecture: Spark operates continuously without the need for the user's device to be awake. Work-Adjacent Focus: It is optimized for tasks that bridge the gap between manual labor and automation, such as summarizing inboxes or organizing spreadsheets. CEO Endorsement: Sundar Pichai positioned Spark as an accessible entry point into agentic AI, contrasting it with more complex systems that require constant user oversight. Real-World Performance Metrics Testing the assistant revealed a mix of high-utility features and frustrating limitations. While Spark excelled at complex research and aggregation, it struggled with specific execution details and integrations. Shopping Research: Spark successfully identified weekly deals and suggested coupon stacking strategies. However, it failed to validate a specific promo code, requiring manual intervention. Packing Lists: The AI provided highly accurate suggestions for a day trip, including weather-appropriate items and event restrictions. However, it failed to export the list to Google Keep, instead offering to create a document or email—a significant usability oversight. Event Discovery: Spark successfully aggregated local events from multiple sources, identifying niche opportunities like the 'Annual Beaver Queen Pageant' that would be missed by manual searching. Newsletter Summaries: The assistant generated summaries with context but missed one requested article and suffered from link redirection issues. The Ecosystem Lock-In Challenge The primary barrier to Spark's adoption is its heavy reliance on the Google ecosystem, creating a 'walled garden' effect that limits its utility outside of Google services. The lack of integration with Google Keep is a major usability gap, as the notetaking app is essential for personal productivity lists. Furthermore, the confusion surrounding its branding—separate from the main Gemini chatbot interface—adds unnecessary cognitive load for users trying to distinguish between 'questions' and 'tasks.' Platform Limitations: The tool cannot be accessed via iPhone hardware buttons, requiring users to manually launch the app. Integration Gaps: Current limitations in MCP (Model Context Protocol) integrations prevent Spark from booking external services like restaurants or flights. Branding Confusion: The industry is saturated with AI names, and Spark's standalone toggle adds to the mental load rather than simplifying it. The Future of Standalone AI Toggles Google's experiment with Spark suggests that standalone AI products may struggle to justify their existence in a crowded market. The future of AI assistants lies in unified interfaces where functionality is integrated seamlessly rather than separated by confusing toggles. For Spark to become a 'must-have,' Google must address the lack of cross-platform accessibility and expand its integration capabilities beyond the Google universe.

The Inevitable Integration of AI in DevelopmentIn 2026, artificial intelligence has become an inseparable tool for developers, yet this reliance may be masking a critical productivity crisis.Researchers at METR discovered that most developers will not participate in studies without AI assistance.This dependency suggests a psychological shift where AI is no longer viewed as an assistant but a requirement.The "Tokenmaxxing" Crisis and Budget BlowoutsThe trend of measuring productivity by token usage, known as "tokenmaxxing," has led to significant financial waste.Amazon shut down its internal leaderboard, Kirorank, after employees gamed the system to run up costs.Uber reportedly exhausted its 2026 AI budget in just four months without measurable project increases.Self-reported data shows a 2x increase in perceived value, but independent analysis suggests 44% of tokens are spent fixing bugs generated by AI.Code review tools indicate AI produces 1.7x more problems than human code.The Hidden Cost of Speed: Maintenance and QualityWhile AI generates code faster, it introduces long-term maintenance costs that developers are currently ignoring.Programmer James Shore warns that trading a temporary speed boost for permanent indenture is a dangerous strategy.Researchers from Singapore Management University have confirmed that AI-generated code can introduce significant long-term maintenance burdens.The Future of Human-AI CollaborationThe industry is moving toward a model where AI is a junior developer that requires constant oversight.Scott Wu (Cognition) admits his AI agent Devin is currently a junior-to-mid-level programmer.Experts recommend that humans must review AI work as carefully as they would a junior developer's code.Software architecture and security design must remain human-centric tasks.

Groq's New Funding Round Groq is looking to raise $650 million in new funding from existing investors, sources tell Axios, as it leans into its inference neocloud business that relies on its homegrown AI chip and systems. The Nvidia Deal and Its Impact In December, Groq struck one of those not-an-acquisition agreements with Nvidia for a reported $20 billion, which involved the departure of some top-level senior Groq employees to the chip giant and the licensing of Groq’s hardware technology to Nvidia. The Focus on Inference Cloud Business The new direction is led right now by Groq’s interim CEO and CFO, Adam Winter and Matt Eng, respectively. The company's inference cloud business lets developers and enterprises host their inference-hungry apps. Inference is the processing that happens after an AI prompt and is currently a much bigger need in the AI world than model training. The Funding Commitment Groq's backers Disruptive and Infinitium have agreed to fill the round should other existing investors not want their pro-rata shares. The $650 million in funding is essentially guaranteed.

Groq's Ambitious Funding Round Groq, an AI chip startup, is looking to raise $650 million in new funding from existing investors, sources tell Axios, as it leans into its inference neocloud business that relies on its homegrown AI chip and systems. The Nvidia Deal and Its Implications In December, Groq struck a not-an-acquisition agreement with Nvidia for a reported $20 billion, which involved the departure of some top-level senior Groq employees to the chip giant and the licensing of Groq's hardware technology to Nvidia. The Focus on Inference Cloud Business The new direction is led by Groq's interim CEO and CFO, Adam Winter and Matt Eng, respectively. The company's inference cloud business lets developers and enterprises host their inference-hungry apps. Inference is the processing that happens after an AI prompt and is currently a much bigger need in the AI world than model training. The Funding Dynamics Groq's backers Disruptive and Infinitium have agreed to fill the round should other existing investors not want their pro-rata shares. The $650 million in funding is essentially guaranteed. The funding round highlights the ongoing investments in AI chip startups and the growing demand for inference capabilities in the AI ecosystem.

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 Massive Deal Cloud data storage giant Snowflake has signed a new $6 billion five-year agreement with Amazon Web Services, the companies announced on Wednesday. This deal is significant, as Snowflake has sold $7 billion worth of its services via AWS Marketplace since its founding in 2012. Driving Growth with AI The growth is driven by AI, with Snowflake offering its AI building tool, Cortex AI, which provides features like text interfaces for database queries and summary reports. The increasing demand for AI processing power has led to a surge in CPU usage, with CPUs handling most tasks associated with AI. The Role of Graviton Chips Snowflake is signing this contract for more access to AWS's home-grown ARM-based CPU chip, Graviton. Amazon CEO Andy Jassy boasted that Amazon's own homegrown AI chips offer "better price-performance" than Nvidia's offerings. The Data Analysis Snowflake has sold $7 billion worth of its services via AWS Marketplace since 2012. The new deal is worth $6 billion over five years. Snowflake's customers are accelerating their spending on AWS, doubling to $2 billion in 2025. The Impact Analysis The deal highlights the growing demand for AI processing power and the increasing competition in the cloud computing market. Cloud providers like AWS are deploying chips as fast as they can to meet the demand. The Prediction The multibillion-dollar deals signed by AWS, including the one with Snowflake, show how AI is lifting the boat for cloud providers. As AI continues to grow, cloud providers will need to invest in more AI processing power to meet the demand.

The Lead A phenomenon dubbed "AI psychosis" is reportedly affecting tech executives, particularly CEOs, who are overestimating artificial intelligence capabilities while simultaneously implementing mass layoffs. This disconnect between perception and reality is creating organizational chaos in the tech industry. The CEO AI Delusion Box founder Aaron Levie has suggested that CEOs are uniquely prone to "AI psychosis" because they're sufficiently distant from the implementation details of AI systems. When executives "play with AI" by developing prototypes or generating contracts, they often make the leap to believing AI agents can fully handle complex work without understanding the limitations. Unlike their technical teams, CEOs aren't responsible for reviewing code, discovering bugs, or training AI models on company-specific requirements. This lack of firsthand experience with AI's limitations doesn't stop them from making decisions based on overoptimistic assessments of AI capabilities. The Layoff Numbers In the first five months of 2026 alone, the tech industry has already seen 115,430 people fired from 152 tech companies. This nearly matches the 124,636 people let go by 275 companies throughout all of 2025, according to industry tracker Layoffs.fyi. The majority of these layoffs have been attributed to AI, though many argue that companies are engaging in "AI washing" - crediting AI productivity gains when other business decisions are really driving the cuts. The ClickUp Experiment Zeb Evans, CEO of project management software startup ClickUp, proudly declared on X that he had laid off almost a quarter of his employees (22%) after implementing approximately 3,000 AI agents for internal work. Evans insisted this wasn't a cost-cutting measure but rather an attempt to create what he calls a "100x org" composed of people who run and review AI agents' work. The Productivity Paradox Research on AI and productivity presents a complex picture. A meta-analysis published in UC Berkeley's California Management Review found "no robust relationship between AI adoption and aggregate productivity gain." Meanwhile, research from the National Bureau of Economic Research concluded that while AI adoption does improve productivity, there's a "productivity paradox" in which perceived gains exceed measured improvements. MIT researchers studying thousands of AI agents found they aren't yet producing human-quality work in many cases. They predict that at the current rate of improvement, large language models will "be able to complete most text-related tasks with success rates of, on average, 80%–95% by 2029 at a minimally sufficient quality level," with additional time needed to outperform humans. The Executive Bottleneck Research published in the Harvard Business Review suggests that when everyone in an organization uses AI to produce more output, the bottleneck simply shifts to executives. Their work awaits authorization of all the content being generated by AI-empowered employees. If everyone is empowered to act, the system risks becoming overwhelmed, as evidenced by OpenAI's experience last year. As Levie advises, CEOs should use AI extensively to understand both its capabilities and limitations. However, with the current trend of mass layoffs and organizational restructuring based on overoptimistic AI assessments, the tech industry may face continued chaos until this balance is achieved.

OpenRouter announced a $113 million Series B financing round led by CapitalG, the growth arm of Alphabet, lifting its post‑money valuation to an estimated $1.3 billion. The round marks a dramatic increase from the roughly $547 million valuation recorded a year ago. Series B Funding and New Valuation Milestone Lead investor: CapitalG (Alphabet) Round size: $113 million Post‑money valuation: ~$1.3 billion Previous valuation (2025): ~$547 million Earlier round: $40 million Series A in June 2025, led by Andreessen Horowitz and Menlo Ventures Scale Metrics: Users, Tokens, and Model Portfolio Active global users: 8 million Monthly token throughput: 100 trillion tokens (≈25 trillion per week) Weekly token growth: 5× increase from 5 trillion tokens six months earlier Model catalog: access to > 400 models from providers such as Anthropic, Google, OpenAI, xAI, DeepSeek Why Multi‑Model Gateways Are Redefining AI Procurement The surge in OpenRouter’s usage reflects a broader shift from single‑model reliance to a flexible, agent‑driven AI stack. Enterprises now prefer a "swappable engine" approach, allowing them to match the most cost‑effective or highest‑performing model to each specific task without vendor lock‑in. Future Outlook: Expansion of Agent‑Driven AI and Competitive Landscape As AI workloads move deeper into inference and autonomous agents, platforms that can orchestrate dozens of models will become critical infrastructure. OpenRouter’s rapid growth suggests it will attract further investment and potentially expand into edge‑deployment services, while traditional SaaS providers may need to integrate similar multi‑model capabilities to stay competitive.

Palantir Accuses Mayor of Prioritising Politics Over SafetyPalantir says London mayor Sadiq Khan is “politicising procurement” by blocking a two‑year, £50 million AI contract for the Metropolitan Police, arguing the move jeopardises public safety.Mayor Blocks £50m AI Procurement Deal with Met PoliceKhan’s office cited a “clear and serious breach” of procurement rules and rejected the plan for the Met to use Palantir’s AI to process intelligence in criminal investigations. The decision was first reported by the Guardian on 21 May 2026.Financial Stakes: £50m Contract and Wider Government Deals£50 million – value of the blocked Met Police contract.£330 million – NHS England deal with Palantir.£240 million – Ministry of Defence agreement.Less than £500,000 – earlier separate AI pilot with the Met to detect rogue officers.Political Fallout and Policing Implications in LondonThe move has split Labour MPs: Rosena Allin‑Khan and Clive Lewis praised the block, while Stella Creasy condemned Palantir’s CEO for “using sexual‑abuse allegations to attack the mayor”. The Metropolitan Police Federation called the AI system “big brother”. Business Secretary Peter Kyle defended Palantir’s capabilities and urged Khan to explain his decision.Future of AI Procurement and Domestic Tech AlternativesKhan’s stance may encourage a shift toward British‑owned AI solutions, echoing Kyle’s call for more investment in domestic firms. Ongoing debates about foreign AI providers could reshape how UK public services adopt advanced technology, with potential impacts on policing effectiveness and public trust.