Breaking AI & Tech News Analyzed

Anthropic, the creator of the Claude chatbot, has publicly urged the world’s top AI companies to devise a coordinated pause on advanced AI development, citing the risk that humans could lose control as systems become increasingly autonomous.Anthropic Proposes Coordinated Global AI SlowdownAnthropic’s research institute will explore a “credible slowdown or pause” in collaboration with other labs.The call follows a blog post on Thursday emphasizing the need for an option to temporarily halt progress.OpenAI counters with a report urging democratic governments, not private labs, to set rules and safeguards.Financial Stakes: IPO Valuation and Market DynamicsAnthropic is preparing an IPO that could value the company at nearly a trillion dollars.The move comes as Anthropic and OpenAI compete to attract investors in the burgeoning AI market.A recent Trump administration executive order asks labs to voluntarily submit their most capable models for government cybersecurity testing before public release.Industry and Regulatory Implications of a PauseA coordinated slowdown aims to prevent “least cautious” players from gaining an advantage while others pause.Anthropic argues that verification mechanisms are needed to ensure no lab secretly advances.Past safety initiatives, such as the 2023 Future of Life Institute’s six‑month halt, have struggled to gain traction.Anthropic’s safety stance includes refusing U.S. military use of its models for domestic surveillance and autonomous weapons, leading to a national security blacklist.Future Outlook: Prospects for Global CoordinationAnthropic’s co‑founder Jack Clark and research head Marina Favaro stress that a pause would buy time for “societal structures and alignment research” to keep pace with AI advances.Experts warn that recursive self‑improvement could enable AI to design successors, heightening control risks.Collaboration between companies, governments, and academia is seen as essential to develop countermeasures against AI‑driven cyber threats.

Executive Summary: Anthropic’s Call for a Temporary Global AI PauseAnthropic announced a proposal for a worldwide “temporary pause” on advanced AI development and pledged to convene policymakers, researchers, and civil‑society actors to discuss the emerging risks of recursive self‑improvement in its Claude model.Anthropic Details Its Latest Claude Advances and the “Recursive Self‑Improvement” NarrativeThe company’s Thursday post highlighted a steady “trend” of increasing capability in Claude, suggesting that with enough compute the system could eventually design and develop its own successor – a scenario long flagged by AI‑safety scholars as a potential pathway to superintelligence.Claude now “runs experiments” and proposes its own coding tasks.As of May 2026, more than 80% of code merged into Anthropic’s codebase was authored by Claude.Anthropic also referenced its unreleased model Mythos, described as “too powerful” for public release.Quantifying Anthropic’s Recent Milestones$1tn potential valuation from the company’s upcoming IPO filing.Embedding of Anthropic engineers inside the US National Security Agency to support offensive cyber operations, as reported by the Financial Times.Claude’s code‑generation contribution surpasses 80% of merged code, indicating a high degree of automation.Implications for AI Governance, National Security, and Public TrustThe juxtaposition of a public safety pause with behind‑the‑scenes collaboration with U.S. intelligence agencies raises questions about Anthropic’s “narrow” definition of AI safety, noted by Steven Murdoch (UCL) and Heidy Khlaaf (AI Now Institute). Critics argue that the company’s actions could undermine credibility and fuel skepticism about the sincerity of its policy outreach.Future Outlook: How a Global Pause Might Shape the AI LandscapeIf policymakers adopt Anthropic’s proposal, the pause could slow competitive pressure among AI labs, allowing regulators to craft standards for recursive self‑improvement and for the use of AI in cyber‑operations. Conversely, without coordinated enforcement, the call may remain symbolic, leaving the industry to self‑regulate amid escalating geopolitical tensions.

Why a Living AI Glossary Matters NowArtificial intelligence is reshaping every industry, but its rapid evolution has spawned a parallel explosion of terminology that can leave even seasoned technologists feeling insecure. TechCrunch’s new glossary aims to provide a single, regularly‑updated reference that translates the most common AI buzzwords into plain language.Key Definitions from AGI to RLHFThe article walks readers through a spectrum of concepts, including:Artificial General Intelligence (AGI) – AI that outperforms humans on most economically valuable tasks, as defined by OpenAI and Google DeepMind.AI Agent – An autonomous tool that can perform multi‑step tasks such as expense filing, ticket booking, or code maintenance.API Endpoints – “Buttons” that let software components interact, enabling agents to automate third‑party services.Chain‑of‑Thought Reasoning – A technique that breaks problems into intermediate steps to improve accuracy.Compute – The hardware (GPUs, CPUs, TPUs) that powers AI model training and inference.Deep Learning – Multi‑layered neural networks that learn features directly from data.Diffusion – The process behind many generative AI models that learns to reverse noise‑added data.Distillation – A teacher‑student method for creating smaller, faster models like GPT‑4 Turbo.Fine‑Tuning – Adding task‑specific data to a pre‑trained model to improve performance.GAN – Generative Adversarial Networks that pit a generator against a discriminator to produce realistic outputs.Hallucination – When models generate inaccurate or fabricated information.Inference – Running a trained model to make predictions, often accelerated by specialized hardware.LLM – Large Language Models that power assistants such as ChatGPT, Claude, Gemini, and Llama.Memory Cache (KV Caching) – An optimization that stores intermediate calculations to speed up inference.Open Source vs. Closed Source – The debate over publicly available model code (e.g., Meta’s Llama) versus proprietary systems (e.g., OpenAI’s GPT).Parallelization – Executing many calculations simultaneously, a cornerstone of modern AI hardware.RAMageddon – The current shortage of memory chips driven by AI data‑center demand.Recursive Self‑Improvement (RSI) – Models that can redesign themselves, a potential step toward singularity.Reinforcement Learning from Human Feedback (RLHF) – Training models with reward signals to improve helpfulness and safety.Tokens & Throughput – The basic units of text processing that determine cost and performance.Quantifying the AI Vocabulary ExplosionThe glossary covers more than 30 distinct terms, each accompanied by concise explanations and links to deeper resources. By cataloguing this breadth, the piece highlights how quickly the AI lexicon has expanded within just a few years of mainstream adoption.Implications for Developers, Investors, and the PublicUnderstanding this terminology is no longer optional. For developers, clear definitions accelerate product building and reduce miscommunication when integrating APIs or deploying agents. Investors gain a sharper lens for evaluating startup pitches that hinge on concepts like fine‑tuning or distillation. Meanwhile, the broader public can better assess claims about “AGI” or “hallucinations,” mitigating hype‑driven misinformation.Future of AI Terminology and Industry AdoptionTechCrunch positions the glossary as a “living document,” promising regular updates as new techniques (e.g., emerging diffusion variants or next‑gen RLHF methods) appear. As AI systems become more autonomous and specialized, the vocabulary will continue to evolve, making ongoing education essential for anyone interacting with the technology.

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 Emergence of Recursive Superintelligence Richard Socher, known for founding You.com and his work on Imagenet, has joined the current generation of research-focused AI startups with Recursive Superintelligence, a San Francisco-based startup that came out of stealth with $650 million in funding. The Vision for Recursive Self-Improvement Socher, along with prominent AI researchers Peter Norvig and Tim Shi, aims to create a recursively self-improving AI model that can autonomously identify its own weaknesses and redesign itself to fix them without human involvement. The Unique Approach: Open-Endedness The startup's unique approach is to use open-endedness to achieve recursive self-improvement. This involves building a system that can automatically generate research ideas, implement, and validate them, potentially leading to a new kind of sense of self-awareness. The Technical Meaning of Open-Endedness Open-endedness refers to the ability of an AI system to create and interact with new concepts, worlds, and agents. Examples include Google DeepMind's Genie 3 and rainbow teaming, where two AIs co-evolve to improve safety. The Future of AI Research and Compute Socher believes that compute will become the only important resource in the future of AI research, and the question will be how much compute humanity wants to spend to solve which problems. The Path to Product Development While Recursive Superintelligence is focused on research, Socher expects the company to develop products that people will love to use, with a positive impact on humanity, in the near future, with timelines potentially being pulled up.

Funding Surge Powers Ineffable Intelligence’s Superlearner QuestIneffable Intelligence announced a $1.1 billion financing round that values the startup at $5.1 billion, positioning it among the elite "pentacorn" AI companies. The capital will fuel the creation of a "superlearner"—an AI system that acquires knowledge solely through trial‑and‑error reinforcement learning.Building a Reinforcement‑Learning Superlearner Without Human DataThe venture’s core mission is to engineer an AI that discovers skills and concepts without any human‑curated datasets. Leveraging David Silver's expertise from DeepMind’s AlphaZero breakthroughs, the team aims to let the system iterate in simulated environments until it autonomously uncovers optimal strategies.Focus on pure experience‑driven learning rather than supervised datasets.Target domains span games, robotics, and scientific discovery.Initial prototypes will run on custom GPU clusters supplied by Nvidia.$1.1 B Funding Round Values Startup at $5.1 BThe round was led by Sequoia Capital and Lightspeed Venture Partners, with participation from Index Ventures, Google, Nvidia, the British Business Bank and the sovereign fund Sovereign AI. Highlights include:Lead investors: Sequoia Capital, Lightspeed Venture PartnersStrategic backers: Google, NvidiaValuation: $5.1 billion post‑moneyComparable rounds: AMI Labs ($1.03 billion) and Recursive Superintelligence ($500 million‑$1 billion)London’s Ascendance as a Global AI HubThe influx of multi‑billion‑dollar rounds signals a shift of AI capital toward the United Kingdom. Factors driving the momentum include DeepMind’s continued presence, supportive government funds like the British Business Bank, and a dense network of alumni launching new ventures.London now hosts three AI startups valued above $5 billion.Proximity to Google’s AI campus and interest from Jeff Bezos’ Project Prometheus further cement the ecosystem.What Success Could Mean for the Future of AI ResearchIf Ineffable’s superlearner achieves human‑data‑free mastery, it could redefine AI development pipelines, reducing reliance on massive curated datasets and accelerating breakthroughs in domains where data is scarce or proprietary.Potential to democratize AI capabilities across industries.May trigger a new wave of reinforcement‑learning‑first models, challenging the dominance of large language models.Founder David Silver pledges all personal earnings to high‑impact charities, linking AI progress to societal benefit.

The UK government has taken a significant step in boosting its tech sector by announcing its first investment in a £500m sovereign AI fund. Technology Secretary Liz Kendall has urged the public to 'make AI work for Britain', despite concerns about job disruption and cybersecurity risks.Kendall acknowledged that 'people are worried about the risks and what it means for their jobs', but emphasized that AI entrepreneurs believe they can create new employment opportunities. The government has taken an undisclosed shareholding in London-based Callosum, a company that helps different types of computer chips work together efficiently to train and operate AI models.The investment is part of a broader effort to support national AI champions and ensure that internationally competitive companies can start, scale, and stay in Britain. The sovereign AI unit, designed to act like a venture capital fund, has also provided access to a network of government-funded supercomputers to help six UK companies develop AI models.These companies include Prima Mente, which is building 'biological foundation models' to tackle diseases like Alzheimer's; Cursive, a company developing autonomous AI agents founded by Google DeepMind alumni; and Odyssey, which develops 'world models', an approach to AI where systems interact with a convincing simulation of the real world.Rachel Reeves, the chancellor, said that by supporting national AI champions, the UK could ensure that internationally competitive companies can 'start, scale and stay here in Britain'. The investment is seen as a key step in establishing the UK as a leader in the AI sector.