Breaking AI & Tech News Analyzed

The Leadership Shift at Anthropic Andrej Karpathy, the AI researcher who co-founded and formerly worked at OpenAI and previously led AI at Tesla, has joined Anthropic. Karpathy announced his move on X, stating that he is excited to join the team and get back to R&D.; Karpathy's Role in Pre-Training Karpathy started this week at Anthropic, where he is working on pre-training under team lead Nick Joseph. Pre-training is responsible for the large-scale training runs that give Claude its core knowledge and capabilities. Karpathy will start a team focused on using Claude to accelerate pre-training research. The Significance of Karpathy's Move Karpathy is one of the few researchers who can bridge the gap between LLM theory and large-scale training practice. Tapping him to build such a team is a clear sign from Anthropic that it believes AI-assisted research, rather than pure compute, is how it stays competitive with OpenAI and Google. Karpathy's Background Co-founded OpenAI and worked on deep learning and computer vision until 2017 Led Tesla's Full Self-Driving (FSD) and Autopilot programs from 2017 to 2022 Returned to OpenAI for one year before leaving in 2024 to start Eureka Labs, a startup dedicated to applying AI assistants to education Anthropic's Recent Hires Anthropic has also brought on Chris Rohlf to its frontier red team, which stress-tests advanced AI models against severe threats. Rohlf is a veteran of the cybersecurity industry with more than 20 years of experience. The Future of AI Research Karpathy's move to Anthropic and the company's focus on AI-assisted research signal a new direction in the AI landscape. As Karpathy stated, "I think the next few years at the frontier of LLMs will be especially formative."

Executive Summary: Forecasts at Risk Amid Budget CutsAs the United States braces for an intense hurricane season and unprecedented summer heat, experts caution that the Trump administration’s proposed 40% reduction in NOAA funding and broader cuts to climate and weather data programs could make federal weather forecasts less reliable when they are needed most.Policy Changes Undermine Data‑Intensive AI ForecastingThe agency launched a suite of AI‑powered global weather models last year, promising faster and more accurate predictions. However, those models are trained on "centuries of weather data," a resource that is being eroded by staffing reductions, satellite de‑commissioning, and fewer balloon launches.NOAA AI model suite introduced late 2025 to improve speed, efficiency, and accuracy.Data cuts include scaling back satellite operations and balloon launches, threatening key observation systems.Budget proposal offers a modest increase for the National Weather Service but a 40% cut to NOAA overall.Financial Impact: The 40% NOAA Funding ReductionThe administration’s budget plan calls for a 40% cut to NOAA’s overall budget while only modestly increasing the National Weather Service’s allocation. This disparity reduces resources for data collection, climate research, and the maintenance of observation networks such as ocean buoys.Broader Consequences for Weather PreparednessReduced data collection hampers the ability of both traditional physics‑based models and newer AI models to predict extreme events. Experts note that AI models, which rely heavily on historical patterns, already "underperform" for unprecedented weather extremes, and further data loss could exacerbate this shortfall.Historical AI model performance lags behind physics‑based models for rare events like the February 2026 blizzard.Cutbacks to climate research threaten the skill of future forecasts, as highlighted by former NOAA chief scientist Craig McLean.Upcoming "super El Niño" conditions could amplify heat records and hurricane activity, increasing reliance on accurate forecasts.Future Outlook: Forecast Reliability and Policy ResponseAnalysts predict that unless the data cuts are reversed or mitigated, the reliability of federal weather forecasts will decline, especially for extreme events. While NOAA maintains that AI tools are an addition—not a replacement—to its existing model suite, the tension between budget constraints and the need for robust data persists. The agency is slated to release its 2026 Atlantic hurricane outlook soon, which will test the resilience of current forecasting capabilities under reduced data conditions.

Anthropic’s Claude Mythos to be Presented to the Financial Stability BoardAnthropic will brief the Financial Stability Board (FSB), chaired by Bank of England governor Andrew Bailey, on the cyber‑defence implications of its Claude Mythos model, which has raised alarm among security experts.Mythos is not being released publicly; access is limited to select tech firms and banks such as Apple and JP Morgan.The briefing follows a report by the Financial Times and confirmation from a source familiar with the discussions.The FSB’s membership includes senior officials from the US, UK, Australia and China.Quantifying Mythos’ New Cyber‑Testing PerformanceThe UK’s AI Security Institute (AISI) noted a “notable capability jump” in the version shown to banks. In the “cooling tower” test, Mythos succeeded in 3 out of 10 attempts – a first for any model evaluated by AISI.Previous iterations had not completed the test.AISI reports that the length of autonomous cyber tasks has doubled within months.Implications for Global Financial CybersecurityThe briefing comes as the International Monetary Fund (IMF) warned that AI‑driven cyber risks are rising for financial stability. Central bank leaders, including Goldman Sachs CEO David Solomon and JP Morgan CEO Jamie Dimon, have already expressed heightened awareness of Mythos’ capabilities.Cyber risk does not respect borders; inconsistent oversight could weaken the interconnected financial system.Experts caution that most breaches still stem from traditional weaknesses such as weak authentication.What the Next Phase of AI‑Driven Cyber Risk May Look LikeAISI is developing tougher hacking tests to track AI progress, while the FSB is expected to issue recommendations for coordinated oversight among regulators. If the trend of rapid capability gains continues, financial institutions may need to embed AI‑specific cyber‑defence measures into their risk frameworks.Potential for tighter collaboration between AI developers and regulators.Increased scrutiny of AI models before deployment in critical infrastructure.

The LeadOsaurus has introduced an innovative open-source, Apple-only LLM server that allows Mac users to seamlessly switch between local and cloud AI models while maintaining data privacy on their own hardware. This development addresses growing concerns about AI token costs and security by providing a user-friendly interface that runs AI in a hardware-isolated virtual sandbox.The Evolution from Dinoki to OsaurusOsaurus evolved from the idea for a desktop AI companion called Dinoki, which Osaurus co-founder Terence Pae described as a sort of "AI-powered Clippy." Dinoki's customers had questioned why they should buy the app if they still had to pay for tokens—the usage units AI companies charge for processing prompts and generating responses. This concern led Pae to develop Osaurus as a solution that allows users to run AI locally on their Macs, accessing files, browsers, and system configurations without relying on cloud services.Technical Capabilities and Model SupportOsaurus can flexibly connect with locally hosted AI models or cloud providers like OpenAI and Anthropic, allowing users to choose which AI models best fit their needs. The platform supports various models including MiniMax M2.5, Gemma 4, Qwen3.6, GPT-OSS, Llama, and DeepSeek V4. It also supports Apple's on-device foundation models, Liquid AI's LFM family of on-device models, and cloud connections to OpenAI, Anthropic, Gemini, xAI/Grok, Venice AI, OpenRouter, Ollama, and LM Studio. As a full MCP (Model Context Protocol) server, it provides access to tools for MCP-compatible clients and ships with over 20 native plugins for Mail, Calendar, Vision, macOS Use, XLSX, PPTX, Browser, Music, Git, Filesystem, Search, Fetch, and more. Recent updates have also added voice capabilities.User Adoption and Market PositionSince launching nearly a year ago, Osaurus has been downloaded over 112,000 times according to its website. The platform distinguishes itself from similar tools like OpenClaw or Hermes by offering an easy-to-use interface for consumers rather than developers, while addressing security concerns through a hardware-isolated, virtual sandbox that limits the AI's scope and keeps users' computers and data safe. Currently, Osaurus' founders, including co-founder Sam Yoo, are participating in the New York-based startup accelerator Alliance.The Future of Local AI and Business ApplicationsOsaurus' founders are exploring potential business applications, particularly in sectors like legal services and healthcare where running local LLMs could address privacy concerns. The team believes that as local AI models become more powerful, they could reduce demand for AI data centers. Pae noted that "the intelligence per wattage—which is like the metric for local AI—has been going up significantly," with local AI evolving from barely being able to finish sentences last year to now being able to run tools, write code, access browsers, and perform various tasks. The vision is for businesses to deploy Mac Studios on-premise, using substantially less power than traditional data centers while maintaining cloud-like capabilities.

SpaceXAI has lost more than 50 researchers and engineers since its February merger, sparking concerns about its AI roadmap.Mass Exodus from SpaceXAI After MergerThe newly rebranded entity, formed when SpaceX acquired xAI, has seen a wave of exits across coding, world‑model research, and the Grok voice team. High‑profile leaders, including team lead Juntang Zhuang, have departed, and rival firms are actively recruiting the talent.Departure Surge Across Core Pre‑Training TeamPre‑training, the foundational step for building large AI models, now operates with only a handful of engineers. At least 11 former xAI staff have joined Meta, while 7 have moved to Thinking Machine Labs, Mira Murati’s startup.Numbers Reveal Scale of Talent DrainMore than 50 total departures since February11 employees defected to Meta7 employees joined Thinking Machine LabsTwo co‑founders left shortly after the mergerStrategic Risks for SpaceXAI’s AI AmbitionsThe loss of pre‑training experts threatens the company’s ability to develop competitive large‑scale models. Internal sources cite Elon Musk’s “extreme work” culture and unrealistic deadlines, which have led to corners being cut on projects like Grok. Additionally, generous share‑sale tenders may be prompting staff to cash out rather than stay for a long‑term build‑out.What the Future Holds for the Merged EntityIf the talent gap widens, SpaceXAI could delay or scale back its model‑training roadmap, potentially ceding ground to better‑resourced rivals. Conversely, the company may double down on financial incentives to retain remaining staff or accelerate hiring from the broader AI talent pool. Stakeholders will be watching upcoming product announcements for signs of whether the exodus has been mitigated.

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.

The Controversial Bet on AI BookkeepingDespite the collapse of his previous startup, Ian Crosby is taking another shot at building a business out of automating bookkeeping. His new venture, Synthetic, aims to create a fully autonomous AI bookkeeper that can generate accrual-based financials without direct human involvement.The Vision Behind SyntheticSynthetic is designed to revolutionize bookkeeping by eliminating the need for human accountants, a stark contrast to current accounting startups like Xero. Crosby maintains an all-or-nothing approach: "We're not going to release anything that's not fully autonomous. It's that or bust."The startup is currently in the design phase, with Crosby acknowledging that his vision may not yet be technologically possible. The company plans to initially serve only AI and other software startups.The $10 Million InvestmentDespite the challenges and Crosby's troubled past with Bench Accounting, Synthetic has successfully raised $10 million in a Seed funding round led by Khosla Ventures. The round also saw participation from Basis Set Ventures and Shopify CEO Tobias Lütke.This financial backing provides Crosby with the resources to wait for foundational AI models to become more reliable for bookkeeping calculations. "I've raised years of cash, so we can just wait it out," Crosby stated.Learning from Past FailuresKhosla partner Jon Chu defended the investment by explaining his tendency to "run towards controversy a little bit." He cited Parker Conrad's journey from Zenefits to founding Rippling (now valued at $17 billion) as an example of how industry narratives can be misleading.Chu conducted thorough due diligence, speaking with several executives who worked with Crosby after his departure from Bench. According to Chu, they "had fantastic things to say about Ian." This feedback, combined with Crosby's subsequent roles at Shopify and founding of Teal (which was acquired by Mercury), convinced Khosla of his growth potential.The Bench Accounting FalloutCrosby's previous venture, Bench Accounting, famously shut down in 2024 before being "bought for scraps." Crosby maintains he wasn't directly responsible for bringing the company to insolvency, stating he was fired by Bench's board in 2021 after turning down a $250 million acquisition offer from Brex.The board reportedly disagreed with Crosby's strategic direction as the business was bleeding cash, and his executive team was frustrated with his direct leadership style. "He took a big swing, made a few mistakes. That didn't go well," Chu acknowledged about Crosby's tenure at Bench.The Path to Autonomous AI BookkeepingWhile Synthetic's prototype works for a narrow group of users, Crosby remains uncertain how it will scale for a broader customer base. He compared the current state of AI bookkeeping to "a self-driving car that can drive down one street versus the self-driving car that can drive down any street.""We haven't driven down enough streets to know if it's going to crash," Crosby explained, highlighting the technical challenges ahead. Despite these obstacles, the founder remains committed to his vision of a fully automated financial future.

Es Devlin’s collective selfie portrait challenges Britain’s fragmentationAt the National Portrait Gallery, artist Es Devlin presents a constantly evolving digital collage built from ordinary people’s selfies. The work is framed as a quiet, non‑verbal invitation for the nation to coexist.How the living portrait merges thousands of UK selfiesParticipants upload a selfie via the project website.Each image is rendered in Devlin’s smoky charcoal‑and‑chalk style.The stylised portraits rotate on a framed screen, appearing and fading in a perpetual carousel.The installation deliberately leaves imperfections – mismatched beards, overlapping features – to reflect the difficulty of seamless unity.Technology behind the portrait: Google Arts & Culture Lab’s AI modelWorking with engineers at Google Arts & Culture Lab, Devlin trained an image‑generation model on her hand‑drawn portraits. The AI translates raw selfies into drawings that retain the tactile feel of charcoal, rather than a simple filter.While the project showcases AI’s creative potential, Devlin acknowledges the paradox of offering her artistic “shadow” to a corporate tech platform amid broader debates over artists’ rights.Why a crowd‑sourced portrait matters for British social cohesionThe timing is significant: Britain is experiencing heightened political fury, algorithmic echo chambers, and loneliness. By foregrounding ordinary faces, the work challenges the dominance of celebrity and political imagery in national narratives.Devlin hopes the fleeting intimacy of a shared glance can prompt a pause in the “age of destruction, fragmentation, separation, isolation” she describes.Future steps: workshops, regional roll‑outs and the debate over AI in artBeyond the gallery, Devlin will run free portrait‑drawing workshops and online classes, with plans to bring the installation to town halls, libraries and schools across the UK.The project also sparks discussion about the role of AI in cultural production – whether it is a tool of industrial capitalism or a means of artistic re‑appropriation.

The LeadWirestock, a company that transitioned from stock photography to AI data provision, has secured $23 million in Series A funding to expand its multi-modal data supply business for AI labs. The company now serves six of the largest foundation model makers and has built a platform with over 700,000 artists and designers contributing creative assets.The Creative Data TransformationWirestock previously helped photographers distribute and sell their work on stock photography services like Shutterstock. In 2023, the company pivoted to becoming a data provider, supplying datasets of images, videos, design assets, and gaming and 3D content to AI labs. The platform operates similarly to freelance marketplaces like Fiverr, with artists completing tasks for data collection.Financial Growth and Market PositionThe $23 million Series A round, led by Nava Ventures with participation from SBVP (co-founded by Sheryl Sandberg), Formula VC, and I2BF Ventures, brings Wirestock's total capital raised to approximately $26 million. The company currently has an annual run-rate revenue of $40 million and has paid out $15 million to its contributors. Wirestock employs 60 people and will use the new funding to hire for research, engineering, and product roles.The Creative AI Data Market ExpansionDemand for data supply services is soaring as AI labs compete to enhance their models. Companies like Surge, Scale AI, and Mercor have built billion-dollar businesses on dataset demand, while new startups such as Micro1, Human Archive, and Human Native AI also partner with top AI model makers. Wirestock focuses specifically on providing data for creative use cases like image and video generation, with plans to expand into audio and music modalities.Future of Multi-Modal Data in AI DevelopmentLooking ahead, Wirestock is building enterprise software for AI labs to collaborate on datasets and plans to continue expanding its creative asset offerings. Freddie Martignetti, founder of Nava Ventures, emphasized the importance of multi-modal data for creating more human-like AI systems. As AI models evolve, the availability and quality of diverse training data will remain critical differentiators in the competitive AI landscape.