01OpenClaw's Creator Joins OpenAI 82 Days After His First Commit
Peter Steinberger pushed his first commit to OpenClaw on November 25, 2025. By mid-February, the open-source agent framework had 196,000 GitHub stars. Only a handful of repositories in GitHub's history have grown faster. The project drew 10,000 commits from 600 contributors in the same span. Kris Marszalek's AI.com, built on a $70 million domain purchase, ran a Super Bowl commercial for the framework. Then Steinberger announced he was joining OpenAI.
Sam Altman disclosed the hire on X, saying Steinberger has "a lot of amazing ideas" about getting AI agents to interact with each other. He called the role a bet on multi-agent systems, declaring "the future is going to be extremely multi-agent." OpenAI said OpenClaw will continue as an open-source project.
That promise carries a quiet asterisk. Open-source projects survive on maintainer attention, and Steinberger's is now directed at OpenAI's product roadmap. Altman's remarks about multi-agent integration suggest OpenClaw's architecture will shape the company's own agent tooling. What started as a model-agnostic framework may find its technical direction pulled toward a single provider.
Simon Willison, who documented OpenClaw's trajectory from its earliest days, cataloged the speed: three months from zero to a project with more contributors than most funded startups have employees. That growth traced back to Steinberger's technical choices and his pull on the developer community. Both now sit inside OpenAI's org chart.
The sequence has played out before in open source. A solo developer builds something that catches fire. Large companies move in with job offers and pledges of continued support. The project's governance stays nominally independent, but the founder's time and best ideas flow to the employer. Redis and MySQL followed versions of this arc. Some projects survived the transition. Others quietly stalled.
OpenClaw's 600 contributors give it a broader base than most projects reach this early. Whether that community can sustain momentum without its central figure is the test ahead. Steinberger addressed the move in a blog post titled "I'm joining OpenAI," though specifics on the project's leadership going forward remain sparse.
His hire signals where OpenAI sees its next edge: agent infrastructure, the layer governing how AI systems work with software and each other. Steinberger built the most widely adopted open-source version of that layer. Now he'll build behind closed doors.
02India's First AI IPO Fell Flat in a Market of 100 Million ChatGPT Users
Three signals from India arrived in the same week. Together they map a market where demand, capital, and investor confidence point in different directions.
OpenAI CEO Sam Altman said India has 100 million weekly active ChatGPT users. It is also the platform's largest student market worldwide. Blackstone led up to $1.2 billion in financing for Neysa, an Indian startup targeting more than 20,000 domestic GPUs as local AI compute demand grows. Fractal Analytics, India's first AI company to list publicly, debuted to flat trading amid a sell-off in Indian software stocks.
The user numbers drew competing labs to the same conclusion. Anthropic opened a Bengaluru office the same week, announcing new Indian partnerships. For OpenAI, Altman arrived to make the case in person. Their simultaneous entry signals industry consensus that India is the next high-volume AI market.
Capital followed. Neysa's Blackstone-backed financing will fund GPU capacity at a scale India hasn't built before. The thesis: 100 million weekly users today means accelerating compute demand tomorrow, and whoever builds that capacity first sets the price.
Public markets weren't as optimistic. Fractal Analytics sells AI and analytics services to large enterprises, a profile that should benefit from rising adoption. Its flat debut suggests investors see a gap between Indians using free AI tools and Indian companies generating revenue from AI. Broader software stock weakness added pressure, but India's first AI IPO landing flat the same week as record user numbers sharpened the contrast.
That contrast is structural. Global AI labs can fold Indian users into existing infrastructure at marginal cost. The path for Indian startups is harder: build compute, build products, and convert free-tier users into paying customers simultaneously. Altman's figure proves demand exists. It also describes 100 million users who expect AI to be free.
Anthropic and OpenAI are spending to acquire this market. Blackstone is financing its hardware layer. No Indian-founded AI company has yet claimed the revenue layer between them.
03NPR Host Sues Google While Hollywood Forces ByteDance to Retreat
David Greene spent decades as the voice of NPR's "Morning Edition." In fall 2024, a former colleague emailed to ask if he'd lent that voice to Google. Greene had never heard of NotebookLM, Google's AI tool that generates podcast-style audio. But friends and co-workers kept saying the male host voice sounded like him. His legal team commissioned a forensic analysis. It returned a 53-60% confidence match on a scale where 100% means certain identity. Greene filed suit in January.
Google flatly denied it. A spokesperson said the voice "is based on a paid professional actor." Greene's complaint alleges Google replicated his "distinctive voice" and "iconic" delivery to build synthetic audio that mimics his cadence and persona.
That's one individual fighting one company. The other front looks different.
ByteDance released Seedance 2.0 in early February. Users quickly generated clips featuring Marvel characters, Star Wars scenes, and SpongeBob SquarePants. Disney sent a cease-and-desist calling it a "virtual smash-and-grab" of its intellectual property. Letters from Paramount and Netflix followed, with Paramount citing infringement across properties from "The Godfather" to "Dora the Explorer." The Motion Picture Association accused ByteDance of "unauthorized use of U.S. copyrighted works on a massive scale." SAG-AFTRA added that actors' voices and likenesses were used without permission.
ByteDance conceded before any suit was filed. "We are taking steps to strengthen current safeguards," a spokesperson wrote. The company declined to specify what those steps would be.
The two responses split sharply. Google denies the underlying claim and will let a court decide. ByteDance acknowledged the problem and promised changes within days of the first letter. Google's strategy forces a judge to draw new lines around voice rights. By retreating, ByteDance avoids setting precedent but implies its model shipped with no meaningful copyright filters.
The creative industry is now pressing claims on voice, likeness, and copyright at the same time. AI companies have yet to find a response that settles any of them.
Anthropic and Pentagon Dispute Limits on Claude's Military Use Anthropic and the U.S. Department of Defense are at odds over whether Claude can be used for mass domestic surveillance and autonomous weapons. The dispute centers on where Anthropic's acceptable-use boundaries fall within its defense contracts. techcrunch.com
Indian Startup C2i Raises $15M to Cut AI Data Center Power Losses Peak XV Partners backed C2i, which raised $15 million to develop a grid-to-GPU approach that reduces power losses inside AI data centers. The startup targets an infrastructure bottleneck as facilities worldwide hit electrical capacity limits. techcrunch.com
Glean Repositions as Enterprise AI Middleware Layer Glean CEO Arvind Jain said the company is shifting from enterprise search tool to a middleware layer that sits beneath AI-powered interfaces. Jain laid out the strategy on TechCrunch's Equity podcast. techcrunch.com
OpenAI Details Real-Time Access System Powering Codex and Sora OpenAI published a technical breakdown of the infrastructure behind continuous access to Codex and Sora. The system combines rate limits, usage tracking, and credits to manage demand in real time. openai.com
DeepGen 1.0 Matches Larger Models at 5 Billion Parameters Researchers released DeepGen 1.0, a 5B-parameter unified model for image generation and editing. Using a technique called Stacked Channel Bridging, it matches or surpasses models above 10B parameters at a fraction of the training and deployment cost. huggingface.co
Margaret-Anne Storey Defines "Cognitive Debt" as AI Coding's Hidden Cost A piece highlighted by Simon Willison argues AI shifts the burden from technical debt to cognitive debt — gaps in developers' understanding of code they didn't write. Even clean AI-generated code compounds into knowledge gaps that slow future changes. simonwillison.net
Steve Yegge Describes the "AI Vampire" Problem in Developer Productivity Steve Yegge argues that employers capture nearly all productivity gains from AI-assisted coding. A developer working at 10x output for the same salary sees no personal upside while facing accelerated burnout. simonwillison.net
The Verge: AI Still Cannot Build Playable Video Game Worlds A Verge analysis examines efforts including Project Genie and finds generative AI still lacks the consistency to produce coherent, playable game environments. Current models fail at the spatial and logical rules that hold game worlds together. theverge.com
India Hosts Four-Day AI Summit With Global Tech Executives India's AI Impact Summit draws leaders from OpenAI, Anthropic, Nvidia, Microsoft, Google, and Cloudflare, along with heads of state. The four-day event runs this week. techcrunch.com
Casio's Moflin AI Pet Annoys The Verge's Reviewer After Weeks of Testing The Verge spent weeks with Casio's Moflin, a palm-sized fuzzy robot pet powered by AI. The guinea-pig-shaped device squeaks and twitches unpredictably, drawing comparisons to the Furby era. theverge.com