OpenClaw creator fielded billion-dollar bids — and nearly lost the whole project to a crypto scam Peter Steinberger’s OpenClaw (formerly Clawdbot, briefly MoltBot) exploded into a phenomenon: a self-modifying, agentic AI assistant that helped spark MoltBook’s viral chaos and an ecosystem of autonomous agents doing “weird and amazing” things online. The repo rocketed to 180,000 GitHub stars in record time and put its Austrian creator squarely in the crosshairs of big tech — and ruthless scammers. Big-tech courting, but with strings attached Steinberger told Lex Fridman that both Mark Zuckerberg and Sam Altman have made concrete offers to acquire the project. Zuckerberg reached out over WhatsApp and the two even spent 10 minutes debating the merits of Claude Opus versus GPT Codex. Altman’s approach included a tangible sweetener: access to massive compute through a Cerebras partnership that could dramatically speed up agent performance. Despite the attention, Steinberger insists his top condition for any deal is that the project remain open source. “Maybe it’s gonna be a model like Chrome and Chromium,” he said. “I think this is too important to just give to a company and make it theirs.” He’s also spoken with Microsoft CEO Satya Nadella and is weighing whether to start a company with VC backing — or keep building independently even as the project bleeds cash. Burn rate and funding choices OpenClaw is expensive to run. Steinberger estimates it’s costing $10,000–$20,000 per month, and he says he routes sponsorship money back into project dependencies rather than taking a cut personally. A previous exit (the sale of PSPDFKit) means he’s not desperate for cash, but the financial hit is real; he says bluntly, “Right now I lose money on this.” A rebrand that almost ended everything The project’s name-change saga nearly killed it. After Anthropic filed a trademark complaint arguing “Clawdbot” was too close to “Claude,” Steinberger renamed the project to MoltBot. In the split second he hit “rename” across browsers, crypto-savvy attackers pounced: bots injected malware via GitHub, hijacked his NPM packages, and turned his Twitter mentions into spam. “I was close to crying,” he said. “Everything’s fucked.” He nearly deleted the repo. The second rebrand — to OpenClaw — required extreme secrecy: decoy names, simultaneous cross-platform account updates, and tight coordination to keep attackers from sniping another rename window. Steinberger called the coordinated hijacks “the worst form of online harassment I’ve experienced.” Agentic engineering, “vibe coding,” and the future of apps Steinberger embraces what Andrej Karpathy calls “agentic engineering” (he dislikes “vibe coding” as a dismissive label). He runs 4–10 agents at once, logged 6,600 commits in January alone, and admits much of the codebase emerged from conversations with AI rather than traditional typing: “These hands are too precious for writing now.” He has bold predictions. OpenClaw-style agents, he says, will render up to 80% of apps obsolete: why open MyFitnessPal when your agent already knows your sleep and stress patterns? Why browse Uber Eats when an assistant can order for you automatically? That vision — plus the project’s rapid rise — explains the intense interest from major platforms. The core dilemma Steinberger is weighing offers and outcomes: join a big company, take compute and resources while preserving an open-source model; form a startup with VC capital; or keep the project independent and keep bleeding money. “I can’t go wrong,” he told Fridman — but the choice will shape whether OpenClaw remains an open commons or becomes a corporate-controlled platform. For the crypto and open-source community, the episode is a stark reminder: influential projects attract both acquisition pressure and supply-chain attacks. OpenClaw’s future will hinge not just on compute and capital, but on the security, governance, and licensing model Steinberger ultimately chooses. Read more AI-generated news on: undefined/news

#AI #BTC