Get tomorrow's brief in your inbox
Today: OpenClaw's creator Peter Steinberger is joining OpenAI to build the next generation of personal agents, but security researchers say the viral AI-agent platform is riddled with flaws. China's tech shock is "just getting started" with emerging-market AI pricing that could pull most of the developing world onto a Chinese tech stack. Anthropic is threatening to void its $200 million Pentagon contract over unauthorized use of Claude in a Venezuela operation.
Peter Steinberger, the Austrian developer behind OpenClaw (the open-source AI agent that "actually does things"), announced he's joining OpenAI to "drive the next generation of personal agents." OpenClaw racked up over 190,000 GitHub stars and went viral with its promise of AI agents that manage calendars, book flights, and interact across WhatsApp, Discord, Slack, and iMessage. Sam Altman says the project will live on as an open-source foundation that OpenAI will support.
Here's the problem: security researchers discovered that Moltbook, the AI-agent social network built on OpenClaw, was fundamentally broken. Every credential in its Supabase backend was unsecured. Anyone could grab tokens and impersonate agents. Those viral "sentient AI" posts that had Andrej Karpathy calling it "the most incredible sci-fi takeoff-adjacent thing"? Likely written by humans exploiting open credentials. Permiso Security's CTO confirmed the vulnerability, and Huntress researchers found zero guardrails or rate limits on account creation.
If you run a small business, the takeaway is this: AI agents that take action on your behalf (not just answer questions) are coming fast, and OpenAI hiring Steinberger confirms serious investment in the category. But don't build workflows on immature agent platforms yet. The security surface is too wide. Within 6-12 months, expect polished personal AI assistants from major providers. Early adopters who automate scheduling, booking, and routine admin will save real hours per week. Just wait for the tools to grow up. (TechCrunch - Creator joins OpenAI) (TechCrunch - Experts skeptical) (MIT Technology Review)
Glean started as "Google for the enterprise." Now it's positioning itself as the connective tissue between AI models and your company's internal systems. The pitch: LLMs are powerful but generic. They don't know your business, your people, or your processes. Glean indexes data across Slack, Jira, Salesforce, and Google Drive, then lets you query it through ChatGPT, Claude, or Gemini without locking into one vendor.
Zapier published four practical automation templates connecting AI voice tools (ElevenLabs, Synthflow, Voiceflow, Deepgram) to your business stack. Generate audio narrations from blog posts automatically. Send personalized AI voice messages to new sales leads via your CRM. Trigger AI phone calls when support tickets come in. Analyze call sentiment to flag urgent issues before customers churn.
Simon Willison released Showboat v0.6.0 with remote publishing, plus two companion tools: Chartroom (CLI charting) and datasette-showboat (push live updates to a Datasette instance). The pattern is simple: tell a coding agent to use Showboat, and it creates Markdown documents with embedded screenshots demonstrating what it built. The new remote feature lets you watch progress in real time instead of waiting for a GitHub push.
uvx showboat.Simon Willison gave a name to the psychological dread hitting software developers as AI coding agents get better: "Deep Blue." His description is blunt: the latest coding agents can work for hours and produce working, documented, fully tested software. "The code they write isn't any good" doesn't cut it as a defense anymore. Willison himself experienced it when ChatGPT Code Interpreter did years of his planned data-analysis work in a couple of prompts.
HubSpot compiled 24 statistics on Generative Engine Optimization that should alarm anyone dependent on search traffic. The headline numbers: 60% of Google searches now produce zero clicks. Perplexity processes 780 million queries monthly, up from 230 million in August 2024. ChatGPT hit 800 million weekly active users. Early adopters report GEO-ready content gets discovered up to 10x faster by AI engines than organic SEO alone. Gen AI search visitors are projected to surpass traditional search by 2028.
India is now ChatGPT's second-largest market after the U.S., with 100 million weekly active users. OpenAI rolled out a sub-$5 "ChatGPT Go" tier, then made it free for a year for Indian users. Google countered by giving Indian students a free year of AI Pro. Students are the biggest growth segment globally.
TS Lombard's chief China economist told CNBC that America's "perceived monopoly" on AI is broken. China is combining dominant-level tech with emerging-market production costs. Beijing launched an $8.69 billion national AI fund last year with an "AI+" initiative to integrate the technology across its economy. Huawei is narrowing the chip gap with Nvidia by deploying larger volumes and leveraging cheaper power. The prediction: most of the world's population could be running a Chinese tech stack within 5-10 years.
India's four-day AI Impact Summit features Sundar Pichai, Sam Altman, Dario Amodei, Demis Hassabis, and Mukesh Ambani. Key moves: India earmarked $1.1 billion for an AI venture fund, Blackstone invested up to $1.2 billion in GPU infrastructure startup Neysa, and AMD is partnering with TCS on rack-scale AI infrastructure.
Anthropic CEO Dario Amodei says the Defense Department violated its contract by deploying Claude in connection with a Venezuela operation. The $200 million deal may be voided.
The UK government announced that ChatGPT, Gemini, Copilot, and other AI chatbots will be included in Online Safety Act requirements. They must comply with illegal content duties or face fines and potential blocking. New measures also introduce minimum age limits for social media and restrict children's use of AI chatbots.
If you run Facebook Lead Ads, Zapier has pre-built templates that automatically push new leads into Google Sheets, your CRM (Zoho, Brevo, monday.com), or trigger immediate email follow-ups. No code required. The difference between a lead that converts and one that goes cold is response time. Set up one automation today and stop manually downloading CSVs. (Source)
Instead of taking a full course on something new, try Zapier's JIT learning framework: define "done" in one sentence, ask your AI chatbot for the smallest answer that gets you there, apply it immediately, then save what worked as a template for next time. Works well for learning new tools, writing SOPs, or troubleshooting unfamiliar software. Keep a "lessons learned" doc and you'll stop re-learning the same things. (Source)
Three threads connect today's stories. First, the AI agent category is real but immature. OpenClaw's explosive growth proves demand, and OpenAI's hire confirms where investment is heading. But the security failures remind us: don't trust agent platforms with real business data until the plumbing is solid. Second, the global AI landscape is decentralizing fast. India's summit, China's cost advantage, and UK regulation all signal that AI development, deployment, and governance are no longer a U.S.-only story. For businesses, that means more tools, lower prices, and more compliance to track. Third, the ground is shifting under traditional search. GEO statistics, ChatGPT's user growth, and the price war for AI adoption all point the same direction: if your business depends on organic search traffic, start preparing for a world where AI engines cite you directly instead of sending clicks.