Agents no longer just think: they pay. And the infrastructure that supports them is starting to crack. In episode 8, Jorge and Antonio walk through the agentic economy end to end — from micropayments to GitHub outages — and close with a profile of the Nobel laureate behind DeepMind.
Stripe, Alipay and Cloudflare: the agent economy
Stripe has launched its infrastructure for agentic payments with a “Total Compensation” program that covers the agent’s mistakes — the equivalent of a new employee’s insurance. In China, Alipay AI has unlocked agent payments at massive scale. And Cloudflare has redesigned its stack to be fully manageable by agents: deployments and network rules delegable without human intervention. The autonomous transaction stops being a demo and becomes a product.
The toilet pivot and the $100 trillion humanoids
The fever continues: after All Birds, now it’s Japan’s TOTO — yes, the high-tech toilet maker — whose stock rises after announcing an AI chip vertical. The humanoid robot market has more substance, projected at $100 trillion by 2035: its big advantage is reusing infrastructure already designed for humans.
Cursor SDK, Codex’s “/goal” and pets with a wink
On tools, two moves: the Cursor SDK lets you integrate background agents inside any product, and Codex debuts the /goal command, with which you define goals the model pursues autonomously until completion. And because it can’t all be productivity: from Claude Code’s /buddy to Codex’s new /pet, terminal pets include winks like an “angry Dario (Anthropic CEO).”
GitHub at 85%: the cost of agentic traffic
GitHub posted an uptime of barely 85% in April, choked by the massive traffic of agents working 24/7. The instability has prompted the departure of figures like Mitchell Hashimoto, and on top of that there’s a zero-day exploit in GitHub Enterprise that allowed privilege escalation with a simple push. The infrastructure that shaped modern software wasn’t sized for this new tireless user.
Demis Hassabis: from Move 37 to Edge models
The episode closes with a profile of Demis Hassabis, 2024 Nobel laureate in Chemistry and co-founder of DeepMind, revisiting the documentary The Thinking Game and AlphaGo’s legendary “Move 37.” In his latest interview, Hassabis predicts top video games created with vibe coding in under a year and bets on “Edge” models (like Gemma) running locally on phones and glasses: privacy and efficiency without going through the data center.
Thanks for joining us one more week on Tokenizados. Subscribe so you don’t miss any update from the world of AI.