Jack Dorsey’s Block just open-sourced Berd, a desktop AI agent manager. The crypto-native response: “Another controlled open-source play.” They’re missing the point. Berd isn’t a developer tool — it’s the first payment rail for autonomous agents.
Context: Block is a payments company. Square, Cash App, and the Bitcoin infrastructure underpin hundreds of billions in annual transaction volume. AI agents are the next frontier of commerce — they can book flights, buy subscriptions, and execute trades. Managing these agents requires a new layer of infrastructure. Today, that layer is dominated by cloud-based tools like LangSmith and Dify. Berd is a desktop app. It’s local-first. It’s open-source, but with strings attached. The industry calls it “controlled open source.” I call it strategic positioning.
Core: The technical architecture of Berd is unremarkable — a combination of a desktop GUI framework, an agent orchestration engine, and local data persistence. The novelty lies in what’s missing. No cloud version. No enterprise plan. No mention of multi-user collaboration. This is a deliberate choice. Block is not trying to compete with LangChain on features. They’re building a sandbox for developers to experiment with agent-controlled payments. The real code is not in the repo — it’s in the APIs that connect to Square’s merchant network and Cash App’s consumer wallet. Based on my experience auditing 50+ Web3 projects, the controlled open-source model is a sign of maturity, not greed. It allows Block to capture community contributions while retaining ownership of the monetization layer: the transaction fee. The narrative is shifting from AI tooling to AI commerce. Berd is the Trojan horse. Developers adopt it for its simplicity and local privacy. But the moment an agent needs to pay for a service, the gateway leads to Block’s infrastructure. This is the same playbook as Square’s early days: give away the card reader, charge on the transaction.
Hunting for the story that defines the next cycle: The timing is no coincidence. Stripe launched its agent toolkit in June 2025. Both companies see the same future: autonomous agents will initiate trillions of dollars in transactions. The winner will be the infrastructure that agents trust to hold and move money. Block’s regulatory moat — KYC, AML, compliance — is a structural advantage that pure AI tooling companies cannot replicate. The contrarian angle: The dominant criticism of Berd is its “controlled open-source” license. Critics call it open-washing. But that misses the forest for the trees. In the age of agentic commerce, the license is irrelevant. The moat is the payment network. Elastic and Redis suffered community backlash because their products were standalone infrastructure. Berd is a gateway to a network. The real risk is not license backlash — it’s whether Block can execute fast enough before agents become native to centralized platforms like OpenAI or Google. The signal is the payment rail, not the open-source license.
Takeaway: The next cycle’s defining narrative won’t be about which AI model is smarter. It will be about which infrastructure can handle the first trillion machine-to-machine transactions. Keep your eyes on the payment rails, not the code. Hunting for the story that defines the next cycle.