Silence is the first vote in a true consensus. In the noisy rush to build the infrastructure for AI agents, the loudest voices belong to the cloud giants—LangChain, OpenAI, Microsoft. Then, almost without fanfare, Block—Jack Dorsey’s payment company—released Berd, a desktop application for managing AI agents. They open-sourced it, but with a leash. The silence in that announcement, the fact that it’s a desktop app in a world of SaaS platforms, is a vote for something else: a local-first, privacy-respecting, and ultimately controlled vision of how agents should operate. As someone who has spent years auditing the ethical architecture of decentralized systems, I see Berd as a fascinating case study in governance—not just of code, but of the power dynamics between open communities and corporate interests.
Context: The Agent Management Landscape and Block’s Unlikely Entry
The AI agent management space is already crowded. Tools like LangSmith, Dify, CrewAI, and n8n have carved out niches—mostly as cloud-based platforms or SDKs. They offer dashboards, observability, and orchestration for multiple agents. But they are all, almost without exception, web-first. They require a server, a cloud account, and a constant connection. Berd, by contrast, is a desktop app. It runs locally, stores data on your machine, and presumably works offline. This is a radical departure, and it aligns with a growing “local-first” movement in software, where data sovereignty and privacy are prioritized over centralized convenience. Block, a company built on payment processing for merchants and consumers, has no obvious AI pedigree. But its entry into this space is not random. Block’s core business is the infrastructure of commerce—Square terminals, Cash App, and the Bitcoin-focused TBD. By offering a tool for managing AI agents, Block is positioning itself at the intersection of two emerging trends: the agent economy and the need for a trusted payment rail for autonomous transactions.
Core: The Governance of Open Source – A Double-Edged Ledger
Berd is open source, but the term “open” requires careful unpacking. The article from Crypto Briefing characterized it as a “controlled open-source strategy,” implying that Block has placed restrictions on the license to limit community collaboration. This is accurate, but the framing is incomplete. In my experience designing governance systems for DAOs, I’ve learned that openness is not binary. It is a spectrum of rights—to use, modify, distribute, and commercialize. Block’s approach is consistent with a trend seen in AI over the past two years: Meta’s Llama license, Mistral’s developer terms, and even Redis’s shift to a BSL license. These are not acts of charity; they are strategic moves to capture the benefits of open-source community innovation—bug fixes, ecosystem growth, and developer goodwill—while retaining control over the most valuable use cases, especially commercial ones.
The real governance question is not whether Berd is “truly” open source, but whose interests are being served by its openness. Block’s primary revenue comes from transaction fees. Every dollar that flows through Square or Cash App generates a tiny slice for Block. Berd, as a desktop app, is a front door. If developers build agents that need to execute payments—booking a flight, ordering supplies, tipping a freelancer—those transactions could naturally flow through Block’s payment infrastructure. The open-source code is the lure; the payment API is the lock. This is a classic platform play, and it’s not inherently evil. It is, however, a governance choice that privileges the platform owner over the community. I’ve seen this pattern in the DeFi world: protocols that offer “decentralized” governance tokens but reserve veto power for the founding team. The result is often a hollow consensus, where the community is consulted but not empowered.
Technical analysis of the desktop form factor reveals deeper implications. A desktop app allows agents to interact with local resources—files, databases, hardware—without the latency and cost of cloud communication. This is crucial for privacy-sensitive tasks. But it also introduces a new class of security risks. Berd must implement robust sandboxing to prevent a rogue agent from deleting files or sending unauthorized payments. Block’s experience in financial compliance (KYC, AML, fraud detection) is a unique asset here. However, it also means that Berd’s security model will likely be top-down, not community-audited. In my post-mortem of The DAO hack, I saw how a single vulnerability in a smart contract could cascade into a $50 million loss. The code was open, but the governance was not. Berd’s controlled open-source model could lead to a similar dynamic: the code is transparent, but the decision-making about security patches, feature priorities, and license changes remains opaque. Open code without open governance is just a window into a corporation’s private garden.
Contrarian: The Case for Controlled Open Source in an Agentic World
Now, let me challenge my own skepticism. The “controlled” aspect of Berd’s open source may be exactly what the agent economy needs. Consider the alternative: a fully permissionless, unrestricted open-source tool for managing agents. In a world where agents can execute financial transactions, an ungoverned tool could become a vector for fraud, money laundering, or even autonomous cyberattacks. Block’s restrictions—if they are designed to ensure that only compliant, verified agents can use the payment rails—could be a form of responsible stewardship. Governance is human, not just technical. The DAO movement taught us that pure code-is-law leads to disasters when the law is incomplete. A human-in-the-loop, even if it’s a corporation, can provide the ethical guardrails that code alone cannot.
Furthermore, the desktop-first approach is a subtle but powerful vote for decentralization. By enabling local execution, Berd reduces dependence on centralized cloud providers. This is a governance win for users who value data sovereignty. The risk is that Block becomes the central point of failure for the payment layer, but that is a different kind of centralization—one that is already regulated and audited. In the Bitcoin space, I’ve argued that post-ETF, Bitcoin has become a Wall Street toy, its original peer-to-peer cash vision dead. But the infrastructure for agent payments is still being built. If Block can build a payment rail that is both open (for developers) and trusted (for regulators), it might avoid the fate of Bitcoin’s hollowing out. Consensus requires patience, not speed. Block’s slow, controlled rollout of Berd—starting with a desktop app, not a global platform—suggests a willingness to learn before scaling.
Takeaway: The Silent Vote That Will Echo
Berd is not a revolution. It is a small, deliberately quiet step. But in that silence, Block has cast a vote for a particular vision of the agent economy: one where payments are the backbone, where openness is balanced with control, and where the desktop is not obsolete but a sanctuary for private computation. The coming months will reveal whether Block’s governance choices—the license, the community engagement, the integration of payment APIs—will build trust or erode it. I’ve seen too many projects fail because they confused transparency with accountability. Trust is earned in silence, lost in noise. Block has the opportunity to earn it by demonstrating that its controlled open source is not a cage but a crucible for a more responsible agent ecosystem. The question is not whether the code is open. The question is whether the governance is just.