From Open Ledgers to Opaque Algorithms: Block's AI Expansion and the Soul of Decentralized Code

When Block reported its latest earnings, the market barely blinked. Cash App and Square delivered better-than-expected results, and the stock ticked up. But buried inside the shareholder letter was a quiet confession: the company is expanding its use of AI across software engineering. Not just for chatbots or customer support, but for writing, reviewing, and shipping the actual code that moves billions of dollars. The market sees efficiency. I see a moral dilemma. For a company that once positioned itself as the friendliest bridge to Bitcoin, Block is now feeding its most important infrastructure into a centralized, opaque, and deeply un-auditable intelligence. That feels less like a bridge and more like a wall. The context matters more than the quarter. Jack Dorsey built Square on the promise of democratizing access to commerce for small sellers. Then Cash App made Bitcoin accessible to millions of Americans, including the ones who had never bought crypto before. In a very real sense, Block became the on-ramp for a generation of users who would eventually learn about self-custody, cold storage, and the quiet joy of holding a private key. Even amid the bear market, Block has held a significant amount of Bitcoin on its balance sheet, signaling a corporate commitment that goes beyond mere treasury management. The company has consistently argued that Bitcoin is not a speculative asset, but a foundational technology for financial sovereignty. That is the narrative that attracted many of us to watch Block's earnings with more attention than we might pay to a typical fintech behemoth. Now, in the middle of a sideways market, the company is telling us a different story. The headline numbers are solid: Cash App continues to grow its gross profit, Square is recovering as consumer spending stabilizes, and the company is cutting costs through AI-driven development. What's not in the numbers is the philosophical pivot. By using AI to write production code, Block is placing its trust in a stochastic engine trained on the vast corpus of human software. It is essentially saying: the best way to maintain a decentralized product is to centralize the creative process. The ledger may be open, but the mind behind it is closed. I have spent the last twelve years watching code evolve from a niche craft into the infrastructure of modern finance. Based on my audit experience, going back to 2017 when I manually reviewed the smart contracts of ICO projects from my apartment in Tokyo, I can tell you that the most dangerous bugs are never the ones you can see. They are the ones that hide in plain sight because the developer's logic seemed reasonable at the time. In the early days, I would trace every line of token distribution code back to its founding assumptions, looking for the disconnect between what the white paper promised and what the ledger actually delivered. That exercise taught me something crucial: code is a mirror of conscience. When a developer writes a function, they are not just manipulating state; they are codifying a value judgment about who gets what, when, and why. Now imagine that developer is a large language model. The mirror becomes a funhouse. The AI does not reason about fairness, transparency, or the spirit of decentralization. It predicts the next token based on billions of examples of existing code. That means it will reproduce the same biases, the same security holes, and the same centralizing tendencies that have always lingered in traditional software. Worse, when AI generates code, the original logic is fragmented across training data. No one truly knows why a particular decision was made. The audit is not the end, but the beginning of a deeper mystery. Let me be precise about the technical trajectory. In a conventional software engineering workflow, a developer writes a function, submits it for review, and a human reviewer questions the assumptions. The code goes through a formal, traceable chain of custody. With AI-assisted development, that chain is broken. The AI might generate a cleverly optimized smart contract that passes all unit tests, but the structural semantics may be unintentionally fragile. This is especially dangerous in the blockchain space, where code is immutable. A bank can roll back a faulty transaction. A smart contract cannot. If Block's AI accidentally introduces a bug into a Cash App bitcoin transfer flow, users will lose funds, and the open ledger will only record the loss, not explain it. The irony is that the broader crypto market is currently obsessed with the wrong sorts of innovations. Let's look at BRC-20 and Runes. These token protocols on Bitcoin have captured the imagination of a retail audience looking for the next meme-driven surge. But anyone with a technical eye can see the issue: trying to force fungible tokens onto Bitcoin's UTXO model is like using a Rolls-Royce to haul cargo. It insults the vehicle and doesn't carry much. Bitcoin is designed for a specific kind of robustness, not as a general-purpose data chain. The recent hype has produced a lot of transactional noise, but very little structural progress. Meanwhile, the industry is also overhyping data availability layers for rollups. The uncomfortable truth is that 99% of rollups don't generate enough data to need a dedicated DA layer. They are building highways for bicycle traffic. Those two trends share a common flaw: they prioritize speculative utility over foundational integrity. Block's AI move is different, but equally concerning, because it attacks the integrity of the development process itself. We are heading toward a world where the code that protects our digital sovereignty is written by an algorithm that does not care about sovereignty at all. It only cares about probability. That is not a recipe for open books, open ledgers, open hearts. Let me offer a contrarian angle. Maybe Block is onto something. Perhaps AI-assisted code generation could actually make decentralized development more accessible. A young developer in Nairobi or Manila could use a large language model to write a robust audited smart contract in minutes, without having to spend years learning low-level Solidity vulnerabilities. In that scenario, AI is a democratizing tool. It lowers the barrier to entry and spreads the ability to participate in blockchain development. The open-source community could then review the generated code collectively, using the transparency of the ledger to compensate for the opacity of the model. That is a beautiful vision. It aligns with the idea of building bridges where others build walls. But it rests on a fragile assumption: that the AI itself is neutral and that the underlying training data is trustworthy. Given that most available training data comes from centralized repositories, corporate codebases, and mostly Western developers, the model will inherently carry the biases of the privileged. Making that code widely available just amplifies those biases. The other contrarian angle is that Block is not actually compromising its principles. It may simply be using AI to automate peripheral tasks, like unit test generation, documentation, or simple refactoring, while keeping core financial logic under human review. If that is the case, my concern is exaggerated. The company's shareholder letter uses the phrase "expanding its use of AI across software engineering," which is vague enough to mean anything. It could mean developer assistants, code review bots, or end-to-end autonomous programming. The lack of specificity is itself a warning. If the company were proud of keeping humans in the loop, it would say so. Instead, it announces the expansion of AI as a cost-saving measure, which suggests the goal is fewer engineers per feature, not more careful engineering per feature. I have seen this movie before. In the 2020 DeFi summer, I launched a project called ChainLit, a volunteer-run digital library aimed at helping non-technical Tokyo residents understand liquidity pools and yield farming. I was full of enthusiasm and managed three Discord servers while writing forty guides in a month. The project failed because I prioritized speed over sustainability. I burned out, the content quality degraded, and the community drifted away. The lesson was painful: evangelism requires structure, not just inspiration. The same principle applies to code. An AI that writes a thousand lines of code per minute is an evangelist without structure. It can produce a flood of information, but it cannot produce trust. Trust takes a human beating in a shared context, with predictable schedules and lasting relationships. That is the real risk of Block's AI push. It is not that the AI will write bad code. AI is already better at catching certain edge cases than many humans. The risk is that the AI will write code that no one fully understands, and we will lose the ability to trace that code back to a conscience. The whole moral foundation of blockchain is that code can be read like a contract and audited like a confession. If the code is generated by a black box, we are back to the original problem that Bitcoin was created to solve: trusting a central party. The only difference is that the central party is no longer a bank; it is a neural network with a billion parameters. Let's zoom out. The market is currently sideways, and that is exactly the moment when institutional decisions matter the most. Choppiness is for positioning. The price action teaches us nothing new. But this earnings report is richer. Block is managing to grow its traditional fintech business while quietly embracing the most centralized technology since the mainframe. It holds bitcoin on its balance sheet, but it is using AI to centralize the development pipeline. The two impulses are in direct conflict, and that conflict is a mirror of the wider crypto industry. We want open ledgers, but we also want fast, polished apps. We want trustless systems, but we also want efficient developers. We want decentralization, but we are willing to outsource the thinking to a company that does not share our values. My takeaway is not that Block is evil, or that AI is going to destroy crypto. It is that the culture of the builder is more important than the tool. Culture is the ultimate consensus mechanism. When a community values auditable code, open review, and human responsibility, it can survive even a misaligned AI. When that community is replaced by automated systems, the consensus dissolves into the noise of the training set. The ledger stays open, but the heart is gone. I started this article with a question about Block's earnings, but the real question is for all of us. In the pursuit of efficiency, are we willing to sacrifice the auditable soul of our own code? Or will we build bridges that preserve human accountability, even as we adopt the intelligence of machines? The next bear market will not kill crypto. The next centralized AI just might, not by attacking the price, but by rewriting the spirit. Let's trace the code back to the conscience while we still can.