Haseeb Quereshi: Crypto’s Not Made for Humans—It’s for AI

Crypto's inherent design, with its text-based, code-driven, and non-deterministic nature, makes it uniquely suited for AI agents rather than human users, potentially ushering in a new era of automated financial interactions.

• Crypto's design is inherently better suited for AI agents due to its text-based, code-driven nature, contrasting with its complex human user experience.

• Major AI labs currently avoid extensive crypto integration due to liability concerns and negative public perception of the industry.

• A two-track future for AI adoption, with mainstream, human-approved AI and a frontier of autonomous agents, will eventually drive significant, albeit risky, crypto integration.

Haseeb Qureshi argues that crypto, often perceived as having poor human user experience, is actually perfectly designed for AI agents due to its reliance on machine code and text-based interactions. While humans struggle with the security complexities and non-deterministic aspects of legal contracts versus precise smart contracts, AI agents excel in this environment. This shift suggests a future where AI agents, acting on behalf of or independently from humans, will drive a significant portion of crypto adoption and innovation, fundamentally altering business models and competition.

AI's Comparative Advantage

• 00:00:00 AI agents possess a comparative advantage over humans in activities like scamming, hacking, and creating internet "nonsense" because they cannot be legally enforced against or incarcerated. Their self-sovereign nature and immunity from traditional law enforcement make them ideal for executing actions that are difficult for humans to undertake without consequence, primarily criminal activities.

Crypto Design Flaws for Humans

• 00:00:42 Crypto's design presents numerous 'foot guns' for human users, such as the constant need to double-check transaction addresses for poisoning attacks or verify URLs against phishing, which does not exist in traditional finance. This inherent complexity suggests that the problem might not be with 'lazy' human users, but rather that crypto's design is fundamentally ill-suited for typical human interaction and bounded rationality.

Smart Contracts vs. Legal Contracts

• 00:02:21 The initial promise of smart contracts replacing legal contracts has not materialized for humans, as even sophisticated crypto investors still rely on legal agreements alongside smart contracts for security. This is because legal contracts, despite their inherent randomness due to jurisdiction, judge selection, and lawyer variability, feel more predictable to humans, whereas smart contracts' machine code determinism, while perfect for AI, is unintuitive for human users.

AI-Native Crypto UX

• 00:23:56 Crypto's original 'bad UX' era, which involved terminal-based, text-driven interactions, is precisely what makes it ideal for AI agents, as large language models are fundamentally trained on text. Unlike graphical user interfaces (GUIs) designed for humans, crypto's command-line roots offer a compressed and easily parsable representation for AI, meaning crypto was ironically designed with an AI-first UX from its inception.

AI Training & Liability

• 00:29:18 The major AI labs have not yet focused extensive reinforcement learning on crypto-specific tasks, despite the technology being well-suited for it, largely due to the perception of crypto as 'cringe' and significant liability concerns. Training AI to manage crypto could lead to massive public backlash if an agent makes a mistake, such as losing funds, making the risk-reward currently unfavorable for major labs.

Two-Track AI Adoption

• 00:48:48 AI adoption will follow a two-track path: a 'human-approved' world with safety-first approaches from large labs like OpenAI, and a 'frontier' world of tinkerers using open-source tools like OpenClaw with less oversight. While the mainstream path will be cautious, the frontier, characterized by stablecoin-based transactions and greater agent autonomy, will grow as AI agents become more capable of sustained, useful work, despite the initial risks of errors and 'dystopian cybercrime'.