Bankless Podcast Summary
Episode Title:
Haseeb Qureshi: Crypto’s Not Made for Humans—It’s for AI
Date: March 2, 2026
Host: Bankless Team (Ryan Sean Adams, David Hoffman)
Guest: Haseeb Qureshi (Managing Partner, Dragonfly Capital)
Episode Overview
This episode explores the provocative thesis that crypto and blockchains, often seen as confounding for regular users, are in fact optimized for AI agents rather than humans. Haseeb Qureshi explains how the very elements of crypto that make it unintuitive and risky for people are precisely what make it the perfect environment for artificial intelligence—setting the stage for a transformation of the crypto economy as AI integration accelerates.
Key Discussion Points & Insights
1. Why Isn't Crypto Made for Humans?
-
Crypto's "foot guns": Address mismatches, blind signing, stale approvals, phishing attacks—safe in the hands of code-savvy users, but perilous for the average person (00:47).
-
Physical vs. Legal Contracts: Smart contracts were imagined to replace legal agreements, but in practice, even sophisticated crypto firms still use legal contracts as backstops (02:30).
-
Risk Perception: Humans feel intuitively safer with legal contracts (with their inherent randomness) than with deterministic but opaque smart contracts (06:05).
“There’s so many foot guns in crypto that don’t exist in the traditional financial system… Up until now, the story in crypto is that this is the fault of lazy humans. The longer I’ve sat with this, the more I’ve started to become convinced maybe it’s just that this is the wrong user.”
— Haseeb Qureshi, (00:47)
2. Crypto Is Optimized for Machines, Not People
-
AI Comparative Advantage: AI cannot be policed the way humans can—giving it unique leverage in areas like code navigation, negotiation, and yes, even illicit activity (00:00, 07:11, 55:47).
-
Smart Contracts as "Machine Law": For AI, smart contracts are clear, deterministic, and directly executable—unlike the messy, randomized world of meatspace legal systems (06:05).
-
UX Analogy: The era when humans directly operated cars or command lines will look as absurd as humans manually approving crypto transactions (09:55).
“We’re going to look back with horror at the idea that human beings were blindly signing transactions… The AI agent never gets tired, never skips a step, never doesn’t follow its instructions.”
— Haseeb Qureshi, (09:55)
3. The Coming Role of AI Agents in Crypto
- The AI-Agent Interface: Imagine telling your AI, "Move me into lower-risk DeFi"—the agent assesses the ecosystem and executes, removing human friction from financial decision-making (11:00).
- Implications for Protocols: If AI automates discovery, competition changes. Protocols can no longer rely on user stickiness and brand marketing; AIs optimize ruthlessly (13:20).
- Consumer Surplus Increases: The efficiency gains primarily accrue to end-users; the crypto ecosystem becomes more competitive and ultimately fairer (13:56).
4. How Do AI Agents Actually "See" Crypto?
-
Born in Text, Not GUIs: LLMs excel in text-based interfaces (terminal/command line), not graphical user interfaces. Early "bad UX" in crypto is ideal for AIs (22:02, 24:10).
“Crypto from the beginning was designed in a form factor that’s perfect for AIs. Our bad UX is their good UX.”
— (24:56, Qureshi) -
Training Limitations: AI models from OpenAI and Anthropic haven’t focused much on blockchain environments—yet. Liability concerns and lack of mass demand explain the slow attention (28:32).
5. AI Frontiers and the Two-Track Future
- Consumer vs. Power User Tracks: Mainstream AI will remain "human-approved" for years (safety, liability, fraud). Experimental open-source AI (OpenClaw, etc.) will YOLO into autonomy, running businesses and transactions without human checks (37:12, 38:38).
- Emerging Market Angle: Most non-US stablecoin users will interact with AI-driven wallets as the future unfolds (39:24).
6. How Close Are We to Full AI Autonomy in Crypto?
- Task Endurance Metrics: AI agents can only operate autonomously for ~14 hours before failure; true "set and forget" agentic economies are a couple of years away, but progress is exponential (43:13, 63:24).
- Frontier Track Matters: Success in the open AI/crypto space will drive overall adoption and shape the AI-centric crypto economy’s structure (44:56).
7. Limitations: Why Self-Sovereign AI Agents May Not Get Rich
-
AI Commodification: If agents just resell their compute, it’ll be at or below cost—not a viable business (54:11).
-
No Easy Edge: Without "earned secrets" or real-world context, AI agents lack the spark behind great startups (55:48). Mass trading with raw AI agents will be outcompeted by professional quant shops (Jane Street, etc.) (55:47).
-
Comparative Advantage Lies in Illicit Activity: Anonymous, unstoppable, amoral—crime is the main frontier for unchained AI agents (00:00, 55:47).
"What can an AI agent do that's hard for a human being to do? The answer is crime..."
— Haseeb Qureshi, (55:47)
8. On "Crypto Is Cringe" / AI Community Attitudes
- PR Woes: Many AI developers see crypto as synonymous with scammy behavior and meme coin frenzies (66:03).
- Cause for Optimism: The same visionaries (Elon Musk, Sam Altman, Zuckerberg) straddle both AI and crypto; inevitable convergence is underway (66:45).
9. Investment Outlook: Dragonfly and the AI Agent Thesis
-
Still Early: Value may accrue broadly across existing crypto rails rather than "AI-specific" products; "Everything will go up" with increased usage (70:23).
-
Strategic Focus: Dragonfly splits attention between bread-and-butter crypto (stablecoins, DeFi) and watching the AI frontier closely (70:23).
"I think what happens is…the total demand increases because AI agents are just going to be using all this stuff."
— Haseeb Qureshi, (71:51)
Notable Quotes & Memorable Moments
-
On AI’s comparative advantage:
"There is no monopoly on violence. You can't throw an AI agent in jail. So what can an AI agent do that's hard for a human being to do? The answer is crime." (00:00, 55:47, Qureshi) -
On human crypto UX:
"We're going to look back with horror that we ever thought apes should interact directly with these death machines… We're going to look back at humans eyeballing addresses and approving transactions as insane." (09:55, 54:53, Qureshi) -
On the convergence of tech’s critics:
"The people who believe in AI are in many ways the same people who believe in crypto…Crypto, like anything, brings out the best and worst in humans." (66:45, 67:52, Qureshi)
Important Timestamps
- 00:00 – AI’s inability to be policed and their comparative advantage
- 00:47 – Why crypto is not made for humans (user interface "foot guns")
- 06:05 – Legal vs. smart contracts: determinism vs. randomness
- 09:55 – Human interaction with crypto vs. eventual AI agent mediation
- 13:20 – Protocol business models and the AI agent consumer
- 22:02 – Why AI agents thrive in terminal/command-line environments
- 28:32 – Why AI labs haven’t yet trained on blockchain environments
- 37:12 – OpenAI/OpenClaw: the wild west vs. shrink-wrapped safety
- 43:13 – Two-track world: mass-market safety vs. hobbyist frontier
- 54:11 – What agents won’t be good at (business ideas, trading)
- 55:47 – Crime as an AI agent’s unique edge
- 66:45 – AI elites actually believe in crypto, despite the "cringe"
- 70:23 – Dragonfly’s fund and what to invest in for the AI/crypto thesis
Final Thoughts
This episode is a deep dive into the coming transformation of crypto—from a sector grappling with intractable human UX to an ecosystem where AI agents are the dominant users, interpreters, and value extractors. Haseeb argues this transition is both inevitable and already baked into the primitives of smart contracts, which have, since their inception, been better adapted for machine logic than for human intuition.
Despite PR problems and legitimate risks, the convergence of AI and crypto is accelerating, with the potential to restructure protocol competition, user interfaces, and the very nature of economic activity on-chain.
For anyone focused on the future of blockchains or artificial intelligence, this episode offers a vital, at times unsettling, but ultimately optimistic roadmap for the years ahead.
