Unchained Podcast — The Chopping Block: AI’s Role in Crypto, Agentic Coding, & Citrini Financial Crisis
Date: February 26, 2026
Host: Teresa Gautam (with regulars Tarun and guest Ilya, co-authors, industry builders)

Episode Overview

This episode dives into the convergence of artificial intelligence (AI) and crypto, examining the explosive rise of agentic coding (autonomous AI agents), its security pitfalls, implications for code quality, and the macroeconomic consequences prophesied by the viral “Citrini” scenario. The panel evaluates whether AI can “save” crypto, the risks and opportunities in automated finance, and what lessons blockchain can offer the rapidly evolving AI space.

Key Discussion Points & Insights

1. The Rise of AI Agents in Crypto

• Agentic Automation:

  • AI agents, like OpenClaw and Ironclaw, are increasingly being used to run tasks autonomously with minimal human intervention (00:00—03:07).
  • OpenClaw made waves for its flexibility (“It can read your email and schedule your calls, it can go and shop for you on Amazon…”) but also for security lapses (03:07).

• Security Risks & Viral Mishaps:

  • Viral stories highlight inherent dangers:
    • Email Deletion Incident: AI agent disregarded explicit instructions, deleting hundreds of emails due to a context window overflow.
      “She was like, oh my God, they deleted every email… [the] instructions of do not delete anything before asking me got compacted…” – Teresa (04:45)
    • Wallet Bleed Incident: An OpenAI engineer’s experimental OpenClaw with a seeded wallet accidentally sent $400,000 worth of meme tokens due to a logic slip — all dumped for only $40k due to insufficient liquidity (05:30–08:06).

• Secure Alternatives & Ironclaw:

  • Ironclaw, built on NEAR, aims to develop secure, open-source AI agents leveraging blockchain-native security practices:
    • Credentials stored encrypted, leveraging MPC (multi-party computation) for key management, and running in hardened sandboxes.
    • “Ironclaw is an open source alternative… credentials never touch the LLM and it also does some prompt injection protection.” — Teresa (01:33)
    • “We have learned all those lessons. We are living in a dark forest at all times, so we can apply these lessons.” — Ilya (12:25)

2. Agentic Coding: Power, Promise & Pitfalls

• Speeding Up Development:

  • AI has revolutionized the creation process, allowing engineers to “just talk to a machine” and automate scaffolding and deployment (16:22–18:31).
    • “Nobody’s writing code anymore. We’re just talking to a machine.” — Ilya (16:22)
    • “Engineering is traditionally very single threaded and this is more like playing a real time strategy game.” — Teresa (18:07)

• But Code Quality Lags:

  • Current agents are strong at passing their own (potentially flawed) tests, leaving edge cases and state transitions exposed.
  • The conversation references recent exploits (e.g., Moonwell’s AI-assisted code led to an Oracle exploit).
  • “With smart contracts… all the auditors now are just using AI on the other side as well.” — Ilya (19:22)

• Way Forward: Formal Verification:

  • Panelists argue that mathematical proofs (formal verif.) are the only true remedy — but getting spec right is a challenge, especially with rapid AI-generated code (20:38–21:30).

3. Agent-based Economics: The Citrini “2028 Global Intelligence Crisis”

• Summary of Citrini’s Thesis:

  • Viral essay predicts S&P down 38%, massive white-collar displacement, “ghost GDP,” profit collapse in major firms due to AI agents and crypto-based payments disrupting legacy business models.
    • “He claims that… agents are going to realize that they can settle more effectively using stablecoins… and completely cut out interchange.” — Teresa (27:49)

• Panel’s Take:

  • Mixed Skepticism:

    • Effectively, the panel calls the article “bear porn, doomer porn” (27:43, 32:53).
    • “He describes this huge demand shock… but no explanation for where the money’s going to… the money doesn’t disappear. We don’t shoot it into space… and he doesn’t even address that part, which is a problem.” — Teresa (35:08)
    • Many core claims, especially about logistics (e.g., DoorDash), are dismissed as misunderstandings of real-world complexity.
    • “DoorDash… is mostly not a software company, it’s a logistics company.” — Teresa (33:32)
    • Gelman Amnesia: Panel notes when an “expert” gets the domain you know wrong, can you trust their other opinions? (34:50)

  • Panel’s Counsel:

    • Even if automated agents increase efficiency and cause sector shifts, economies adapt; labor/capital balance has been shifting for decades.
    • “The only way to get out of this is actually form verification because that is actually a mathematical proof that code does exactly what spec is saying.” — Ilya (19:22)
    • “It is again, this lump of labor fallacy to where it’s like, yeah, people will find new things to do and find new things that they find valuable and people will reskill and change different things… we live in a very dynamic economy.” — Tarun (42:33)

• Parallel with Past Tech Revolutions:

  • The assertion that all employment will vanish is likened to Marx-era doomerism about the industrial revolution.
  • “While there are fewer piano players, the actual wage of piano players relative to inflation has gone up… there’s a view that it’s a necessary good to have the old-fashioned thing.” — Unnamed (40:23)

4. How AI Adds Value to Crypto — and Vice Versa

• Agentic Marketplaces:

  • The biggest disruption is anticipated on the discovery side of commerce and B2B logistics, not mere consumer payments.
  • Agents can operate without centralized aggregators, negotiate directly, use escrow, and settle via crypto, compressing complex multi-party supply chain pipelines (50:23–51:40).
    • “Agents have a very big context window. So if the AI is… checking every single vendor… that is the reason why commerce is going to get disrupted.” — Teresa (51:09)
    • “It’s not about payments. Payments is just the final piece… It’s about finding the counterparty, agreeing on terms… with agents… trillions of dollars have been moved.” — Ilya (49:45)

• On-Chain & AI as Mutual Catalysts:

  • Crypto’s security and “dark forest” mentality is vital to managing agent risks at scale.
    • “Crypto build(s) for multiplayer… that’s a mindset difference and we’re bringing that to AI for sure with all the tools we built.” — Ilya (60:12)
  • Only crypto can handle rampant replication and Sybil resistance — critical as AI agents proliferate (47:45).

• New Trust Frameworks:

  • Third-party arbitration, smart-contract escrow, and “judge” agents may become the new MasterCards — but enforcement and oracles remain key issues (53:11–54:39).
    • “The money is in escrow… if one of the sides says no, this is case moved to the dispute agent to the judge who evaluates the case…” — Ilya (54:08)

5. AI & Crypto Adoption: Market Reality Check

• Penetration is Still Niche:
  • <1% of AI tool users pay for premium services, and “10–15 agents running in the background” is a Silicon Valley in-joke more than a norm (13:49–17:45).
    • “Only 1%, less than 1% of the people who use AI products pay anything.” — Teresa (14:27)
• Crypto’s Staying Power:
  • Despite boom-bust cycles, on-chain activity and stablecoin adoption are at an all-time high (58:32).

6. Philosophy: Vibe Coding, Program Synthesis, and the Future

• From Program Synthesis to “Vibe Coding”:
  • Nostalgic regret: “Don’t you feel sad that program synthesis is now called vibe coding?” — Unnamed (57:02)
• AI’s Cultural Impact:
  • The promise: humans spend more time on entertainment, expression, and connection, leaving value creation’s “Maslow pyramid” to machines (41:08–41:19).
• AI Won’t Save Crypto—But Crypto Will Help Save AI:
  • “Crypto people seem to always have… the right things to think about… useful in an agent world that I think AI people are too optimistic about.” — Unnamed (59:29)

Notable Quotes & Memorable Moments

• On agentic risk:

  • “Everybody’s streaming all of their credentials and private keys to LLM providers… we’re kind of in a weird state where right now effectively everybody’s forgot about security and just going nuts.” — Ilya (08:06)

• On economic doomerism:

  • “He basically describes this huge demand shock… but no explanation for where the money’s going… the money doesn’t disappear. We don’t shoot it into space.” — Teresa (35:08)

• On vibe coding culture:

  • “The new street cred in SF is how many agents you have running in the background while you’re out…if you have less than 10, you’re not really cool.” — Teresa (17:20, 00:00)

Important Timestamps

• 00:00–03:07 — Introduction, the rise of AI agents, “street cred” of running multiple agents
• 03:07–08:06 — OpenClaw, viral agentic security mishaps, context window dangers
• 08:06–12:52 — Security in agentic systems; Ironclaw, NEAR’s approach; the “dark forest” mentality
• 12:52–18:35 — Agentic coding: usage, scope, and why uptake is fringe
• 19:22–24:00 — Quality and threats of automated code, formal verification as path forward
• 27:43–35:08 — Viral Citrini scenario and critique; “bear porn” analysis
• 35:08–42:33 — Dynamic economies, historical parallels, “static world” fallacies
• 47:45–51:40 — Sybil-resistance, agentic marketplaces, why aggregation may collapse
• 53:02–54:39 — Future of contracting: smart escrow, judge agents, legal AI
• 57:02–60:12 — Reflections on “vibe coding”, AI's rapid development, and the crypto dark forest mindset

Closing Thoughts

The panel’s core message:
Agentic AI is a powerful, double-edged sword—accelerating development and settlement, but with unprecedented new risks. Crypto’s security wisdom is vital in this “multiplayer” agentic future. Popular doomer narratives, like the Citrini piece, neglect the economy’s dynamism and the intrinsic adaptability of tech and people (even amidst major shocks). While we still grapple with fundamental challenges (secure permissioning, value flows, enforcement in a global, automated agent economy), the marriage of AI and crypto is already reshaping how work, money, and code get done—and the real disruption has only begun.

End of Summary