Untangling Web3

🎙 Untangling Web3 The podcast that simplifies tomorrow's technology, today. Follow and subscribe to join our journey to untangle Web3.

Kalshi spent years fighting for CFTC oversight. That regulatory work - the kind the industry complains about - is the only reason this story has a "caught before he started" chapter at all. Polymarket had no equivalent. The gap between those two outcomes is exactly one KYC check.

A US soldier tried to bet classified intel on Kalshi. Kalshi's ID checks caught him and blocked his account. So he walked over to Polymarket and placed the bets anyway.

Army Specialist Kristoffer Teigland is now facing federal charges for using military intelligence to bet on whether Nicolás Maduro would be removed from power. One platform's compliance stopped him. The other had no door to close.

Running a tech event? Untangling can help with: • live podcast recordings • panel hosting + moderation • guest speaking • content capture for clips + social We bring real experience making complex tech topics into compelling live conversations — AI, blockchain, digital identity, IoT, and beyond. Looking for support at your next event? Let's talk: theuntanglingweb3podcast@gmail.com

Sam Bankman-Fried is building his pardon argument around exactly this. Hard to dismiss when the asset is Cursor and the buyer is Elon Musk. If they left $3 billion on the table here - what else did they sell?

FTX's bankruptcy estate sold a 5% stake in Cursor for $200,000 in 2023. SpaceX just agreed to buy Cursor at a $60 billion valuation. That stake is worth $3 billion today. The creditors got none of it.

The administrator's job was to recover money for victims as fast as possible. So they sold a $200,000 seed bet in an AI startup like it was a rounding error. The buyer of that stake just made a 15,000x return. FTX's creditors got cents on the dollar.

Struggling with AI trust? 🤖 Alec and Jack discuss how Sideck's agent-based system verifies AI model performance, generating proofs that anyone can check. It's like an OS with skills as apps, ensuring your workflows run as specified. This week on Untangling Web3, we’re joined by Leo Fan (@leofanxiong) - Founder & CEO of Cysic ( @cysic_xyz), a company building at the intersection of zero-knowledge proofs, AI, and high-performance compute.

In this episode of @untanglingweb3, our founder Leo @leofanxiong breaks down a core problem: From AI to blockchain, most systems still run as black boxes. However, verifiable compute and ZK change that. Want to know how? Listen here: youtube.com/watch?v=EPslS5…

Watch or listen now: Youtube: youtu.be/EPslS5Iuf94?si… Spotify: open.spotify.com/episode/3dt6l4… Apple: podcasts.apple.com/gb/podcast/unt…

🔎 What you’ll learn: 🔹 Why AI has a growing trust problem As systems become more autonomous, “just trust the output” stops working. 🔹 Why decentralisation alone isn’t enough Distributed compute ≠ correct compute. You need mathematical guarantees. 🔹 How zero-knowledge proofs change the game Prove a computation happened correctly—without revealing the data. 🔹 From DeFi → ComputeFi We verified money. Next step: verify the machines making decisions about it. 🔹 Why this matters now AI agents are moving from assistants → actors. That shift makes verifiability critical infrastructure. Feels niche today. But so did DeFi… before it didn’t. 👇 Question: Would you trust an AI agent to execute real actions without proof of what it did?

🚨 New Episode: Untangling Verifiable Compute with Leo Fan Watch now: youtu.be/EPslS5Iuf94?si… AI is starting to act, not just respond. It can execute trades, trigger workflows, move funds… But here’s the catch: 👉 You still can’t verify what it actually did. Most AI today is a black box. That’s manageable for content. It’s a real risk for finance, agents, and autonomous systems. So the real question becomes: Can we prove compute the same way we proved transactions? This week on Untangling Web3, we’re joined by Leo Fan (@leofanxiong) - Founder & CEO of Cysic ( @cysic_xyz), a company building at the intersection of zero-knowledge proofs, AI, and high-performance compute. With a PhD in cryptography from Cornell and deep experience across hardware and protocol design, Leo is focused on one core idea: 👉 making compute verifiable by default. @AlecJBurns @JackODavies_

If this theory holds, every AI company's safety team stops being a product function and becomes a legal defense team. The question isn't whether AI can cause harm. It's whether the company behind it can be charged for the output. That answer doesn't exist yet.

Florida AG James Uthmeier issued criminal subpoenas to OpenAI over ChatGPT's alleged role in planning a mass shooting. Not a civil suit. Not a fine. Criminal subpoenas - asking whether a software company can be held liable for what its model said.

Section 230 has protected platforms from user behavior since 1996. But that law was written for message boards. A model that actively generates a plan on request is a different thing entirely. Uthmeier is betting a criminal court will see the difference.

The $292 million cross-chain exploit didn't just drain one platform. It reminded everyone that DeFi protocols are stacked on top of each other. When one cracks, the question isn't "is my platform safe?" — it's "what is my platform built on top of?" That's a much harder question to answer.

=Billions left DeFi platforms in days — not because of a bank run, not because of a regulator — because one protocol got hacked and the fear spread on its own. That's not a crypto problem. That's a contagion problem.

TradFi spent decades building circuit breakers, deposit insurance, and lenders of last resort — all to contain exactly this kind of cascade. DeFi has none of that. The stress test arrived before the safety net did.

Hosting a tech event? Untangling can help with: • live podcast recordings • panel hosting + moderation • guest speaking • content capture for clips + social We bring hands-on experience making complex tech topics compelling — from AI and blockchain to digital identity, IoT, and beyond. Looking for support at your next event? Let's talk: theuntanglingweb3podcast@gmail.com