Allocators say crowding is up but dispersion rewards fundamentals, with Hyperliquid cited as proof

Crypto x AI: coordination and privacy over raw efficiency

Arthur Cheong of Defiance Capital said the crypto-AI overlap struggled with "false fitting" since 2023 but has matured, pointing to Nous Research's Hermes open-source agent framework and Venice AI, a privacy-focused inference router he compared to Open Router, as examples with "hockey stick" growth even if token spend remains smaller than competitors. Rajiv Patel-O'Connor of Framework Ventures said AI is giving startups operational leverage, and highlighted portfolio company USDAI, which finances GPU cluster purchases for startups or public companies, as a crypto financing opportunity tied to AI capex buildout.

Smac of Compound argued crypto AI shouldn't try to compete on efficiency against centralized labs, but on coordination, permissionlessness and different economic models. Arthur added that crypto-AI projects tend to build in encryption and privacy by default, something he thinks the broader AI world will eventually need.

On whether AI agents will actually use blockchains, the host pushed back on this being "force fitted in retrospect." Rajiv floated distributed ledgers for agent-swarm communication, suggesting someone should "crawl Arweave or Filecoin" to see if agent swarms already use them. Arthur said permissionlessness cuts both ways: it could let a "rogue AI agent" exploit the system, but also noted a rogue agent "is not going to use a bank API... it's going to use crypto."

Sizing into illiquid, consensus-crowded positions

The host framed the current era as a barbell between venture and liquid allocation, especially around a privacy narrative where assets like Zcash and Near-adjacent tokens are publicly tradable yet hard to size into. Arthur said building size hasn't been a major issue below roughly $500 million to $1 billion in AUM if investors accept slippage, budgeting for "a 5 to 10% variance" on entry and exit. Rajiv agreed, saying sizing is harder when liquidity is thin or teams won't do OTC deals, and noted crowding means ideas become consensus faster as participants converge on similar research.

Smac said Compound shifted from pure seed/pre-seed investing toward liquid exposure partly because the asset class moves faster than a handful of private deals per year allow, and emphasized sizing positions to target "venture style returns" rather than levered beta exposure.

Dispersion rewards fundamentals, Arthur calls it the "golden age" of long/short

Arthur said this has been the best year in recent memory for active managers because dispersion is finally being rewarded: he estimated more than 80% of liquid crypto hedge funds he knows have outperformed Bitcoin and ETH, with most funds up "20 to 100%" year-to-date even as the majors lagged. He pointed to Hyperliquid as evidence that an asset can stay consensus among "smart money" for over a year and keep outperforming rather than peaking, calling it "incredibly encouraging."

"This is a year that... on the street I think probably more than 80% of the liquid funds I know... outperform Bitcoin, ETH." — Arthur Cheong

Arthur also argued this is a "golden age" for long/short strategies in both crypto and equities, since failing projects can be shorted repeatedly as they keep dropping roughly 50% every six months, something he said surprised even fundamentally-driven investors. He cited how bets against the US Treasury kept paying off despite looking "oversold" on technicals. On whether Bitcoin remains the right benchmark, Arthur said it still is for most LPs since alternative indices like CCI 30 haven't gained traction, and that LPs care mainly about absolute returns and Bitcoin outperformance. Rajiv and Smac offered softer views, with Smac noting ETFs like IBIT set a "bare minimum baseline" funds must beat.

Seed versus growth: founder quality and AI's effect on capital needs

Arthur said late-stage deals in both equity and crypto suffer from heavy overlap among investors chasing scarce opportunities, while seed stage has more opportunity precisely because "everyone got so burned," though founder quality is harder to find as AI draws away talent. He argued crypto and AI are among the only sectors where a seed investment can still return 50x within one to two years.

Rajiv disagreed somewhat, saying the number of outlier founders at early stage "has not dramatically decreased" even though total seed-stage founder count has fallen due to right-sizing after prior overcapitalization, and noted growth-stage companies now show more durable traction rather than pure narrative-driven raises. Smac said Compound will stick to pre-seed and seed, arguing some previously "written off" crypto sectors may simply not have had the right timing or founder yet.

On capital intensity, Arthur asked whether AI lets crypto startups need fewer engineers, potentially enabling billion-dollar companies run by solo founders. Smac said this is already happening, with AI-native team members retained and others seen as "not net additive," while Rajiv said a chunk of the typical $5–10 million seed round once spent on headcount is shifting toward compute and API credits, though enterprise sales and marketing still require people.