Where Is The Crypto Alpha Hiding This Cycle?

Nous Research

• Arthur Cheong · DeFiance CapitalBullish

  Nous Research, which builds the leading open-source agent framework Hermes, is now seen as a legitimate AI team.

  Arthur cites Nous Research as the developer of Hermes, which he calls the largest open-source AI agent framework. He notes the team had been building crypto-AI projects and is now considered a legitimate AI team. He presents this as evidence that crypto-AI teams are gaining credibility.

  “the developer of the largest open source AI agent framework, Hermes, is actually built by Nous Research... And I think that now they are, like, being considered very legitimate AI team as well.”at 2:07

Venice.ai

• Arthur Cheong · DeFiance CapitalBullish

  Venice AI, a privacy-focused crypto inference router, is taking off with a hockey-stick growth curve.

  Arthur describes Venice AI as essentially a crypto version of OpenRouter, where users route inference with a privacy focus. He says it has been really taking off. Although its token throughput is still far below OpenRouter's, its growth curve is still a hockey stick, which he cites as a prime example of crypto-AI traction.

  “Venice AI, which is essentially a crypto version of open router where you can route inference with privacy focus... And I think that that actually has been really taking off... the growth chart is still a hockey stick.”at 2:30

USDAI

• Rajiv Patel-O'Connor · Framework VenturesBullish

  Crypto-backed USDAI offers better GPU cluster financing terms than conventional lenders.

  Rajiv argues crypto is good for capital formation, and the AI capex build-out needs financing. USDAI, one of his firm's portfolio companies, does mid-market cluster financing. A startup or public company buying about $150M of GPUs can get more attractive rates and a smaller down payment than elsewhere.

  “We think crypto is really good for capital formation... One of our portfolio companies, USDAI, is doing like mid market cluster financing... you can go to them and potentially get more attractive rates with less of a down payment relative to what you might see elsewhere.”at 3:15

Hyperliquid

Across the shows: 1 speaker in other episodes, 0 bullish and 1 bearish calls, past 30 days

• Arthur Cheong · DeFiance CapitalBullish

  Hyperliquid stays a consensus name yet keeps outperforming because its fundamentals are strong.

  Arthur notes Hyperliquid has been a consensus name for more than a year, and historically consensus names in crypto tended to peak and underperform. Hyperliquid still keeps rising and outperforming despite its high market cap and hype. He argues that strong fundamentals and growth can outperform even when smart money is aligned.

  “Hyperliquid, it was been a consensus name for more than a year... it's very consensus and yet it still continue to go up and outperform... It's just because an asset is consensus among the smart money doesn't mean it can't outperform anymore.”at 20:55

Venture capital & startups

• Arthur Cheong · DeFiance Capital

  Late-stage crypto and AI rounds offer about 50x upside in 1-2 years, with little competition.

  Arthur says late-stage deal activity is sparse, and everyone got burned, so C-stage is a lot of opportunity right now. Founder quality is hard to find, and AI is pulling good founders away, so investors underwrite a lot of risk. He argues that crypto and AI can deliver about a 50x uplift in one to two years at the C stage, which he says few other industries can match.

  “I think C stage is a lot of opportunity right now because everyone got so burned... you can get the 50x uplift pretty fast. I think you can get it in one to two year if you've done the C stage”at 27:56

• Rajiv Patel-O'Connor · Framework Ventures

  Outlier seed founders remain, and durable growth-stage traction favors more seed investing.

  Rajiv argues the number of outlier founders has not dramatically decreased, though fewer seed founders exist after right-sizing. He sees growth-stage companies showing more durable traction, not needing to raise on narratives. His hunch is there is more opportunity at the seed stage, and growth generalists will likely keep entering as these companies compound.

  “The number of outlier founders has not dramatically decreased... I still think that like the number of extremely high quality founders at early stages is still there... It seems like a lot of the growth stage companies have traction that's like quite durable versus like needing to raise purely off narratives.”at 29:05

• Rajiv Patel-O'Connor · Framework Ventures

  AI credits are replacing startup headcount, so a $10M seed round stretches further unless training models.

  Rajiv says a large share of early-stage capital historically went to headcount. A meaningful part of that is now shifting to compute and API credits. Enterprise sales still need people, but a $10M raise gets a company much further, unless it trains its own models, in which case it does not go far at all.

  “historically, a huge chunk of that has gone to head count. A meaningful chunk of that head count is probably being reallocated to compute or API credits and stuff like that... that 10,000,000, I think, gets you a lot further unless you are training auto models”at 32:03

AI & jobs

• Rajiv Patel-O'Connor · Framework Ventures

  AI gives early-stage companies operating leverage, letting them progress without adding headcount.

  Rajiv argues AI is transforming how companies operate. He says his firm is seeing operational leverage it had not seen before, especially at early-stage companies. These companies can make much more progress without hiring many more people or adding headcount spend.

  “AI is obviously transforming the way that companies operate. So we're seeing sort of operational leverage that we haven't seen previously, sort of for early stage companies where they're able to make a lot more progress without needing to hire a lot more people.”at 2:53

• Smac

  Portfolio companies are running with fewer staff as AI-native team members replace slow adopters.

  Smac says his firm has seen portfolio companies operate with fewer employees because team members are becoming much more AI-native. Those who leverage the tools are kept, and those slow to adopt are not net additive. He cites this as a real effect inside his portfolio.

  “I think it's definitely happened. Like, we've even seen portfolio companies who are like, we can actually operate with less employees... Team members are becoming much more AI native and and using these tools.”at 31:26

Bitcoin & crypto markets

• Arthur Cheong · DeFiance Capital

  Most crypto liquid funds are beating Bitcoin and ETH this year, while the majors are down year to date.

  Arthur says most crypto liquid funds he knows outperform Bitcoin and ETH this year. BTC and ETH are still down year to date, while most of those funds are up roughly 20% to 100%. He argues this shows the market is rewarding fundamentals and skilled managers, and that dispersion is healthy for active managers.

  “probably more than 80% of the liquid funds I know... In crypto, they outperform Bitcoin, ETH... year to date, they are still down and I think year to date, most of the crypto funds I know is up, like, from 20 to a 100%.”at 20:27

• Arthur Cheong · DeFiance Capital

  This year, long and short legs in crypto and equities both profit as failing projects keep falling.

  Arthur argues that this year, in both equities and crypto, traders can make money on both the long and short legs. Projects that are failing and near shutdown keep dropping roughly 50% every six months, so shorts can be re-established. Fundamental drivers have been stronger than positioning, which would normally make a long-short book lose money in a bull market.

  “think this year in both equity and crypto, think you can actually make money both on a long leg and a short leg because there are some projects that are really not making it... you're shorting those, you actually gonna keep getting like a 50% down every six months”at 24:00

AI: overall outlook

• Smac

  Crypto's edge in AI is coordination and permissionlessness, not compute efficiency; value capture remains unproven.

  Smac argues that crypto AI is unlikely to win on efficiency today. He says crypto's real strength is coordination and economic incentives, plus permissionless, distributed systems. Pushback against centralized frontier labs and rising privacy concerns create room for that. He notes whether crypto projects can capture value from distributed or smaller-model inference is still an open question.

  “if crypto AI is going to compete on efficiency today, that's probably not the place that makes the most sense... I think there's a lot of kind of underlying societal trends that make the things behind crypto more appealing. And that's kind of around coordination.”at 4:23