Dropping a podcast with @benfielding, Co-Founder and CEO at @gensynai

Ben started his PhD in deep learning in 2015, well before the current wave of attention around AI. In this episode I wanted to walk through how he thinks about the progress we have made, the economics of building bigger models, and what comes next.

We get into why he questions whether the money going into centralized AI will be paid back through product usage, how open models affect that business, and why having customers who keep using your product matters more than staying ahead on a benchmark.

We also walk through how Gensyn has changed. The team moved from focusing on decentralized compute to collecting truthful information from the world. Ben explains why they made that decision and how Delphi uses markets to bring information together and check it against what actually happens.

That leads into forecasting. Ben thinks predicting future events could become a much bigger use of AI. We discuss how models could learn from being right or wrong, why people need a reason to share what they know, and what it would mean to have a probability calculator helping with everyday decisions.

We discuss

- Ben's background in deep learning and how Gensyn started

- The economics of bigger AI models and whether the spending makes sense

- Open models, competition, and building products people keep using

- Why Gensyn moved away from focusing on compute

- Verifying machine learning work and making training auditable with open-1b

- Delphi, information markets, and learning from real outcomes

- Rewarding people for sharing the evidence behind their predictions

- Forecasting, better decisions, and where decentralization helps

Timestamps

0:00 – Can AI Predict the Future?

1:14 – Ben Fielding and the Origins of Gensyn

7:26 – Scaling AI Beyond Bigger Clusters

13:29 – Why Startups Can Take a Different Approach

18:49 – Trust and Verifying Machine Learning

23:59 – The Economics of Centralized AI

27:29 – Why Gensyn Moved Away from Compute

31:09 – open-1b and Auditable AI Training

35:47 – Open Models and Closed Labs

43:29 – AI Products, Customers, and Neoclouds

50:13 – From LLMs to Forecasting

53:44 – Training Models to Make Better Predictions

57:10 – Delphi and Information Markets

1:02:26 – Rewarding the Evidence Behind a Prediction

1:04:34 – Where Crypto and AI Actually Overlap

1:11:35 – The Next Wave of AI

Enjoy!

Vikram Sekar, writer of Vik's Newsletter and co-host of Semi Doped, joins me to discuss AI agents, memory, and the hardware behind the data center buildout.

We break down why training and inference need different hardware, why copper becomes harder to use as GPUs move farther apart, and why the industry is turning to optical interconnects.

Vik explains the tradeoffs across SRAM, DRAM, high bandwidth memory, and storage. We discuss why useful work per token matters more than speed alone, how much context an agent needs, and why he sees memory as AI's biggest bottleneck.

Enjoy.

00:00 Highlights

01:18 Vikram Sekar and Semi Doped

02:55 AI agents and hardware demand

18:57 AI spending and the risk of overbuilding

25:10 Training versus inference

36:25 Why distance limits copper bandwidth

41:18 Co-packaged optics and the push beyond copper

46:04 New approaches to optical interconnects

52:24 The AI memory hierarchy

56:39 Useful work per token

57:51 How much context does an agent need

1:03:35 DRAM, HBM, and memory on logic

1:05:54 The next generation of inference chips

1:08:25 AI's biggest engineering bottleneck

Follow Vik

https://x.com/vikramskr

Vik's Newsletter

https://www.viksnewsletter.com/

Semi Doped

https://www.semidoped.fm/

Vik's institutional research

https://www.semiexponent.com/

Armani Ferrante, Founder and CEO of Backpack, joins me to talk about tokenized stocks and the next phase of global capital markets.

We discuss the different tokenization models, why redeemability and liquidity matter, and Backpack's ambition to connect traditional markets with Solana. We also get into the licensed path, Hyperliquid and Kraken, and competing with the largest exchanges.

Armani explains the tokenomics of Backpack's own BP token, how growth milestones shape its allocations, and the equity-conversion option he describes for token holders who contribute to the product. We also discuss the corporate treasury and a potential future IPO.

Enjoy.

00:00 Highlights
01:26 Armani Ferrante and Backpack
03:18 Why tokenized equities matter
13:05 The different models for tokenized stocks
19:33 Backpack's approach to tokenized stocks
28:58 Backpack's place in the market
31:47 Why Backpack chose the licensed path
35:25 One global liquidity layer
40:00 Hyperliquid, Kraken and derivatives
49:29 Competing with the incumbents
53:47 Why a token launch is a starting point
1:00:04 Backpack's three allocation buckets
1:02:31 The treasury lockup and a potential IPO
1:07:07 The option to convert BP tokens into equity

Follow Armani
https://x.com/armaniferrante

Backpack
https://backpack.exchange/

Dropping a podcast with @mdudas, Managing Partner at @6thManVentures

A crypto business can make money without that money reaching its token holders. In this episode I wanted to talk with Mike about what gives a token value and who benefits when the business grows.

We get into how he built 6MV, why he invests in applications, and how he thinks about returning money to investors when a business might not stay ahead for long.

A big part of the conversation is about tokens and equity. Mike thinks teams need to explain what people are buying, where the revenue goes, and how token holders benefit. We talk about buybacks and how much control holders really have.

We also get into why trading apps need better assets and fewer ways for users to lose money, his views on Solana, Hyperliquid, and Ethereum, and where crypto fits into AI.

We discuss

- How Mike built 6MV

- What makes a crypto business last

- When to sell and return money to investors

- Stablecoins, lending, and prediction markets

- Social trading, slippage, and keeping users

- Solana, Hyperliquid, and what gives ETH value

- Tokens, equity, buybacks, and what holders actually own

- Why Mike calls XRP and Cardano “memes for suits”

- AI agents, payments, and compute markets

- Whether AI companies are worth their valuations

Timestamps

0:00 – What Gives Tokens Value?

0:58 – Mike Dudas and the Origins of 6MV

6:20 – Liquidity and Returning Capital to Investors

10:16 – Real Adoption Beyond Crypto Token Prices

16:56 – Crypto Apathy and the Rise of AI

21:43 – Why 6MV Changed Its Consumer Investing Thesis

26:34 – Social Trading, Slippage, and Better Assets

36:25 – Solana, Hyperliquid, and the Chain Landscape

39:37 – The Problem with ETH’s Value Proposition

43:16 – Tokens, Equity, and Where the Money Goes

46:21 – Governance Tokens, XRP, and Cardano

49:40 – Where Crypto Meets AI

55:38 – Are AI Valuations in a Bubble?

58:42 – The Next Five Years of Crypto

Enjoy!

Dropping a podcast with @pequityresearch

Mr. P has been doing some great work breaking down the AI buildout and following where hyperscaler capex actually goes. In this episode I wanted to walk through the full stack: logic, memory, power, and networking.

We get into why memory could become the largest line item in the AI bill, what old GPU rental prices tell us about compute demand, and where value is going to accrue as the physical constraints get harder to solve.

At the center of the conversation is his view that AI spending cannot grow forever. Long-term contracts might change the shape of the next memory downturn, but they don’t eliminate the cycle. And signing a 10-year contract does not mean anyone can actually see 10 years of demand.

We also get into why he’s excited about optics, where NAND and HBF fit as agents use more memory, and his views on Chinese open source and the future of US model development.

We discuss:

• Why he thinks 10-year demand visibility is bullshit

• Memory’s growing share of hyperscaler capex, and why estimates vary so much

• Why older GPUs are still renting and what that says about compute demand

• How long-term agreements, pricing floors, and prepayments actually work

• Power as a bottleneck, and why identifying a constraint isn’t the same as finding an investment

• Copper vs optics, and where networking value accrues

• Agents, NAND, and where HBF fits in the memory hierarchy

• CXMT, Chinese open source, and the risks of slowing frontier model development

Timestamps:

0:00 – Why AI Spending Can’t Grow Forever
1:13 – P Equity Research’s Background
5:48 – Where Hyperscaler Capex Actually Goes
8:00 – Is Compute Still Tight?
12:08 – Memory’s Share of the AI Bill
17:20 – Why the Memory Cycle Isn’t Dead
18:37 – Inside a Long-Term Agreement
28:30 – What Happens If Customers Cancel?
32:17 – The Rising Cost of the AI Buildout
33:43 – Copper vs Optics
38:49 – Power and Gas Turbines
39:29 – US Models and Chinese Open Source
42:36 – Agents and NAND
47:00 – Where HBF Fits
52:23 – What P Is Most Excited About
56:27 – CXMT and China’s Memory Industry
1:04:00 – Closing Thoughts

Enjoy!

Mr. Bubble is world class at explaining very technical concepts in layman terms. In this episode I wanted to chat through the 0 to 1 of the chips, racks, packaging, the memory hierarchy, scaling up and scaling out.We spend a lot of time mapping computer-architecture primitives directly onto the AI factory and examining where value is actually going to accrue as packaging, HBM, and the next memory tier mature.At the center of the conversation is the belief that the bottleneck is no longer designing a faster chip, it is the physical stack from silicon to rack. While the frontier labs settle into a more mature “closed demand” phase focused on their own silicon, their own scale-up domains, and capturing the value themselves.We also get into his view on AI alignment and the future of Labs.We discuss:

• The 0 to 1 of chips: what actually has to be true before a die is useful
• Why packaging is the new Moore’s law
• Racks as the real unit of compute
• The memory hierarchy after HBM
• Scaling up vs scaling out, and why they get confused
• What the labs actually want from the supply chain
• AI alignment, and whether the labs are aligning the world or just themselves
• The future of Labs: closed frontier, custom silicon, and who gets the model

Timestamps:0:00 – Introduction & Why Hardware Is the Real AI Conversation8:00 – 0 to 1 of the Chip18:00 – Packaging as the New Moore’s Law28:00 – Racks, Trays, and the Unit of Compute38:00 – Memory Hierarchy: HBM, DRAM, and the Next Tier50:00 – Scaling Up vs Scaling Out1:02:00 – AI Alignment1:10:00 – The Future of Labs1:17:00 – Closing Thoughts and Where Value Accrues Next

My conversation with Rob Hadick.As General Partner at Dragonfly, Rob has one of the clearest views on how blockchain is evolving from speculative crypto into the actual infrastructure of global capital markets. In this episode we dig into why finance, payments, asset issuance, and markets are the only parts of crypto that are truly scaling and how the industry is quietly becoming TradFi’s onchain upgrade.We spend a lot of time mapping traditional capital markets primitives directly onto blockchain rails and examining where value is actually going to accrue as tokenization, stablecoins, and onchain trading mature.At the center of the conversation is the belief that blockchain is no longer building a parallel financial system — it is becoming the settlement, issuance, and trading layer for the existing one, while crypto itself settles into a more mature “capital markets +” phase focused on real assets, institutional flows, and sustainable business models.We discuss:- The current state of crypto as capital markets infrastructure and the decline of pure speculative narratives- Why finance, payments, and tokenization are winning while most other crypto applications struggle- The architectural parallel between traditional capital markets and on-chain systems- Tokenized assets = Securities- Stablecoins = Cash / settlement- DEXs & on-chain venues = Exchanges- Prediction markets = Information markets- Why institutions are moving on-chain and what they actually want (control, privacy, segregated markets)- Token vs equity: where value accrues in a non-Clarity Act world- The mass extinction event in crypto VC and why Dragonfly is doubling down on financial infrastructure- Stablecoins, RWAs, and the real path to “tokenization of everything”- Prediction markets (and why Polymarket matters) as the next interface layer- Sustainable business models and where value will ultimately captureTimestamps:0:00 – Introduction & State of Crypto as Capital Markets2:00 – Why Speculative Narratives Are Fading7:00 – Finance, Payments & Tokenization as the Only Scaling Verticals12:00 – Institutional Adoption & What Wall Street Actually Wants18:00 – Token vs Equity Value Accrual25:00 – Blockchain as the New Settlement & Issuance Layer35:00 – Prediction Markets, Information & the Next Interface45:00 – Crypto VC Consolidation & Dragonfly’s Thesis55:00 – Real-World Assets, Stablecoins & On-Chain Markets1:05:00 – Closing Thoughts: Where Value Accrues NextEnjoy!

My conversation with Tarun Chitra.

As a co-founder of Gauntlet and GP at Robot Ventures, Tarun has one of the sharpest frameworks for understanding market structure across both crypto and AI. In this episode we dig into why open-source AI is unbundling faster than most people expect, and how the resulting stack looks surprisingly similar to DeFi.

We spend a lot of time mapping the AI infrastructure layers directly onto crypto primitives and examining where value is actually going to accrue as models, harnesses, routers, and inference providers separate.

At the center of the conversation is the belief that AI’s unbundling is creating a new competitive order-flow market (data centers competing like nodes, MEV-like dynamics for tokens/GPUs) while crypto itself has settled into a more mature “TradFi plus+” phase focused on trading, payments, and bringing real assets on-chain.

We discuss:

- The current state of crypto as TradFi+ and the decline of speculative narratives

- Why AI is killing Bitcoin mining economics and weakening ETH value accrual

- The architectural parallel between AI stacks and DeFi

Harnesses = Wallets

Routers = DEX aggregators,

Models = Protocols

Inference Providers = LPs

- Why open-source models are unbundling faster than traditional software

- Agents as the next interface layer and the potential unbundling of ETFs

- Cryptography, verifiable compute, and turning GPUs into digital assets

- Onchain compute trading as the real crypto × AI opportunity

- Sustainable business models and where value will ultimately capture

Timestamps:

0:00 – Introduction & State of the Crypto Market (TradFi+)

2:00 – Speculative Narratives Fade, Trading & Payments Remain

7:00 – AI’s Impact on Bitcoin Economics & Data Center Opportunity Cost

9:00 – ETH Value Accrual, Solana Positioning & DeFi Token Sustainability

15:00 – Trading Design Space & On-Chain Volume Upside

25:00 – AI Unbundling Thesis: Open-Source Models vs Data Centers

35:00 – The DeFi Mapping (Harnesses, Routers, Models, Inference)

48:00 – Agents, Preference Expression & Unbundling Traditional Products

1:00:00 – Real-Time Harness Generation & Active Learning

1:05:00 – Cryptography, Verifiable Compute & On-Chain GPU Markets

1:10:00 – Closing Thoughts: Where Value Accrues Next

Enjoy!