Feb 14, 2025 · 1h 1m · sourcery
How Writing Led To Investing In Ramp, Scale AI, Replit, Varda | Packy McCormick, Not Boring · Sourcery with Molly O'Shea
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of Sourcery, host Molly O'Shea speaks with Not Boring founder Packy McCormick about his transition from newsletter writer to venture capitalist, his investment thesis around hard tech 'vertical integrators,' and his pragmatic view on artificial intelligence.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Molly holds 17.1% of the talking time here. How this is scored →
Molly as informed peer 3.9 Guest teaching 2.8 Guest disagreement 1.6 Molly pushing back 1.1
speaking balance: gold is Molly, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 54:00 Rejection of AI Worship and Anti-Human Doomerism
Packy aggressively criticizes extreme AI cultists and doomers, labeling their fatalistic surrender to machine intelligence as bizarre, anti-human behavior.
Hardest push from Molly ▶ 56:15 Molly's Pushback on AI Switching CostsO'Shea directly rejects the premise of enduring model moats, pointing to DeepSeek's rapid adoption to argue foundation models are interchangeable commodity software.
Biggest teaching moment ▶ 29:00 Packy's Grid Economics ReframePacky systematically deconstructs grid mechanics, explaining why power generation costs decrease while transmission interconnect queues and transformer shortages drive consumer power bills higher.
Molly holds their own ▶ 17:00 Molly Drawing on Upfront Diligence PrinciplesO'Shea leverages her venture capital diligence experience at Upfront Ventures, citing Mark Suster to critique over-intellectualized venture theater and market maps.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Molly as informed peer | Guest teaching | Guest disagreement | Molly pushing back | Why |
|---|---|---|---|---|---|---|
| Episode Welcome and AI Voice Cloning Discussion | 2 | 1 | 1 | 1 | Molly O'Shea introduces Packy McCormick with friendly banter regarding his ElevenLabs AI clone and early subscriber milestones. Packy playfully corrects her on the podcast name ('Boring News' vs 'Not Boring News') before discussing his newsletter's monetization timeline. The tone is highly collaborative and conversational. | |
| Investment Thesis and The Barbell Strategy | 3 | 3 | 1 | 0 | Packy explains his barbell fund strategy, splitting capital between vertical integrators combining hardware and software and high-risk frontier bets like crypto. O'Shea prompts him on fund mechanics and how writing fuels his investment pipeline, offering supportive open-ended questions. | |
| Profiling Startups: From Early Fits to Ramp Super Bowl Ads | 4 | 2 | 1 | 0 | O'Shea mentions her familiarity with Packy's deep dives and diligence style from her own VC background. Packy describes writing extensive profiles on Meter and Ramp, explaining his preference for rich literary journalism over simplistic market maps. | |
| Research Methodology and The Founder Compression Algorithm | 3 | 3 | 0 | 0 | O'Shea asks about Packy's drafting process and whether he ever pulls investments mid-research. Packy explains his research workflow and characterizes exceptional founders as a compression algorithm for complex industry diligence. | |
| Sponsor Break: Passes Monetization Platform | 2 | 2 | 0 | 0 | Following a quick Passes platform promo, Packy expands on founder traits, focusing on a founder's ability to inspire great work from top talent and navigate the historical idea maze. | |
| Cutting Through Venture Theater and Cultivating Intuition | 6 | 2 | 1 | 2 | O'Shea draws upon her venture background at Upfront Ventures, quoting Mark Suster on narrative creation versus fundamental truth in venture diligence. Packy agrees, decrying performative venture theater while admitting his writing serves as intellectual preparation for rapid intuitive decisions. | |
| Early Investment Mistakes and Meeting Somos Internet | 3 | 3 | 1 | 0 | Packy explains early investing errors involving compounding miracles before detailing his fast investment into Colombian telco Somos Internet after an in-person meeting. He emphasizes how previous research in nuclear and networking refined his pattern recognition. | |
| Standout Portfolio Investments and Earth AI | 4 | 3 | 2 | 1 | O'Shea and Packy discuss portfolio winners and the dynamic where small check sizes disproportionately correlate with write-offs. Packy pushes against consensus VCs who misjudged Earth AI's capital requirements, explaining how vertical integration unlocks high cash generation. | |
| Sourcery VC Newsletter and Channel Outro Promo | 5 | 3 | 1 | 2 | O'Shea transitions from an ad read to pressing Packy on how vertical integrators achieve defensible moats. When Packy distinguishes his thesis from pure deep tech science risk, O'Shea playfully challenges him by citing his own deep tech portfolio investments. | |
| Case Study: Base Power Company and Decentralized Grid Storage | 3 | 4 | 0 | 0 | Packy delivers an educational deep dive on Base Power Company, explaining grid transmission bottlenecks and how residential battery packs create a decentralized power utility. O'Shea prompts him to unpack the founding team and economic model. | |
| Go-To-Market Execution and Building True Moats in Hard Tech | 5 | 2 | 1 | 1 | O'Shea contrasts traditional low-capex SaaS venture investing with capex-heavy hard tech inflections. Packy agrees, contrasting durable moats in companies like SpaceX against the transient, easily commoditized revenue curves of generative AI startups. | |
| Hard Tech Founder Psychology and Fuse Fusion Case Study | 4 | 3 | 1 | 0 | O'Shea highlights the psychological resilience required of hard tech founders and references American dynamism trends in LA. Packy uses Fuse as a case study, detailing how generating interim pulsed-power radiation revenue de-risks fusion power development. | |
| Spicy Take: YC Request for Startups vs. Inbound Founder Passion | 4 | 4 | 4 | 2 | Packy delivers a contrarian critique of Y Combinator's Requests for Startups, arguing that manufactured prompts produce inferior founders compared to organic obsession. O'Shea probes for confirmation bias in top-down ideation before yielding to Packy's thesis on inbound obsession. | |
| Downstream Late-Stage Capital and Series A/B Mispricings | 5 | 3 | 1 | 2 | O'Shea brings up capital dislocations for downstream Series B and C funding in capex-heavy businesses. Packy outlines why downstream Series A rounds are often mispriced when tier-one investors wait for single binary technical milestones before aggressively bidding up valuations. | |
| Debunking AI Hype: Why AI is Super SaaS, Not God | 4 | 4 | 4 | 2 | O'Shea quotes Packy's provocative statement that AI will become super SaaS rather than God or Satan, prompting his views on current model capabilities. Packy strongly dismisses scaling-law hype, explaining that newer models are merely incremental optimizations rather than conscious agents. | |
| The Anti-Human Bias in AI Cult Culture | 4 | 2 | 5 | 2 | O'Shea asks whether Packy views AI purely as a marketing stunt. Packy clarifies forcefully, rejecting both doomerism and AI techno-religion as anti-human zealotry while asserting his belief in software-driven productivity gains. | |
| DeepSeek, Model Commoditization, and ChatGPT Brand Power | 6 | 3 | 3 | 4 | O'Shea challenges the uniqueness of frontier AI models, citing DeepSeek's rapid adoption and arguing foundational models are pure commodities where users switch freely based on price. Packy pushes back by highlighting OpenAI's massive consumer brand equity and Super Bowl presence. | |
| Brand Power in Deep Tech, SMRs, and SpaceX Heritage | 3 | 3 | 1 | 1 | Packy explains how brand and flight heritage become the ultimate defensibility in high-stakes deep tech like SpaceX and SMR nuclear reactors, illustrating the point with an anecdote about his children. O'Shea wraps up the episode inquiring about future plans. |
Statements from this episode (33)
McCormick: Not Boring Capital manages about $50 million in total AUM
“Oh, like, 50 total AUM now, yeah.”
McCormick: Not Boring focuses on hardware-software vertical integrators beating incumbents
“Doing kind of this barbell where it's mostly these vertically integrated companies, I've written a bunch about these, we can talk about them but kind of thesis being companies that use both kind of hardware and software, build vertically integrated systems It …”
McCormick: Ramp filmed its Saquon Barkley Super Bowl ad in four days
“We got word that we got a super bowl slot and we've been like spending the whole past day trying to coordinate with Saquon and his people to get this thing done. And they filmed it in like You know, they did the whole thing in, like, three or four days.”
McCormick: Reducing investment decisions to market maps or one-sentence theses fails
“And like anybody who like boils it down to like a market map or like a one sentence thesis on a company, I think is like just a very tough way to invest.”
McCormick: AI will prove concise answers matter less than immersive storytelling
“I think, like, AI will make us see, like, answers are, like, really not important, but you want to read Robert Caro, because, like, you want to read Thousands and thousands of pages that, like, really make you feel and understand what it was like to be in the …”
McCormick: Literary nonfiction and deep journalism are best for early-stage underwriting
“That, like, Jim Simons at Rentac versus, like, Robert Carroll, like, who would be a better earlier stage, early stage investor? And I think it's actually, like, Robert Carroll. Like, I think, like, literary nonfiction done really, really well, or journalism do…”
McCormick: Backing the right founder is a compression algorithm for research
“I do think, like, and people say this, and it's cheesy, and it's, like, also true, is that the founder really matters, and, like, there's definitely a midwit curve on that that I've written about before, which is, like, You think founders really matter and the…”
McCormick: Founder magnetism and fear of disappointing them drives peak performance
“It all comes down to people and like the team that you build and like what you get out of your people in the end. So if you have this like magnetism Or, like, there's almost, like, people have a fear of disappointing you and, like, they really want to, like, d…”
McCormick: Much of VC Diligence Is Performative Theater to Justify Deals
“Cause I do think there's like just so much venture theater. Where, like, you just have to do a bunch of stuff because, like, your boss, you know, your GP expects it, or their LPs expect it. Like, there's just a bunch of, like, shit that you have to produce to,…”
McCormick made a recent investment decision solely over a three-hour coffee
“It's one of the most recent decisions I made. It was, like, over a three-hour coffee. I didn't write a single word. You know, I met the guy and read the deck and loved him and invested.”
McCormick: Startups requiring compounding miracles to succeed never work out
“I can see like if this happens and then this happens and then this happens and then that happens and like things go just right here, like this could be a really big company and like it never works. Like you can't like, you know, there's the compounding miracle…”
McCormick: Low-conviction smaller checks account for most Not Boring failures
“A lot of my smaller checks are the ones that have failed already. It's the ones where I'm like, I don't like fully, but like, maybe like if it does, it'd be like really huge and like, it's a pretty cheap price.”
McCormick: Earth AI won't need future venture funding if its model works
“If this model goes well, like, They do not need to raise again. Cause they can sell state little, little stakes in their minds and use it to fund.”
McCormick: Vertical integrators take integration risk over science risk to outperform incumbents
“There's a lot of like deep tech that I think you're taking science risk. And like, I think with vertical integrators, you're taking integration risk where there's like five different kind of like cutting edge-ish, but proven technologies that you're trying to …”
McCormick: Anduril hit $28B valuation after doubling revenue
“Mandel just raised it a two X step up again, and they're at twenty eight billion and They're not that far off now. They're like a quarter of the way to the big defense prime's kind of market caps and continuing to grow really fast and like what doubled revenue…”
McCormick: Electricity is not getting cheaper due to rising transmission costs
“Electricity is not getting cheaper though, because transmission ends up getting more and more and more expensive over time. So like getting it from point A to point B, there's like long delays on the interconnect queue. It takes forever to build transmission. …”
McCormick: Base Power rents batteries and monetizes by trading power
“They're a retail electric provider, so they like are your power company when you install the battery. They, because of that, they don't have to sell it to you. They just rent you the battery essentially, and they, and you pay for the install cost, but then the…”
McCormick: SpaceX suffered only ~50% dilution despite raising $10B in equity
“I think they're like 50% dilution over the past 20 years, even though they've raised ten billion dollars of, you know, of equity aside from secondaries.”
McCormick: Most fast-growing AI software startups will flatten and fail to IPO
“By choosing to, like, not play in this space, I'm definitely missing a trillion dollar company or two, but I think for the most part, like, I think there's gonna be, like, a lot of these types of things happening, or even, like, these types of curves, where th…”
McCormick: Fuse monetizes pulsed power radiation before reaching fusion energy
“If you go with this one particular architecture, what you're really doing is selling radiation by building these pulse power generators, and you can use that for Radhart Electronics and nuclear arsenal testing and, like, all these things that can, like, genera…”
McCormick: Fuse aims to become the SpaceX of the NNSA
“His whole thing is they want to be the NNSA, the National Nuclear Safety Administration's I think that's the right acronym. They want to be the NNSA's SpaceX essentially, like you know, like SpaceX was to NASA, they want to be to the NNSA.”
McCormick: YC 'Request for Startups' Prompts Rarely Lead to Great Companies
“Like when YC puts out the request for startups, I'm like, I can't imagine that anything good is going to be built against these categories starting now because you've read the YC request for startups.”
McCormick: Supersonic Private and Commercial Flight Will Emerge Within 10 to 15 Years
“I think there's a path where we wake up in 1015 years, and, like, first it's supersonic private planes, and then supersonic commercial planes, and, like, that is a long path to kind of get there, but, like, the technology is there, and actually, like, just bat…”
McCormick: Anduril's early round was crowded at a $100M valuation
“But that, like, early story that, like, nobody wanted to touch it was actually, like, a pretty crowded round at a hundred million dollar valuation.”
McCormick: Series B and C capital is easy once category leadership is clear
“But I think actually the B and then the C, like, once they get to a certain point and do some of those things, and it becomes clear to others that, like, if anybody wins this category, it's going to be this company, then capital, I think, actually becomes, lik…”
McCormick: DeepSeek represents doing the same thing cheaper, not a qualitative leap
“My DeepSeek point is that, like, This, like, crack team that figured out all these algorithms, breakthroughs, and whatever, essentially just figured out how to do the same thing cheaper. Like, there hasn't been any leap that I've seen that I've been like, yeah…”
McCormick: AI could create a winner-take-all dynamic in corporate cards like Ramp
“AI is the reason that a company that might not have been winner-take-all before would be winner-take-all now. Like, I think RAMP is a really interesting example of, like, There might be a winner take all dynamic in corporate cards at some point. If like that s…”
McCormick: AI will become 'super SaaS' rather than godlike AGI
“I called it super sass in, in that like 20, 25 prediction. And I think it is like, I think it's really amazing and like in no way godlike. And it's just like, I feel like I'm being psyopped whenever people are like, oh my God, this is like, this is, this chang…”
McCormick: Commoditizing intelligence shifts economic value to higher stack layers
“When something becomes commoditized, we're like, it could be answers or intelligence or whatever, like there will be value at the next layer of the stack. And like, that's just how the world works.”
O'Shea: AI models are a commodity product with zero customer loyalty
“They're using it, like, it's, like, it's commodity business, we all know that. Like, we've known that from the beginning, so I don't know why we go through these circular phases of, like, thinking it's gonna change the world, and it's, like, a power law thing,…”
McCormick: Consumer brand equity creates real value in AI despite commoditization
“That is a brand that like, you know, he might be with, like talking to ChatGPT for the rest of his Life, you know, or like some version of it. And so I do think there's real value in brand.”
McCormick: Nuclear reactor markets will ultimately be decided by brand power
“I think like nuclear ends up coming down, will end up coming down to like brand, like who got something working early on and like, you know, nobody ever got fired for buying like reactor company X that is good enough and cheap enough and whatever else and safe…”