The Most AI-Pilled CEO We Know

The Most AI-Pilled CEO We Know The Most AI-Pilled CEO We Know Brex co-founder and CEO Pedro Franceschi believes most people still underestimate how much AI will change the way companies are built. AI isn't just another tool, it's a new foundation for building products, teams, and companies. In this episode of Lightcone, Pedro shares why he thinks we're only months into a platform shift as significant as the invention of electricity, how AI has changed the way he works, and why every founder should be "token maxing" to understand the limits of the technology firsthand. He explains why the CEO needs to be the chief AI officer, how Brex is rebuilding itself around AI, and why founders should rethink what's possible when intelligence is available on demand. Apply to Y Combinator: https://www.ycombinator.com/apply Work at a startup: https://www.ycombinator.com/jobs Chapters: 01:13 – How Pedro Became AI-Pilled 04:08 – The Electricity Analogy 05:21 – Free the Claw 06:56 – Making AI Safe for Enterprise 10:57 – Why Most Companies Are Behind 13:09 – AI Teammates, Not Chatbots 14:22 – The Case for Tokenmaxxing 18:24 – The Company of One 20:54 – The One Thing AI Can't Replace 28:06 – Building Customer World Models 32:58 – Rebuilding Brex Around AI 39:02 – The CEO Must Be the Chief AI Officer 43:50 – Building Company AGI 51:43 – Why We're Still So Early to YC for lunch, it sent our entire team >> is uh you they've been treating the LLM what the LLM sees because it's about you my encounter with LMS, which was so was someone someone gave me an API access to GPT3 and uh and I was playing experience of like buying a movie ticket card. It was provisioned through an API know they they they go to the ALM know what tools the model calls and the you can also just make an HTTP request premise is you analyze you you HTTP of web documents HTTP traffic is use an LLM as a judge and the LM recruiting agent at BRS called Jim. Uh automatically 2% use an LM so we sort of do Uh I don't see any reason why a YC the LM as a judge was for us the security team at BR very rigorous and very good at what they do for a long building HT proxies. We are in the Hopefully someone's going to build a YC chatbot with a few MCPS um or a G Suite I think uh where our thesis was if you and the thesis was how to actually build people a few MCPS and let them go partners at YC. Sounds like a great models we you can use in an API, right? ACP helps on that. So, yeah. >> Yeah. ACP is good. position to to be able to spend on LLM and then that is like totally will compound. YC Startup School is well known like your like MVP like had no web UI, right? It was just like all have because no one needs HTML. It's approach just have a really simple MVP API. Braxton early days no UI just like like I don't know where you buy napkins that you can put into an LM and that LM selling to finance teams and finance remember that LM are not magic like LM and I think the biggest pitfall of LMS asked an LM a question it gave you like jobs or whatever. It's like as long as there are limits on uh RAM LMS work for people that look very be you know long inference that basically the thesis is that there's the argument to be long inference. Uh and yes there's a lot of margin in revenue growth. I think what's really know we're redesigning our KYC process have to do all these checks to KYC the customer and KYC historically is customer? But when you have KYC for free, you can you can KYC a lead versus CF very large public company uh uh this bad and so if shortly after the of the company along with you on this example of KYC that we were saying like the KYC team would never think of using the KYC technology to score a lead. The store specifically like do you buy into like build its own little company AGI bunch of learnings on this uh and and actually goes in and looks at KYC developer UI right now is like voice like build more client or more UI or >> I saw you brain LSD. >> Yeah, LSD. Yeah. lateral synactic drift. kind of like in this cone LSD mode a soul MD for like my open core agent. with AI as your SLN buddy, what are the limitations of an LLM uh degree you're working for the LM to some put the LM as almost the founder and the

Content is a YC Startup School talk about building durable startups in the AI era. The actionable market-relevant bits are mostly high-level: (1) intelligence/AI inference is getting much cheaper, (2) moats shift away from “model choice” toward distribution, product loops, data/workflows, and founder execution, and (3) US export restrictions on frontier AI matter. No explicit company mentions or investable calls, so tickers are inferred by theme (AI compute stack, hyperscalers, and export-control-exposed semis).

Talk-level, largely qualitative discussion about AI startups vs Big Tech, with mentions of LLM limits, “world models,” robotics, and continued need for large-scale GPU compute. Actionability is low because there are no concrete catalysts, numbers, or near-term company-specific claims; the most tradable takeaway is a continued AI compute/infra demand narrative (GPU/accelerators, foundry, advanced packaging).

YC talk argues “Physical AI” (AI applied to the physical economy via multimodal sensing/robotics/automation) is the next platform shift; content is conceptual with limited concrete catalysts, but maps to tradable beneficiaries in GPUs/edge compute, industrial automation, and sensor/vision stacks.

Interview-style content about Opencode (open-source Claude Code alternative) claiming rapid adoption (13M MAUs, 20x growth) and heavy token usage, framed around (1) open-source models becoming “good enough,” (2) enterprise adoption of coding agents, (3) model-choice flexibility and token economics, and (4) platform risk illustrated by Anthropic allegedly attempting to block Opencode, which backfired via attention/distribution.

Interview-style content about Photoroom (private) describing how Y Combinator increased founders’ ambition and execution mindset; little concrete product/financial data and no public-company catalysts. Limited direct trading actionability beyond a broad “AI image editing / creator tools / e-commerce enablement” narrative.

YC Startup School talk with Dust co-founder argues no single AI lab will dominate; model-agnostic application/platform layer may be a moat. Notes funding being absorbed by frontier labs, raises small by design, and highlights margin compression at the token/model level, making unit economics challenging for AI apps that resell model tokens.

YC Startup School talk: Supabase grew rapidly by offering an open-source, Postgres-based alternative to Firebase/RDS with very fast time-to-value; claims a $500M round and $10B valuation; positions “open source wins the LLM/agent era” and suggests AI agents are becoming core users. Supabase is private, but narrative has read-through to public cloud, database, and devtool vendors.

Podcast-style discussion with PostHog CEO James Hawkins on startup strategy (ambition as GTM, product expansion, founder mindset) and some broad AI/dev tooling themes (LLMs, “recursive AI loop,” intent data, AI-assisted pull requests). No concrete company-specific news, financials, or tradable catalysts.