Paul Copplestone

1. Key Themes

Agents Are Now the Primary Customer, Not Humans

Supabase's growth is being driven by agents choosing and operating the product. Garry Tan framed it as an evolution of YC's famous t-shirt: "increasingly all of us are realizing it's Make Something Agents Want. And your company really exemplifies that. You know, something like 70% of the new activity is from agents." 00:00:58 Paul described the scale: "databases under management now approaching 50 million, and this time last year was 8 million, so it's just a huge, huge inflection." 00:02:01 Revenue will grow 4x this year, and sign-ups went from roughly 3 million to 13 million in 12 months.

The Rise of the "Builder" Class: Domain Experts Who Never Needed to Code

A new customer segment is emerging: people with deep domain knowledge and pent-up ideas who previously couldn't recruit an engineer. Paul: "this category of people have all these pent-up ideas that they could never convince a software engineer to like help them build, and now they don't have to convince anyone." 00:05:40 Harj Taggar gave a concrete example of a fintech founder with a $1B+ exit: "now, in side projects, he's just building his own stuff. And he has all the knowledge from how to build a fintech company, so he can actually build pretty sophisticated just like for his own personal finances." 00:06:29 Paul's anecdote of a ~50-year-old Brazilian taxi driver who built "an entire taxi app from his phone... a little mini Uber for his local market" 00:08:49 shows the long tail reaching blue-collar workers.

Knowing What to Build Is Now the Scarce Skill

Garry described YC's shifting thesis: from "you need a technical co-founder no matter what" to a new view. "The much harder thing to do is to know what to build. You know, being able to build is something that anyone can sort of have on tap now." 00:07:39 Later he framed the mismatch: "the capabilities of the models are like going parabolic, and then the number of people who know how to actually have an impact on society with it is like more or less constant." 00:41:27

Winning Agents Requires Re-Engineering Developer Experience for Machines

Paul laid out a two-stage model: "there's like two stages to winning the agents. The first is being picked by the model, and then the second is like, if they get stuck, you don't want them to switch away." 00:31:02 Tactics include code-first/CLI-first design, writing error messages that tell agents how to self-remedy, building eval suites to test which models build best on Supabase, and publishing skills. Paul: "pretty much now a lot of our engineering effort is going into like testing specific use cases rather than the building." 00:31:39

Open Source as an Agent-Era Moat

Jared Friedman asked how much being open source mattered; Paul: "It's hard to know the counterfactual, but I think it was critical." 00:03:29 Garry's reasoning: "the agents actually really like open source because when something goes wrong, it can just look at the source." 00:03:57 Paul added the enterprise angle: "with security being so important... a lot of the, say, enterprises are going to want to inspect the source code to know that it's secure. And if you've got closed-source code, then you probably won't want to adopt them as a vendor." 00:04:15

The Personal Database Becomes the Store of Record for Life

Harj described his "control tower" app: "I would never have thought about wanting a personal database I think a year ago, cuz it wouldn't have made any sense. But now... that's actually like the most valuable thing." 00:13:28 He noted that vibe-coded personal finance can beat entire categories: "that was like a whole company, like several startups just doing personal finance, and now you can just like build it for yourself." 00:10:04 Paul noted agent lock-in is a risk, so data portability matters: "I'm slowly extracting it out... so that if I want to attach different agents on top, then I know that I can use any agent." 00:12:10 Harj's conclusion: "the database is actually the thing that's valuable. You want to own your own database and then build all the applications or agents on top of it." 00:12:36

Serving Both Ends of the Scale Spectrum: "Prototype to Petabytes"

Supabase is pursuing two opposite directions at once. Paul: "we've got two products that cater on the two ends, so Multigres that scale up to petabytes and then Turso scale down to like these prototype and SQLite-type workloads." 00:17:32 The acquisition of Turso enables a "Supabase Spark" tier for the flood of getting-started workloads. His scale math: "ChatGPT... 1.2 billion weekly active users, and let's say all of them want to launch something every month, that's 1.2 billion databases a month." 00:15:21

The Stripe/AWS Model: Win Day One, Scale Forever

Paul on why capturing the earliest stage matters: "something like 40% of AWS's revenue comes from like startups that started on them and then scaled up... to convince people to migrate from one database provider to another is near impossible. And so we just want to be there on day one, make sure that they have a great time, and then make sure that they actually don't want to leave." 00:24:53 Harj: "It's the same business model as Stripe." 00:25:24 Paul notes the complication: "What does day one look like? That's rapidly shape-changing." 00:25:29

Rapid Internal AI Adoption Through No Policy

Garry cited a Supabase internal poll in which engineers writing 80% of their code with agents went from 10% to 60% in six months. Paul: "we don't have an AI policy internally. We just let people do whatever they want, and we give them access to whatever tools they want to use." 00:18:20 The side effect is a testing explosion; Garry noted YC startups like "Blacksmith, Depot, Ubicloud" are riding the CI/test-volume surge. 00:19:14

Small Teams, Huge Revenue: The Vertical SaaS Reset

Garry observed that vertical SaaS companies at "$5 to $10 mil ARR even, which like historically is a lot of ARR" are struggling to raise Series A, but "you can get to 10 to 15 mil, you know, often 25, 30 mil, 50 mil with like 10 or 15 people." 00:43:32 Paul's math: "If you got 15 people, 2 million ARR per head, and with 50% margins, you're all making a million a year." 00:44:20 Garry's advice to bummed-out founders: "you're probably going to look back on this as a blessing, cuz you don't need the money." 00:44:31

2. Contrarian Perspectives

Everyone, Not Just Developers, Will Own a Database

Conventional wisdom says databases are infrastructure for engineers. Jared noted the historical absurdity: "the idea of a database company being a direct-to-consumer business would have just like sounded like a syntax error." 00:26:13 Paul's counter: "I certainly think that it'll be the case that everyone in the world with access to AI will become a builder of some sort." 00:00:00 This extends to a consumer-scale vision: a billion-plus databases per month.

The Dashboard Is Dead; "No One Visits Dashboards Anymore"

Supabase began as a dashboard company, an Airtable-like UI on Postgres. Paul reversed completely: "this year, I just saw that actually, no one's coming to the dashboard... I just kind of put out a product memo to the team that said, 'It has to be code-first, Git-native.'" 00:20:35 Even humans building on it won't be interacting directly soon: "those humans pretty soon won't actually be interacting." 00:27:29

Product-Led Growth Has No Ceiling (Contra Early Investor Advice)

Paul: "our early investors were saying, 'Ah, PLG, product-led growth, it's going to tap out at this revenue and you'll have to start moving into this segment of the market,' and there just appears to be no ceiling so far." 00:05:40 Even more unusual: top-of-funnel acceleration normally degrades quality metrics, "but actually this year all of them are kind of going up and to the right, which is a bit insane." 00:04:43

Scarcity Has Flipped: Building Is Free, Taste and Domain Knowledge Are Not

Against the 2022-2025 "you need a technical co-founder" orthodoxy, Garry argues the scarce resource is knowing what's worth building: "Can you imagine all the things that never would have built, exactly for what you said? It's like, 'Oh, we know that people would buy this. We know that there's a problem here. But at the margins, those people couldn't convince an engineer to work on it.'" 00:07:39

Raising Capital Is Becoming a Negative Signal of Achievement

Paul on the shift in startup status: "the badge... to get funding at the start, felt like, you know, I was achieving something great. But actually, you know, to actually not fundraise now would, in my mind, be like, 'Oh, that's achieving something great.'" 00:44:56

3. Companies Identified

Supabase

Open-source Postgres backend; 13M users, 1M+ databases created per week, ~50M databases under management. Mentioned as the default database inside Claude Code and Codex and as the subject of the episode. Paul: "this year, like revenue will grow 4x." 00:01:38 Tagline: "Prototype to petabytes." 00:16:03 Garry: "Supabase of all the companies has really been able to ride high through every single wave." 00:45:20

Turso

SQLite-compatible (and Postgres-compatible architecture) database company acquired by Supabase. Paul: "We made an acquisition recently, Turso. They've got an architecture that's compatible with Postgres, and actually, it's compatible with SQLite as well. I think SQLite's a phenomenal database." 00:16:03 Powers the new low-cost "Supabase Spark" tier.

Multigres

Supabase's horizontal-scaling layer for Postgres. Paul: "Multigres that scale up to petabytes." 00:17:32

OpenClaw

Agent framework that Harj and Garry use as a personal agent. Harj: "getting really addicted to OpenClaw." 00:10:04 Garry: "when I saw OpenClaw... it was a very different way for you to work with AI, cuz then it's like, oh, the AI's smart, it's not trying to hurt you, and then you should just let it like do stuff." 00:32:25

G-Brain (Garry's open-source retrieval project)

Open-source RAG plus graph retrieval layer on Postgres/pgvector, built on Supabase. Garry: "it's like, RAG retrieval plus like a graph database that creates like structure out of unstructured data. And so it's like dynamic schemas." 00:32:25 Includes a "G-Brain Evals" repo with 40 categories of retrieval tests.

Parse and Firebase

Mobile backend-as-a-service predecessors, acquired by Meta and Google. Paul: "they like blazed the trail for us to come along and actually have a legitimate database business. Because in 2014, it would have been ridiculous not to use RDS." 00:22:07

MongoDB

Credited with showing database companies can exist outside the hyperscalers. Paul: "Mongo came along, Mongo started moving to Atlas, and a lot of these database companies kind of showed that you don't need to just use one of the three hyperscalers." 00:22:07

Heroku

Developer-experience-first platform that popularized Postgres. Paul: "kudos to Heroku. I think they popularized Postgres." 00:24:17 Harj noted it "sort of... couldn't quite bridge that gap" to larger scale. 00:24:14

AWS

Analog for the "win startups on day one, scale with them" model. Paul: "something like 40% of AWS's revenue comes from like startups that started on them and then scaled up." 00:24:53

Stripe

Model for generational business built by serving the fastest-growing companies early, and for founder-led future orientation. Harj: "It's the same business model as Stripe." 00:25:24 Paul: "companies like Stripe... have this real like someone like a founder who just cares about like almost focusing on what the new ways of building are." 00:45:34

Bolt and Lovable

Vibe-coding platforms that built on Supabase's API in 2024. Paul: "it turns out that was exactly the right move for 2024 when the likes of Bolt and Lovable and all of these people started building on top of us." 00:20:35

Blacksmith, Depot, Ubicloud

YC startups benefiting from CI/test-volume explosion. Garry: "they're all vertical, cuz everyone's doing CI, and the number of tests that you have are just like through the roof." 00:19:04

Capy

Agent orchestration tool Garry uses, with "captains" for always-on autoresearch loops. Garry: "Capy has this concept of captains... I have one that is always going that's like come up with new categories." 00:37:04

Exa

Search API Garry used for bulk research download. Garry: "Use Exa and download and diarize every paper you can find in AI research." 00:35:48

Aside Browser

Browser-use tool Garry uses. Garry: "I use Aside browser now for browser use." 00:34:45

Garry's List

Garry's agentic newsroom project that spawned G-Brain. Garry: "Garry's List, which is like agentic... an agentic newsroom, and I had to build up like basically a research system for RAG and for like policy and papers." 00:32:25

Palantir

Cited as a reference for dynamic-schema structure from unstructured data. Garry: "dynamic schemas, like a little bit like what I remember Palantir was back in the day." 00:32:25

Airtable

Paul's early product inspiration. Paul: "let's build a UI that looks like Airtable on top of Postgres, because I actually loved Airtable as well." 00:20:35

PostgREST

Postgres API tool Supabase helps maintain. Paul: "PostgREST is a very famous tool for the Postgres ecosystem. And we helped to maintain that, so we kind of got sprawl across many communities." 00:03:24

Mint and Monarch

Personal finance apps Harj says his custom app matches or beats. Harj: "better, I think, is at least feature parity to Mint or Monarch." 00:10:04

Claude Code and Codex

Coding agents where Supabase is the default database. Garry: "default database inside tools like Claude Code and Codex." 00:00:36

Visa (and Stripe) / OpenUSD

Paul's "OpenUSD" anecdote referenced Stripe and Visa announcing a product of that name. Paul: "suddenly Stripe and Visa and everyone announced OpenUSD." 00:39:41

Grokbot

Mentioned by Garry as analogous to Capy's captains. Garry: "similar to Grokbot's bots." 00:37:04

AlphaSignal

Mentioned as a research-discovery tool. Garry: "Today I could use like AlphaSignal or something." 00:35:48

4. People Identified

Co-founder and CEO of Supabase. Serial founder who previously built startups wrapped around blue-collar workers. Mentioned for adapting through three product eras. Garry: "Supabase of all the companies has really been able to ride high through every single wave." 00:45:20 His dogfooding discipline: "my kind of heuristic to make sure I'm building or dogfooding correctly is that I just don't look at the code." 00:27:29

Garry Tan

YC President and CEO; builder of G-Brain and Garry's List. Hosted; relevant as an extreme power user. Garry: "I have half a million markdown files on like every communication I've ever had." 00:32:25

Harj Taggar

YC partner; built a personal "control tower" app on Supabase. Harj: "I call it control tower, but it's just like sort of every section of my life." 00:10:04

Jared Friedman

YC partner and host. Mentioned for the "conference just for agents" idea. Jared: "I want to like send my agent to Supabase Select to like listen to all the awesome new features and figure out how to integrate them into my code base." 00:28:49

Boris Cherny

Creator of Claude Code, referenced for his vision of coding. Jared: "his vision of the future of coding was more like, you just talk to Claude like a friend, and you don't have to know about evals and all this stuff." 00:39:25

Steve Yegge

Author of "software factories" concept. Garry: "reading about Steve Yegge's software factories and things like that, and then not being able to totally figure it out." 00:38:54

The 50-year-old Brazilian taxi driver (unnamed)

Built a local ride-hailing app from his phone on Supabase. Paul: "he had the energy of a like teenager building something." 00:08:49

The fintech founder with $1B+ exit (unnamed)

Harj's example of a domain-expert builder writing his own bank integrations. 00:06:29

Paul's Executive Assistant

Briefly mentioned as catalyst for his personal-agent obsession. Paul: "after a week of my EA being away, and me learning exactly what they can do." 00:27:29

5. Operating Insights

Rewrite Your Error Messages as Agent Instructions

Paul: "inside the error messages, you are literally telling them how to fix this error or to go and read some docs to fix the error. So a lot of it is just picking your use case and just trying to do these loops on improving your own error messages so that you can one-shot basically anything that you want." 00:30:01 He noted this wasn't obvious even to other dev-tool founders at a dinner.

Dogfood as a "Normie of Next Year," Without Touching Code

Paul's method: "I'm trying not to like point it towards GitHub or anything like that. And I'm trying to code like a normie or someone would do next year, so that I know... is it going to have a good time with Supabase in 1 year's time?" 00:41:00 His rule: never look at the code, and only check the architecture to verify correctness. He also chats with his agent by phone call.

Build Evals as the Engine of Autonomous Improvement

Garry's loop: take production traces, extract an eval dataset, run a separate eval repo with 40 categories, then "just auto research and loop, and anything that improves your stats on this eval, like ship it into master by default. And anything that you tried that didn't work, write a markdown file that explains why." 00:37:04 A second always-on agent generates new eval categories.

Skip the AI Policy, Fund the Tooling, Invest in Testing

Paul: "We don't have an AI policy internally. We just let people do whatever they want." Engineers' adoption leapt from 10% to 60% as "the amount of testing that we're doing and the tools that we're adopting for testing has been pretty useful to make sure that they actually know that the end result is going to be good." 00:18:20

Reserve Roadmap Capacity for Ecosystem Partner Bets

Paul: "we try to partner with the key players in the space... And we try to leave some capacity within the team, which is also very difficult... not a lot of the bets actually win... But the ones that do work then end up meaning that we ride this awesome wave together." 00:47:34

6. Overlooked Insights

Postgres Indexes Could Become Agent-Generated and Dynamic

In passing, Paul relayed an idea from a dinner: "putting an agent on a loop to create the perfect data structure for the search. Like, if you're searching this data set this many times and it's taking 10 seconds, then why don't you just put an agent on a loop to create a new type of index or like find the best index for returning that in less than a second?" 00:36:21 Garry agreed: "dynamic retrieval is pretty clearly going to happen the second you have really, really smart codegen." 00:36:58 This implies database performance tuning, a historically high-skill DBA function, could be automated into self-optimizing storage, which could change who captures value in the database layer.

"Conferences for Agents" and Distribution Aimed at Agents, Not Humans

Jared's offhand idea of sending an agent to Supabase Select, with Harj adding "you can sort of airdrop them like the markdown files for everyone," points to a new go-to-market surface: developer relations whose audience is the agent and whose artifact is markdown skills and docs. Combined with Paul's point that winning means "being picked by the model," this suggests vendors competing for inclusion in model weights and agent skill libraries, not just developer mindshare. Garry's "The Artifactory" name 00:29:42 hints at a category nobody has built yet.