Paul Copplestone
Related people
Episodes
Notes
Y Combinator · What Happens When Everyone Can Build Software
-
How Do You Design Software That AI Agents Want to Use?
Paul Copplestone says you design software for AI agents by putting the fix inside the error. A generic developer-tool error, the old 500 that only says something is wrong, leaves a person to go search. An agent needs the message to say how to fix the error, or which docs to read, so the agent can remedy itself and one-shot the task.
Supabase got there by picking a use case and looping on its own error messages. He describes two stages to winning the agents. The first is being picked by the model. The second is the moment the agent gets stuck: give it the tools to fix the problem so it does not switch to something else. Agents are smart enough now to figure a lot out, and the product still h
-
Why Build Software for Personal AI Agents?
Paul Copplestone is building Supabase for personal agents because the humans using it will soon not be the ones interacting with the product. He is very bullish after a week of his EA being away, the week he learned exactly what personal agents can do, and he is obsessed with building for them. The company is code first, built for humans who do code first, and the work is actually for the agents.
He builds by talking to it. He sits on a phone call with it all the time, chatting and asking it to make changes. His heuristic for building and dogfooding correctly is that he does not look at the code. He does not allow himself to check the architecture, except when he wants to check that it is a
-
What Is the Stripe Model for Databases?
Paul Copplestone says the Stripe model for a database company is to get people on day one and then scale with the fastest growing companies in the world. Doing it for a generation produces a generational business. It is basically the best business model on earth if you can get it.
He might get the figure wrong, but something like 40% of AWS revenue comes from startups that started on them and then scaled up. Startups that begin on a provider and later grow are the category he wants. Convincing people to migrate from one database provider to another is near impossible, so the aim is to be there on day one, make sure they have a great time, and make sure they do not want to leave.
Day one is
-
What Did a Brazilian Taxi Driver Build From His Phone?
Paul Copplestone says a taxi driver from Brazil built an entire taxi app from his phone, a little mini Uber for his local market.
After a conference, the driver came up to him. He does not know why the driver built the app. The man is a taxi driver himself. The driver could build it, and he was extremely grateful he could do it. The driver was like 50 years old and had the energy of a teenager while building.
He thinks there is a lot of uncertainty around what the future of work will look like. Among builders there is excitement about the future. People who are uncertain are discovering a skill set and opportunities. Previously they were just blue collar workers.
-
Why Are Non-engineers Building Software in 2026?
Paul Copplestone says builders are coming online in 2026, including people without a software engineering background, because the models are getting so good at building. Signups are accelerating, and activation and conversion are rising at the same time.
Faster signups usually bring a decline in activation and conversion. In 2026 signups, activation, and conversion are all going up and to the right. Software engineers still make up a huge portion of users, but the builder segment is growing fast.
Early investors said growth would tap out at a revenue level and he would have to move into another segment of the market. There appears to be no ceiling so far, and he cannot see where it stops.
-
How Fast Did Supabase Grow in a Year?
Paul Copplestone puts Supabase at 13 million signups, up from around 3 million 12 months earlier, with databases under management approaching 50 million versus 8 million a year ago. More than a million databases are created every week. He calls the inflection in the numbers a huge shift toward agents. The question he answers puts about 70% of the new activity on agents.
Signups moved from around 3 million to 13 million. Managed databases moved from 8 million to approaching 50 million over the same stretch, while creation runs at over a million databases a week. The new activity is mostly agents, and the inflection in the numbers is that shift toward agents.
More builders are coming online
Latent Space: The AI Engineer Podcast · Why Agents Love Supabase
-
How Much Does Supabase Spend on AI Coding?
Paul Copplestone says Supabase does not set one AI coding policy for the whole company. Product teams already choose their own tools and habits, and coding agents were left the same way. A few cost rules are starting to line up. He still wants each team to decide.
Shawn Wang asks how much the company spends and whether the bill is a worry. Copplestone answers with who decides, not with a company-wide cap. Because the company is remote, decisions are written down, so he can ask an assistant what the bill is. The figure he gives is about $300,000 a month for about 450 people. He first says 600 and corrects it to 450. Hiring is fast, so the bill moves with headcount when the tools give the tea
-
Why Can't an AI Agent Read a Folder of SQL Migrations?
Paul Copplestone says an agent cannot see the shape of a database from a folder of SQL migrations. Each change adds another file. A thousand files is a history log, not a picture of the table. The agent has to replay them, or someone has to boot Postgres, run the migrations, and inspect the live database. Both paths are too slow when the agent is editing local files.
Supabase is turning the schema folder in a local project into a declarative description of the database as it is now: tables, policies, and comments. The library is called PG Delta. It works with any Postgres database, with extra hooks for Supabase. Agents mostly read code and need to change it quickly, so the current shape has
-
Why Were AI Coding Agents Breaking Postgres?
Paul Copplestone says a database is harder to leave than a frontend tool, which is why Supabase can stay stickier than a coding app such as Lovable. Shawn Wang had asked whether a product chief ends up more attached to the coding app or to the database. Lovable was also the first of the coding apps to stop its agent from doing damage.
Early agents stuffed image data into Postgres and turned off row-level security just to make a project run. Lovable wrote skills that blocked the moves. He treats the restraint as the useful part. Newer models have stopped making the same mistakes on their own, so the failure is less common than it was.
-
Why Won't Companies Migrate a Production Database?
Paul Copplestone says a move from one production database to another almost never happens, and he would bet it almost never will. Earlier Postgres companies tried to win by pulling people off the system they already run, and the efforts failed. A bigger database is not enough to make a team move.
The plan is to land on day one, on a new project, and still be there much later. The enterprise deals he describes are greenfield: many new databases, not a transfer of the system that already makes the money. Multigres can take a large workload if a team does choose to move, but he does not expect migration to be a focus for about the next five years.
-
Why Do AI Coding Agents Keep Choosing Supabase?
Shawn Wang says his tracking of model answers shows frontier models recommending Supabase as the default integrated backend: the database, auth, functions, and the other APIs in one place. He traces the lead to being in the market for three or four years before the models took off, already the open-source alternative people compared with Firebase.
Paul Copplestone calls the timing an accident and says the job now is to keep the lead. The early habit of spending time with the community left posts, videos, and podcasts about how to build on Supabase, and open source made that material easy for a model to read.
-
Why Do AI Coding Agents Activate Faster Than Human Signups?
Paul Copplestone says the latest jump in Supabase signups came from the models getting better, not from a marketing spend. Most of what they spend on attention is still memes.
A flood of lighter traffic usually hurts activation and the share of people who start paying. Both went up. The agent finishes signup and activation in one step, so activation is almost 100 percent. The new users also started paying at a higher rate than developers who signed up the old way.
Shawn Wang pulls up the signup chart and says the curve keeps forcing a new axis. The unusual part, in Copplestone's account, is that activation and payment rose together as the audience got broader. Friction at the front door wa
-
Why Is Supabase Preparing for Billions of Databases a Month?
Paul Copplestone says Supabase already launches about 5 million databases a month, and the agent path they are planning for is billions a month. A dedicated Postgres cluster for every tiny app would run out of machines. Most of the new databases are small, vibe-coded prototypes, until one of them gets traction.
They acquired Turso for that scale-out. Turso's SQLite work is a file format that can take many writers, which fits a huge number of small databases better than a heavy cluster. Supabase keeps the Turso name on SQLite and adds a small instance, Spark, that uses the same client libraries. When a project takes off, it moves up to full Postgres and later to Multigres. The handoff has to
-
Why Did Supabase Rewrite the Postgres Operator?
Paul Copplestone says every Postgres operator they tested could lose data under some condition. An operator is the software that runs a cluster: extra nodes, failover, and backups. Supabase looked across the ecosystem and found a failure case in each one.
Multigres is the ground-up rewrite of that layer. The original goal was sharding for Postgres. What shipped first is an operator they are willing to trust. They ran a thousand sequential failovers and it did not lose data, and they tried to break it in the other ways they could think of. The release is open source. Teams running serious workloads can ask to use it on the platform. That access starts invite-only.
-
Why Do AI Agents Need Compute Next to the Database?
Paul Copplestone says Supabase Edge Functions, which run on Deno, stop at about two seconds. That limit is too short for the jobs coding agents actually run, such as chunking text and writing embeddings.
Supabase Compute is the longer-running replacement announced for the conference. It is serverless, it sits on the same private network as the database, and a task can open a Postgres connection from there. It is not limited to Deno. A team can use Rust or Python, or run an agent in the instance. He compares it to a container-style Lambda: a full machine beside the data, rather than an edge function that dies at two seconds.
-
Where Did the Supabase Co-founders Meet?
Shawn Wang says Paul Copplestone and Ant Wilson, the Supabase co-founders, went through Entrepreneur First in Singapore, each on a different company, and stayed after the cohort as housemates.
Paul Copplestone says seven people lived in one house, each running a different startup, before Supabase. An accelerator usually pairs a technical founder with a domain expert. Everyone in the house was a builder. Pairing two builders got more done on developer tools, because a developer is already the subject-matter expert. The co-founders kept building together from there, and they hired most of the other people who had lived in the house. Rory, one of them, went on to run go-to-market.
-
How Does Supabase Pay Its Sales Team?
Paul Copplestone says Supabase pays sales only for the growth a help call adds on top of growth the account would have shown anyway. About 60 percent of every Y Combinator batch uses Supabase, and the loyalty they want is the kind Stripe has kept with developers for more than a decade.
The team watches product triggers: an important logo, a customer scaling up, limits, errors, or spend past a line. In the working example, a thousand triggers a week become a thousand emails to people with problems, and about 500 people reply. Accounts that do not reply are the control group, and they grow 25 percent month over month. People who reply get a call whose only job is to help, with a script drawn
-
Does Supabase Track Open Source Adoption?
Paul Copplestone says Supabase deliberately does not measure how far the open-source project has spread. There is no tracking in the open-source build, and anyone can run it.
Enterprises still show up and describe what they already operate. Large pharmaceutical companies have told Supabase they are running about a hundred instances and asked for a product. Control Tower is a view over a fleet of databases. It repackages Supabase for platforms, the product AI companies use when they launch databases for their own users, for a single company that wants many internal projects.
Control Tower is not on the website. Copplestone says the launch might be the next day, and he is not sure they are a