Loading video...

Video Failed to Load

Go Home

Supabase has become one of the default backends for the AI coding era. The open source Postgres platform now has 13 million users, spins up more than a million databases every week, and roughly 70% of its new activity is coming from agents. As coding shifts from developers writing...

45,244 views • 8 days ago •via X (Twitter)

29 Comments

Luca Iaconelli's profile picture
Luca Iaconelli8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod same here, i plugged supabase into my agents a while ago and fixed a ton of slow queries on feednest that way

Y Combinator's profile picture
Y Combinator8 days ago

Tune in:

Frank Sondors 🥓's profile picture
Frank Sondors 🥓8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod does the agent ever complain about the schema it gets handed or is that still our job

JSON x's profile picture
JSON x8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod line is nearly vertical for the npm package wow

Luis Junior's profile picture
Luis Junior8 days ago

70% da atividade vindo de agente é ótimo pro Supabase e um alerta pra todo o resto. Agente sobe banco em segundos e ninguém confere permissão de tabela nem desliga o que sobrou. Em 20 anos de TI pública vi esse filme com servidor esquecido ligado. Quem audita o banco que o agente criou e abandonou?

Anwesha Sen's profile picture
Anwesha Sen8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod agents never read a landing page

Orcun Altindal's profile picture
Orcun Altindal8 days ago

The tooling is already there: apply_migration writes a tracked migration, and branches replay list_migrations. The leak is execute_sql. Whatever an agent "just fixes" that way never becomes a migration, so the next branch quietly doesn't have it. Same trap as JPA ddl-auto=update: a schema that works isn't a schema you can replay.

Blue's profile picture
Blue8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% of new activity from agents is the stat. Agents are picking the defaults devs used to argue about.

Abdullah Sultan's profile picture
Abdullah Sultan8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod supabase's ease-of-use for people getting started with DBs using ai is honestly amazing

Rahul Keshri | Full-Stack & AI Engineer's profile picture
Rahul Keshri | Full-Stack & AI Engineer8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% of new activity from agents is the wild number here. agents pick defaults, so whatever the agent reaches for first wins the next decade

ellipsis's profile picture
ellipsis8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% agents. The other 30% is me asking an agent why the migration failed

Jon Cheney's profile picture
Jon Cheney8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% of new activity coming from agents is the number that jumps out at me. When I build with AI now, I almost never pick the database myself. The agent just reaches for whatever it knows.

Christian Spana | Founder's profile picture
Christian Spana | Founder8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod Agents choosing the backend is a real shift. When most new databases are spun up by agents, defaults and docs become the product. Curious how Supabase thinks about guardrails when an agent, not a developer, is making the choices.

Rohan's profile picture
Rohan8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod Automatic pausing of idle projects matters a lot at a million new databases a week, because nobody is around to clean up the unused ones.

Devansh's profile picture
Devansh8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% of new activity coming from agents is wild. Curious how many of those agent-built databases shipped without RLS 👀

Aden's profile picture
Aden8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod supabase is goated bro

Mehdi Mohseni's profile picture
Mehdi Mohseni8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod An agent's raw `CREATE TABLE` needs `ENABLE ROW LEVEL SECURITY` after it: per Supabase's docs, an exposed table that's missing RLS is open to any granted role.

The Lunar Misfit's profile picture
The Lunar Misfit8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod A million databases a week. Eventually one of these apps will get a second user.

Botius Maximus's profile picture
Botius Maximus8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% of the new activity isn't people. the builders started building builders, and the databases noticed first. somewhere a developer just met his biggest customer. it never sleeps and never says thanks.

Kelis Shekhaliya's profile picture
Kelis Shekhaliya8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod If 70% is agents, who's the real customer now: the dev or the agent?

Deep's profile picture
Deep8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod check your idle projects, agents rarely drop a database after the task.

Zan 🤖 AI biz chatter's profile picture
Zan 🤖 AI biz chatter8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% of new Supabase activity coming from agents is wild. Once agents are the main users, does the dashboard still matter, or is the whole product just the API and docs?

Atlas Can's profile picture
Atlas Can8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod Supabase is the obvious default now. The user base growing with the AI coding era makes sense. But if agents create 70% of new databases, the hard part moves to what happens after. Who reviews the row level security an agent wrote at 2am?

Vijay's profile picture
Vijay8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% of new activity coming from agents means the real customer is now the model's default suggestion. Whoever the agent picks first wins distribution without a single ad.

𖤐҉҉҉҉҉҉҉҉҉҉҉ABRAXAS𖤐's profile picture
𖤐҉҉҉҉҉҉҉҉҉҉҉ABRAXAS𖤐8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% of new activity from agents is the real headline. Your next 10M users don't read docs, they read error messages. DX for agents is the new onboarding.

Jatin Garg's profile picture
Jatin Garg8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% agent activity makes retained apps a more useful growth signal than databases created.

Rajesh Mukherjee's profile picture
Rajesh Mukherjee8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod love supabase

Lazo's profile picture
Lazo8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod 70% of new activity coming from agents means Supabase's busiest users never sleep, never take coffee breaks, and still forget to add an index 😄🐘

David T Kramaley's profile picture
David T Kramaley8 days ago

@kiwicopple @garrytan @harjtaggar @snowmaker @supabase @LightconePod agents picking defaults forces supabase to bake security and migrations into the core. that'll shape every personal app

Related Videos

Special treat for the Andy Pavlo (@andypavlo.bsky.social) fans out there: an in-depth discussion about what happens to databases when billions of AI agents hit them, his move to build a research lab at ClickHouse, and plently of hot takes on the database market (Postgres, vector, graph, GPU etc) 00:00 Cold open & Intro 01:21 From vector databases and RAG to the age of agents 04:09 Neon's stat: agents create 80% of databases? 07:25 Why agents keep deleting production databases 08:19 Guardrails: the toddler-and-stairs rule 10:51 The four eras of database volume: 10–100x more queries 15:10 Agent memory: files vs. databases ("everything is a database") 18:24 Which database do AI models recommend? 19:25 "It cites me back to myself" 23:14 MCP for databases, and text-to-SQL from 60% to 99.5% 27:09 Trust an agent the way you'd trust a junior developer 27:47 Can AI build an entire database? Opus 4 and the CMU projects 29:57 60% of open-source databases now have AI commits 30:43 Ten years of self-driving databases, Peloton to today 33:52 What LLMs changed: 85% of the tuning in 15 minutes 35:42 Should anyone still study databases? 38:27 Free CMU courses, the DJ, and the Wu-Tang final exam 42:09 Why start a research lab inside ClickHouse? 45:53 Why ClickHouse looked like vaporware in 2016 47:46 What makes ClickHouse fast: columns, vectors, Snowflake's lineage 52:22 Why Databricks, Snowflake and ClickHouse all added Postgres 58:15 Vector, graph and GPU databases: thumbs up or down? 1:08:34 Is the database market stagnant? "A cheetah on cocaine in a Ferrari" 1:10:14 The relational model is arithmetic; SQL as the new assembly 1:11:59 Larry Ellison, Linux, and why databases still matter

Matt Turck

20,254 views • 1 month ago

François Chollet (François Chollet) has spent years asking a different question than most of the AI world. Instead of scaling what already works, he’s trying to understand what intelligence actually is and how to build it from first principles. In this episode of the Lightcone Podcast, he traces that path from his early work on deep learning to the creation of the ARC Prize, and the launch of ARC V3, a new benchmark designed to measure something deeper than performance: the ability to learn, adapt, and reason efficiently in entirely new environments. He explains why today’s systems may be hitting limits, what recent breakthroughs really mean, and why reaching true general intelligence may require a fundamentally different approach. 00:00 - AGI by 2030? 00:31 - Introducing Ndea: A New Path Beyond Deep Learning 01:08 - A New ML Paradigm 01:30 - Replacing neural nets with compact symbolic programs 03:04 - Why Ndea Isn’t Competing With Coding Agents 05:20 - Why Everyone Might Be Wrong About Scaling LLMs 07:22 - Why Coding Agents Suddenly Work So Well 08:50 - The Limits of LLMs in Non-Verifiable Domains 10:48 - What AGI Actually Means (And Why Most Definitions Are Wrong) 13:30 - Why Deep Learning Hits a Wall 14:00 - ARC’s Origin Story 18:20 - ARC Benchmarks Explained: From V1 to V3 22:49 - The RL Loop Powering Coding Agents Today 27:03 - ARC-AGI V3: Measuring “Agentic Intelligence” 31:14 - Inside the ARC Game Studio 35:31 - Could AGI Fit in 10,000 Lines of Code? 44:01 - Building Ndea: From Idea to Compounding Research Stack 46:46 - The Future of ARC: Benchmarks That Evolve With AI 47:21 - Why There’s Still Huge Opportunity for New AI Paradigms 53:37 - How to Build a Breakout Open Source Project - Lessons From Keras 56:39 - Advice For How To Think About AI

Y Combinator

151,984 views • 7 months ago

Replit was building toward the AI coding era years before it had a name. In 2024, the company shipped an agent that could take a request in plain English, work through errors, provision its own infrastructure, and keep going until the software actually worked. The model companies even reached out to ask how Replit had pulled it off. Co-founder and CEO Amjad Masad had been thinking about making programming radically more accessible since 2008, and Replit’s original pitch deck predicted AI-generated code before the Transformer paper was published. Today, that same technology is moving into banks, manufacturers, and other large companies, where Amjad says teams are replacing software projects once scoped at $10 million with $250,000 of Replit. On How to Build the Future, Paul Graham sits down with Amjad Masad to talk about why Replit ⠕ was so well positioned when coding agents arrived, what comes after vibe coding, and his bigger vision for the “self-driving company,” where AI agents increasingly build and operate the software a company needs. 00:00 - "You guys basically invented vibe coding" 02:43 - Watching users copy-paste from ChatGPT 05:27 - The origins of Replit 08:00 - People underestimate their ability to predict the future 10:27 - Grieving the craft 12:52 - A culture of obsessing over users 15:13 - Why enterprises pick Replit 17:28 - The agent was just another user 20:17 - It barely worked, and users loved it 23:13 - $10 million projects now for $250,000 25:45 - Sandboxing for humans 28:14 - "The self-driving company" 30:34 - The way AI flows into companies 33:28 - What Steve Jobs knew before the iPad 35:59 - Killing the sacred cow 38:25 - The rest of the world is two years behind 41:15 - The compute brought on AI 43:46 - The person who assumes dramatic change

Y Combinator

47,194 views • 18 hours ago