Neon Postgres

Now active for all Neon databases: 𝐂𝐨𝐦𝐩𝐮𝐭𝐞 𝐩𝐫𝐞𝐰𝐚𝐫𝐦𝐢𝐧𝐠 is the first of several features that make compute restarts invisible. Context: We auto-upgrade Postgres minor versions and ship security patches for millions of databases/month. For always-on databases, this requires a compute restart that has 1sec P99. But the compute cache still needs to warm up. Compute prewarming dramatically improves post-restart performance. This is also the first big feature to ship in tandem on Neon and Databricks Lakebase. Thanks to Hans Norheim and team for leading the effort and providing an incredible write-up. (see comments)

Agents can write your code. Now they can provision your database too. 𝚜𝚝𝚛𝚒𝚙𝚎 𝚙𝚛𝚘𝚓𝚎𝚌𝚝𝚜 𝚊𝚍𝚍 𝚗𝚎𝚘𝚗 One CLI command to get a Neon account, database, and credentials. We're excited to participate in Stripe Projects as a co-design partner! https://lnkd.in/eMiBM3jg

We talk a lot about "Serverless Postgres," but what does that actually mean for your daily workflow? By separating storage from compute, Neon gives you a database that actually behaves like modern software: 1. Autoscaling: It scales up when your traffic spikes and scales to zero when it doesn’t, so you don’t pay for idle compute. 2. Branching: Create full copies of your database in seconds. Use them for development, preview deployments, or safe AI sandboxes. 3. Point-in-Time Recovery: Treat your data like code. If something breaks, just roll back to the exact second before the error happened. All the Postgres you love, with none of the infrastructure headaches. Check out the breakdown in the video below!

Agents are great at writing code, but too much of the other work like configuration of local dev environment and provisioning infrastructure, is still on the developer. Specific (YC F25) solves this with an agent-first cloud. Agents get automatic local dev environments and a Terraform-style IaC layer that gives them the infrastructure context they need to develop, deploy, and maintain apps safely. Also the database layer runs on Neon!

Big news: The Neon free plan just got a lot more collaborative. You can now add unlimited team members to your organizations on the free plan. Whether you’re hacking on a weekend project with friends or bringing a collaborator into a new idea, you don't have to worry about seats. Just invite them and start building. We also added unlimited free plan organizations, so you can keep those team projects separate from your personal ones. Check out the video to see how to get your team set up!

WARNING: Reading this post may cause a little jolt of anxiety every time you cancel a query with CTRL+C in psql. George MacKerron is back with another deep dive into one of Postgres’s strangest corners. https://lnkd.in/ewjpEafF

How QwikBuild ships thousands of apps over WhatsApp - each with its own isolated Postgres database - without a huge team or budget? They lean into Neon’s architecture: https://lnkd.in/g5-bjbdq QwikBuild started as a WhatsApp-based AI app builder: describe the app you want in chat, get a live web app in minutes. Every generated app needs real, persistent storage. Their model: 𝐎𝐧𝐞 𝐍𝐞𝐨𝐧 𝐩𝐫𝐨𝐣𝐞𝐜𝐭 𝐩𝐞𝐫 𝐚𝐩𝐩, 𝐩𝐫𝐨𝐯𝐢𝐬𝐢𝐨𝐧𝐞𝐝 𝐢𝐧𝐬𝐭𝐚𝐧𝐭𝐥𝐲 𝐯𝐢𝐚 𝐀𝐏𝐈 Because Neon decouples compute and storage, – idle apps scale all the way down to zero – while storage persists independently – and once traffic is back, compute restarts quickly, autoscaling automatically to match demand Neon's architecture makes it practical to run large fleets where many databases sit inactive, without runaway costs or engineers babysitting infra.

We’ve been partners with Encore from the early days. Branching powers Preview Environments + now Neon also backs apps built by Leap, Encore’s agent: https://lnkd.in/gAhJCBR5

Cursor just launched plugins. The Neon plugin gives you much more than “deploy Postgres” - use it to create environments with real data, test migrations safely, and run instant rollbacks for your apps: https://lnkd.in/gShq4ibS

We ran the resource usage of every production database on Neon through RDS rightsizing algo to see what instance size each would need: • Neon uses 2.4× less compute • Neon is 50% lower cost • The same workloads running on provisioned instances sized using AWS recommendation would hit 55 performance degradations per db per month