To see how we achieved 71% (top 5) on SWE-bench Verified, see this post. Terminal-Bench is an open-source benchmark for evaluating how well AI agents perform on complex tasks that are rooted in the terminal. The tests range from resolving mangled Python dependencies, removing all API keys from a cod...

Learn how to build a Supabase powered AI assistant that combines PostgreSQL with scheduling and external tools for long-term memory, structured data management and autonomous actions.

You may be familiar with this wonderful illustration and accompanying blog post by Henrik Kniberg about good MVPs: It’s a very visual way to illustrate the age-old concept that that a good MVP is not the one developed in isolation over months or years, grounded on assumptions about user needs and go...

We’ve been supporting metrics export to Datadog for a few months, and we just extended this capability to any OLTP-compatible third party – but it has taken us much longer to ship any sort of Postgres log export. You might wonder: why did it take so long? The question is valid, and the answer deserv...

Exporting metrics has been a most requested feature by our users, and for good reason. Having Postgres metrics show up in the same dashboards as your app and infrastructure saves you time and headaches. This post walks through how we built native metrics export in Neon using OpenTelemetry Collectors...

Back in 2014, I had a brush with cervical cancer. We fortunately caught it when it was stage 0, the point at which it's not even called cancer, and is called adenocarcinoma instead. I went in for surgery, a procedure called cervical conization, where the doctor basically scrapes the potentially canc...

The biggest obstacle consumer packaged goods (CPG) and retail companies face isn’t technology or capital. It’s finding talent that can bridge the gap between business strategy and technical execution in real time.

Imagine you ask AI/LLM how to do a particular task by describing what you want to accomplish in code comments.test-examples/spec2.cy.js12345678// this is a Cypress end-to-end testit('changes the label after the click', () => { cy.visit('/') // TODO: there is label element with id "foo" // and a butt...

Today, we’re excited to launch Warp 2.0, the first Agentic Development Environment. In Warp 2.0 you get: The top overall coding agent: #1 on Terminal-Bench (52%) and top-5 on SWE-bench Verified (71%). It features a fundamentally new and superior user interface compared to IDE and terminal coding age...

Dive into rapid tech team growth at TechLeadConf 2025, featuring lessons from Swizec's experience with Plasmidsaurus's explosive expansion. Explore tech debt, AI, and leadership insights from industry experts.

23 game showcases in June, and just as many recommendations! We’re talking about all the big announcements and big summer vibes with Alex, Armando, Matt, and… Matt???

In this blog post, we explore ways in which we can make regular expressions easier to use.

At Neon, our docs team does a little bit of everything. We work on technical documentation, sometimes UI copy, changelogs, reviews, and the occasional regex-heavy cleanup across hundreds of pages. It’s a lot of small, steady work – often, exactly the kind of work you wish an AI agent could just take...

Like many companies that use the cloud, Wealthfront uses AWS Lambda. And while our lambdas look pristine in the web console and perform their highly specific duties well, there’s quite a bit that happens before our lambdas get to that state–I’m referring to CI/CD. However, depending on the given lam...

There are many ways to learn Python online, but there are also many people out there that want to learn Python for multiple reasons - so hey, why not add one more free Python course into the mix? I'm happy to finally release ProficientPython.com, my own approach to teaching introductory Python. The ...

Code is read more than it is written, but it is skimmed more than it is read, so we should write code that is clear at a glance.

SWE-bench is the primary benchmark for evaluating LLMs and AI agents on coding tasks. It assesses a system’s ability to fix problems pulled from real-world GitHub issues on large, complex open-source codebases. Using these realistic coding tasks lets SWE-bench evaluate several factors. There’s the L...

CloudFormation's promise CloudFormation is a declarative infrastructure management tool, meaning you describe what you want, upload, and the service will figure out what changes it needs to make to match reality with the configuration. This works well for some resources. For example, a Lambda functi...

The weird thing about engineering management is that you feel kinda useless. Yet if you stop, projects stop.

One of my colleagues wrote a pretty awesome tool called claude-swarm that orchestrates multiple Claude Code instances as a collaborative AI development team. At Shopify, we are attempting to use it to generate Ruby unit tests at some scale with an army of AI test agents (think a “Ruby Expert” paired...

Six months ago I failed a basic coding interview at a FAANG. Yes, I was a Principal Engineer, and yes, I was paid absurd amounts of money, yet I couldn’t implement a diameter of a binary tree as a “warm up” exercise, 10 lines of code. The interviewer was very nice about it, and was equally surprised...

JavaScript has two common patterns: Maps: We check the existence of a key via .has() before retrieving the associated value via .get(). Arrays: We check the length of an Array before performing an indexed access. These patterns don’t work as well in TypeScript. This blog post explains why and presen...