Denislav Gavrilov — AI engineer, writer, and speaker. Claude Code experiments (Clopus), an autonomous Kubernetes SRE (A2W Lucas), and writing on building with AI agents.
Databricks bought Neon and turned it into Lakebase, so an agent can get its own branch of production in seconds and the rows it writes land in a governed table you can query. I walk through what that gives you, where it stops (nothing merges back), and what it means for getting an agent past read-only.
Denis ran the same task through ten isolated Claude Code agents at once — each with its own git worktree and its own Postgres branch — and diffed what all ten did to the database. They agreed completely on the schema, then split 5/5 on two decisions nobody asked them to make. Where they split is the cheapest eval you will ever run.
After two years on Supabase, I moved all my new projects to Neon and wrote down why: they are an agents-first company that ships everything to the API, they develop at very high velocity, and branching.
pastmonday.com is a real SaaS: auth, billing, uploads, and AI chat. A software factory of four Claude Code skills built it in about 40 minutes, and every backend service on the bill is Neon, including the brand new Object Storage and AI Gateway.
Neon releases features to the API before anywhere else, which makes it perfect for agentic coding: your agent can build what the dashboard doesn't have yet. To prove it, I had Claude Code build a real hard spending cap (the one Neon only offers as alert-only), with two interactive widgets you can play with.
Continuing on Part 1 (setting up your developer environment) and Part 2 (defining your project to Claude Code), we are now jumping into Part 3, in which we are going to be building the landing page for our application.
Assuming you have followed part 1 of this series, you now have a beautiful, customized development environment, waiting for you to start using it.
Everyone says you can't code if you are not a developer. I say: this is false, and today anyone can build anything! This is the first part of a 6-posts series on how to build any app with AI agents.
Continuing on my agentic workflows exploration, I expand on my previous Supabase implementation of preview environments, and move to Neon for true branching. I get a solid implementation.
Practical exploration (and recreation) of the `Mini Shai-Hulud` supply chain attack.
I make an iOS app, but I am concerned non-iOS users might want to engage with the idea so I find a way to allow web payments with Stripe & bind them to a RevenueCat entitlement. Apple accepts it.
I decide to learn the basics of AI/ML. I do, with Prime Intellect's "Hosted RL Training", Modal, and Codex. I fine-tune a LoRA model with synthetic data in the shape of a surveillance agent.
A company sends a cold email to my girlfriend's sister. They ask for $400 to create a landing page for her new dental studio. I told her to drop the email in spam and made her a Github account instead
Complete self-hosted natural language programming (""vibe coding"") platform with Github, ArgoCD, and Claude Code. Essentially, a far better and functional Replit/Lovable.
A practical introduction to GitOps & the power of preview environments: a fully functional isolated environment, spawned automatically on every pull request.
I put Claude Code in a cronjob in a k8s namespace. It is tasked to monitor an application and in the unfortunate case of application error (degraded pods), it is tasked to do a hotfix and document it.
A Claude Code instance runs without any human action for 24-hours. I gave it short-term (sqlite3) & long-term (qdrant) memory, as well as access to a browser.
I attempted to create a fully-autonomous Claude Code. While I did not reach full autonomy, I got pretty close. In this piece I share the stats, how it went, and well.. how it made me feel.
Notes on building a self: Purpose through escapism.
In the age of AI, those creating patterns will be rewarded with high amounts of satisfaction, while those with no soul to sign their work with, will be stuck in a no-reward purgatory of copying others
I ordered tartare, then I put it in a pan and posted photos of it on social media. Then ~5 million people hated me for it. Here's a story about food, technology, and the people.