Blog

Every time the chat broke, Claude recommended streaming as the fix. I kept saying no.

The author details why existing platforms couldn't meet their need for a custom 'hub' with a shared identity layer across multiple subdomains. This led them to build 16channels.com from scratch, leveraging AI for efficient project scoping.

AI as a database editor, using tool calling, can replace traditional forms by directly modifying data through natural language conversations. This enables AI to make judgment calls and iteratively refine content, contrasting with forms that demand upfront answers.

AI agents often struggle with code reviews due to anchoring bias, where prior context leads to confirmation instead of critical analysis. A structured process is proposed, emphasizing analyzing code diffs first to derive behavior before consulting descriptions or tickets.

The post explains how to implement "fire-and-forget" async AI jobs using Next.js 15's `after()` API, ensuring a fast user response. It details handling individual job failures and updating the UI via client-side polling based on data presence.