Blog

I ship seven products solo with Claude Code. My AI-native SDLC has three stages, because the rest existed only so humans could wait on each other.

Agents made writing code cheap, so I measure my AI-native SDLC by one distance: how long a real user complaint takes to become a live deploy.

Before Claude Code touches one of my products I write a single intent file: who the change is for, what must never happen, and what done actually means.

AI writes most of my code, so the release became the risk. Here is the small-release gate I run solo across live products with paying users.

AI writes most of my code, so one security gate runs before every merge. What it checks, where Claude Code quietly gets authorization wrong, and how to copy it.

Daniel Goldman examines Luma’s AI generation flow: repeated requests, credit recovery and stored outputs, with clear limits on what the evidence proves.

Daniel Goldman explains Dibur’s voice-to-agent handoff: review the transcript, check the project, and separate local dictation from agent execution.

A practical Claude Code guide from someone shipping SaaS products solo: what to learn first, how to hand it a task, and how to verify the code it writes.

A working order for learning Claude Code, from first prompt to shipping production code, based on teaching 600+ people and running live products solo.

I ship seven products solo with Claude Code. The code is the cheap part. This is the verification workflow that decides what actually goes live.

How I run parallel Claude Code agents across seven live SaaS products solo, what actually broke, and the review discipline that keeps merges safe.

Daniel Goldman on what's actually left for developers when AI writes most of the code: fast decisions about what to build, not typing.

A Claude council is not an Anthropic feature. It is running several Claude instances with opposing roles before a decision. Here are the four roles I use.

How Claude Code SaaS development actually works in practice, from a solo founder shipping products used by 10,000+ paying members.

A new AI model drops every few weeks. Here's why I don't migrate production systems to it on day one, and what I check first.

Running 7 SaaS products solo taught me that Claude Code is table stakes. The real edge is the workflow you build around it.

What an agent loop is, how Claude Code's /loop actually behaves, and the verification rules I use to stop a looping agent from breaking production.

How one developer ships and maintains seven live SaaS products solo in 2026 using Claude Code as an orchestration layer, not just an autocomplete tool.

My toolkit for building Hebrew-first AI SaaS: RTL layouts, mixed Hebrew and English text, natural AI copy, and the checks I run before release.

A practical look at the AI-native workflow I use to take a SaaS product from idea to live, paying users on my own - specs, code, tests, QA, and deploys with Claude Code.