https://www.rajasleem.com/tags/build-in-public/Recent content in Build-in-Public on Raja Sleemhttps://www.rajasleem.com/images/rajasleem-og-thumbnail-1200x630.pnghttps://www.rajasleem.com/images/rajasleem-og-thumbnail-1200x630.pngHugoen-usWed, 26 Aug 2026 09:00:00 -0400https://www.rajasleem.com/posts/vibe-coding-is-not-a-yes-or-no-question/Wed, 26 Aug 2026 09:00:00 -0400https://www.rajasleem.com/posts/vibe-coding-is-not-a-yes-or-no-question/I had a hunch the vibe coding debate was all blanket verdicts. So I counted 51 takes before publishing mine. I was half wrong, and the half I was wrong about made the real pattern more interesting.https://www.rajasleem.com/posts/agentic-workflow-part-8-trust-but-grep/Wed, 19 Aug 2026 09:00:00 -0400https://www.rajasleem.com/posts/agentic-workflow-part-8-trust-but-grep/Almost half the questions my AI pipeline stopped to ask me were already answered on the hard drive. The plan reasoned flawlessly, from a premise nobody spent five seconds checking.https://www.rajasleem.com/posts/agentic-workflow-part-7-the-paperwork-problem/Wed, 12 Aug 2026 09:00:00 -0400https://www.rajasleem.com/posts/agentic-workflow-part-7-the-paperwork-problem/I finally pointed my autonomous pipeline at two real repos. One paid 4.7x more per line of code than the other, and the culprit was not the model. It was my own specs.https://www.rajasleem.com/posts/agentic-workflow-part-6-obedience-is-futile/Tue, 04 Aug 2026 09:00:00 -0400https://www.rajasleem.com/posts/agentic-workflow-part-6-obedience-is-futile/The bottleneck in an agentic workflow is still you. Not your attention this time, your context window. Here is the guardrail I built so my own limits stop quietly steering what I build.https://www.rajasleem.com/posts/agentic-workflow-part-5-two-ways-to-run-a-loop/Wed, 22 Jul 2026 09:00:00 -0400https://www.rajasleem.com/posts/agentic-workflow-part-5-two-ways-to-run-a-loop/I have an autonomous build loop. Now there are two ways to run it: a portable setup of agents and skills, or a dynamic workflow that spins up many agents at once. I'm testing both.https://www.rajasleem.com/posts/agentic-workflow-part-4-letting-it-run/Wed, 15 Jul 2026 09:00:00 -0400https://www.rajasleem.com/posts/agentic-workflow-part-4-letting-it-run/Once agents could build to a shared contract and be checked against it, I stopped driving every step by hand. This is the loop that runs a batch of features mostly on its own.https://www.rajasleem.com/posts/agentic-workflow-part-3-when-agents-diverge/Mon, 06 Jul 2026 09:00:00 -0400https://www.rajasleem.com/posts/agentic-workflow-part-3-when-agents-diverge/Run several agents on the same project and they each build something reasonable that doesn't match the others. Here's the fix: write the rules down once, before anyone builds.https://www.rajasleem.com/posts/agentic-workflow-part-2-surviving-the-context-window/Sun, 07 Jun 2026 09:00:00 -0400https://www.rajasleem.com/posts/agentic-workflow-part-2-surviving-the-context-window/One Claude session is never enough for real work. This is how I hand off between sessions, push work onto sub-agents to save context, and finally see all the planning files at a glance.https://www.rajasleem.com/posts/agentic-workflow-part-1-from-prompts-to-a-process/Mon, 01 Jun 2026 09:00:00 -0400https://www.rajasleem.com/posts/agentic-workflow-part-1-from-prompts-to-a-process/The first thing I fixed about working with an LLM wasn't the model. It was me starting from scratch every session. Here's the simple process that changed that.https://www.rajasleem.com/posts/ai-fatigue/Sun, 31 May 2026 12:00:00 -0400https://www.rajasleem.com/posts/ai-fatigue/The tools got faster. My brain didn't. Why four hours of building with an LLM wipes me out, and why I doubt the 'spin up 200 agents' future.