Writing about building software, AI-native products, and long-term thinking.https://ralak.me/ https://ralak.me/blog/voice-is-a-system-not-a-setting/https://samuelralak.com/blog/voice-is-a-system-not-a-setting/Why every AI writing tool defaults to the same generic register, what voice actually is at the systems level, and the case for AI as an editing aid rather than a generation engine. Notes from building Hadithi.Sun, 26 Apr 2026 00:00:00 GMT https://ralak.me/blog/ai-will-confidently-build-you-the-wrong-thing/https://samuelralak.com/blog/ai-will-confidently-build-you-the-wrong-thing/An 851-line spec matrix, a deliberate workflow, and the argument that AI-assisted development doesn't need better prompts — it needs better thinking. A case study in building Smith, the AI orchestration library behind Hadithi.Sat, 21 Mar 2026 00:00:00 GMT https://ralak.me/blog/ive-used-my-own-ai-to-write-every-post-for-the-last-two-weeks/https://samuelralak.com/blog/ive-used-my-own-ai-to-write-every-post-for-the-last-two-weeks/Three posts in two weeks after months of silence. How building Hadithi and dogfooding my own AI writing tool turned friction into a workflow and broke a long publishing drought.Tue, 17 Mar 2026 00:00:00 GMT https://ralak.me/blog/the-emperors-new-agent/https://samuelralak.com/blog/the-emperors-new-agent/OpenClaw just passed Linux on GitHub's star leaderboard, but a 58.9% safety pass rate and 0% intent understanding score tell a different story. A firsthand account of what happens when you trust an AI agent with a real goal.Wed, 11 Mar 2026 00:00:00 GMT https://ralak.me/blog/the-new-senior-skill-reviewing-code-you-didnt-write-and-dont-understand/https://samuelralak.com/blog/the-new-senior-skill-reviewing-code-you-didnt-write-and-dont-understand/AI has made writing code faster, but it has made senior review harder. Why the real bottleneck now is judging whether plausible code actually belongs in your system.Wed, 11 Mar 2026 00:00:00 GMT