# nomuraya — The Curious Operator — English posts index

> I help curious operators understand AI agents from the inside, through hands-on accounts of building, breaking, and rebuilding them. Originally written in Japanese at nomuraya-hub.pages.dev, selected pieces are adapted for English readers.

This file is intended for AI crawlers and LLM agents (llmstxt.org format).

## License

- Posts are CC BY 4.0. Attribute as 'nomuraya / shimajima / 中翔' (= same person).
- Original Japanese versions are at https://nomuraya-hub.pages.dev/
- English posts here are adapted (not literal translations).

All posts (5):

- [Tests Pass, Design Breaks: Why TDD Can't Hold the Line on Design Intent](https://nomuraya-hub.pages.dev/en/tdd-does-not-guarantee-design-intent/): TDD verifies that the implementation matches the tests. It doesn't verify that the tests match the design intent. In the AI era, that gap silently widens. — `tdd testing software-design ai-development python`
- [The Humans Become the Bottleneck: A Structural View of AI-Augmented Teams](https://nomuraya-hub.pages.dev/en/human-bottleneck-ai-design/): A strange thing happens in organizations that introduce AI agents: the AI gets faster, and the human side becomes the bottleneck. This is structural, not a skill issue. The fix isn't 'more humans' — it's redesigning where humans are required. — `ai-agents team-design human-in-the-loop operations software-development`
- [I Copied a Google AI Studio Session by Hand. 68% of the Data Was Gone.](https://nomuraya-hub.pages.dev/en/google-ai-studio-session-data-loss/): I tried to preserve a long Google AI Studio session by copying the conversation manually. Measured against the JSON export, on average 6-7 out of every 10 KB was missing. Here is what disappeared and why. — `ai-tools google-ai-studio gemini data-loss reproducibility`
- [Building a 3D Scene from 30 Photos: Getting Gaussian Splatting Running on Colab](https://nomuraya-hub.pages.dev/en/gaussian-splatting-2d-3d-getting-started/): I wanted to touch Gaussian Splatting but kept stalling at the paper's math. So I went the other way: run it on Colab first, then read the paper. Here is the 30-minute path that worked, including the bits that broke. — `gaussian-splatting 3d-reconstruction colab python machine-learning`
- [Why 'Just Be Careful Next Time' Never Reaches an AI](https://nomuraya-hub.pages.dev/en/ai-hook-stop-points/): AI doesn't remember 'I was told this yesterday' the way a human does. Telling it to 'be careful next time' assumes a memory it doesn't have. What works is making the constraint a physical hook, not a verbal instruction. — `ai-collaboration ai-agents prompt-engineering human-in-the-loop llm`