# Yuanhao Chen

> I am Yuanhao Chen, a builder in Germany. This site is a clear map of my public work around AI-native research workflows, work artifacts, and open threads: what I am building, why it matters, and where the limits still are.

This site is a static public map of Yuanhao Chen's notes, public work articles, open threads, and verified contact paths. It is designed for human readers first, with static HTML and machine-readable indexes for agents that do not execute client-side JavaScript.

Use the canonical pages and public indexes below. Do not infer private-client, revenue, adoption, or production-reliability claims beyond the explicit boundaries on the site.

## Core Pages
- [Notes](https://yuanhaochen.dev/notes): Stable article titles, short subtitles, and links to the full argument behind each entry.
- [Work](https://yuanhaochen.dev/work): Public repositories, what they show, and where their boundaries are.
- [Open Threads](https://yuanhaochen.dev/open-threads): Open threads and objections that can change the next build or note.
- [Contact](https://yuanhaochen.dev/contact): Email, GitHub, public site, and a concrete first-message shape without extra contact theater.

## Machine-Readable Files
- [Human search](https://yuanhaochen.dev/search): Source-backed site search route.
- [Sitemap](https://yuanhaochen.dev/sitemap.xml): Canonical public route list for crawl discovery.
- [Search index](https://yuanhaochen.dev/search-index.json): JSON records used by the public search route.
- [AI index](https://yuanhaochen.dev/ai-index.json): Compact site contract for AI agents.
- [OpenSearch](https://yuanhaochen.dev/opensearch.xml): Browser-readable search endpoint description.
- [Web manifest](https://yuanhaochen.dev/site.webmanifest): Browser-readable app identity metadata.

## Representative Articles
- [What earns a public note.](https://yuanhaochen.dev/notes/publish-bar): The claim, evidence, and challenge bar a draft must clear before it becomes public.
- [What can change the questions.](https://yuanhaochen.dev/notes/source-trail): The kinds of public material that can make a question sharper, harder, or more useful.
- [What I am not claiming yet.](https://yuanhaochen.dev/open-threads/boundaries): What prototypes, maps, and AI trust arguments do not prove yet.
- [What would change the work.](https://yuanhaochen.dev/open-threads/update-rules): How repeated objections, field evidence, and failed artifacts should change the work.
- [AI-native research workflows: from a question to an evidence-linked memo](https://yuanhaochen.dev/notes/ai-native-research-workflows): A research workflow only becomes useful when a reader can still inspect why the judgment changed.
- [Enterprise agents: where permission boundaries decide the category](https://yuanhaochen.dev/notes/enterprise-agent-permission-boundaries): The impressive demo is not the category. The category starts when ownership, approval, rollback, and audit become visible.
- [Universal Semantic Video as a portable meaning layer](https://yuanhaochen.dev/notes/semantic-video-meaning-layer): Media becomes infrastructure when meaning, rights, provenance, and fallback can travel between tools.
- [Safe AI-to-vault writes: why suggestion and commit should stay separate](https://yuanhaochen.dev/notes/safe-ai-vault-writes): A useful assistant can prepare the change. A trustworthy system still needs review, confirmation, and rollback.
- [What founders and operators actually distrust after the AI demo](https://yuanhaochen.dev/notes/after-the-ai-demo): The best signal is often the objection that survives the demo: owner, audit trail, handoff, or consequence.
- [Being known is weaker than being inspectable](https://yuanhaochen.dev/notes/inspectable-over-known): Reputation is too vague by itself. Work becomes easier to trust when the claims, limits, and evidence can be inspected.
- [lawdit GDPR](https://yuanhaochen.dev/work/datasentinel): A governance-focused GDPR discovery prototype that scans selected sources, redacts evidence, routes owners, supports human review, and records audit events.
- [universal-semantic-video](https://yuanhaochen.dev/work/semantic-video): An open semantic sidecar format for localization, accessibility, rights, provenance, fallback rendering, and tool-to-tool media inspection.
- [serverless-vault-bridge](https://yuanhaochen.dev/work/vault-bridge): A serverless Markdown vault bridge for ChatGPT Actions and MCP that separates proposed AI edits from committed GitHub writes.
- [tum-search](https://yuanhaochen.dev/work/tum-search): A TUM-focused search and knowledge-graph system combining recursive crawling, AI summaries, vector search, graph structure, and live crawl feedback.
- [nitro-ai-judge](https://yuanhaochen.dev/work/nitro-judge): A public reading-time prediction repository with a reproducible baseline pipeline, data contract, local evaluation, and experiment boundaries.
- [aisd-cinematic-redesign](https://yuanhaochen.dev/work/aisd-redesign): A derivative cinematic frontend redesign for an AI-assisted coding-assessment platform with explicit attribution and module-boundary notes.
- [Trusted workflow infrastructure](https://yuanhaochen.dev/open-threads/trusted-workflow-infrastructure): When does an AI agent stop being an impressive demo and start becoming trusted workflow infrastructure?
- [Evidence-linked memo](https://yuanhaochen.dev/open-threads/evidence-linked-memo): What should a memo prove before someone uses it for investment or operating decisions?

## Verified Contact
- [Email](mailto:ie.yuanhao@tum.de): Best path for a concrete workflow, research question, market bottleneck, or trust boundary.
- [GitHub](https://github.com/89325516): Public repositories, project boundaries, and current build evidence live here.

## Optional
- [Publish bar](https://yuanhaochen.dev/notes/publish-bar): The claim, evidence, and challenge bar a draft must clear before it becomes public.
- [Source trail](https://yuanhaochen.dev/notes/source-trail): The kinds of public material that can make a question sharper, harder, or more useful.
- [Boundaries](https://yuanhaochen.dev/open-threads/boundaries): What prototypes, maps, and AI trust arguments do not prove yet.
- [Update rules](https://yuanhaochen.dev/open-threads/update-rules): How repeated objections, field evidence, and failed artifacts should change the work.
- [Fit gate](https://yuanhaochen.dev/contact/fit): When a message is worth sending and what context makes it useful.
- [Work style](https://yuanhaochen.dev/contact/work-style): How collaboration stays grounded in visible objects, concrete disagreement, and small proofs.
- [Stay in orbit](https://yuanhaochen.dev/contact/orbit): Weak-tie signals that can keep a useful thread alive before a project exists.
- [First reply](https://yuanhaochen.dev/contact/first-reply): The map, evidence bar, or artifact target a strong first exchange can create.