> For the complete documentation index, see [llms.txt](https://yvechat.gitbook.io/wp/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://yvechat.gitbook.io/wp/problems.md). # Problems The AI market solved access. It left custody unresolved. *** ### The Custody Problem #### 1) Your reasoning lives on someone else's server Every conversation with a hosted assistant is written to infrastructure controlled by the vendor. The record of how a person thinks, researches, and decides accumulates in a place they cannot inspect, move, or switch off. #### 2) Your credentials leave your machine Third-party AI tools routinely ask users to paste an API key into a web form. That key then travels to a server the user does not operate. Both Google and OpenAI instruct developers in writing to keep keys server-side and never expose them in client-side code, guidance that a browser-based key form directly contradicts. Source: [Google Gemini API key guidance](https://ai.google.dev/gemini-api/docs/api-key) #### 3) Your identity is a database row Access to a workspace depends on an email address and a password held by the vendor. Account recovery, suspension, and deletion are decisions made by someone else about an identity you were only ever renting. *** ### The Lock-In Problem #### 4) One vendor, one model, one bill Committing to a single assistant means committing to whichever models that vendor licenses, at whichever price they set, with whichever policies they apply. Model quality moves quickly, and a workspace tied to one provider inherits that provider's ceiling. #### 5) Inference margin is charged on top of inference cost Subscription pricing bundles the cost of the model with the vendor's margin. Users pay for tokens twice: once in the underlying provider cost, and again in the platform fee layered above it. *** ### The Trust Problem #### 6) Tool use happens without visibility Assistants increasingly browse, execute code, and call external services. The Model Context Protocol specification, which standardizes this behavior, states that tools "represent arbitrary code execution and must be treated with appropriate caution" and that hosts "must obtain explicit user consent before invoking any tool". Products that run tools silently operate outside that guidance. Source: [MCP specification, Security and Trust & Safety](https://modelcontextprotocol.io/specification/2025-06-18/index) #### 7) Local models remain out of reach for most people Running a model on your own hardware is the strongest available privacy guarantee. It is also gated behind runtime installation, model selection, quantization choices, driver versions, and command-line operation. The people who would benefit most from local inference are the least likely to reach it. *** ### Why Now Three conditions arrived together. Consumer hardware now runs capable open-weight models locally. Ollama reduced local model serving to a single installer. And a large population already understands custody as a first principle, because they learned it with money. The audience, the hardware, and the belief are all in place. *** Every one of these problems has the same shape. Someone else is holding something that belongs to you. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://yvechat.gitbook.io/wp/problems.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.