bedrockscience.com

What It Is

Bedrock Science LLC is a privately held, self-funded software company based in Washington State, USA. It positions itself as building “privacy-first AI software for knowledge work.” Its core focus is not simply another chatbot, but enabling modern AI assistants to use more persistent personal context for individuals and small teams, while avoiding the need to hand sensitive materials over to a third-party cloud.

Core Capabilities and Privacy Design

According to the official website, its first product is still under development, with macOS as the initial priority and plans to support iOS and more platforms later. The product is designed around a user-owned and user-controlled local file vault that provides durable memory for AI assistants. Data stays on the user’s own devices and in their own files, and the file format remains readable, which should in theory help with long-term portability and reduce vendor lock-in. The website does not disclose what AI models are used, whether a local model is built in, whether external LLMs are supported, or whether it includes retrieval-augmented generation or multimodal capabilities, so its actual AI capabilities cannot yet be assessed.

Pricing, Trials, and Integrations

There is currently no public information on pricing, free tiers, trial policies, or payment methods. APIs, plugins, and third-party tool integrations have not been disclosed either. The website only emphasizes that the files can be read by any AI system, which sounds more like a product philosophy around open data formats and portability than a clearly launched integration ecosystem.

Pros and Cons

Its strengths lie in its clear positioning: it targets privacy-sensitive knowledge workers and addresses the problem of long-term AI assistant memory and user control over personal context. Local storage, user-owned files, and readable formats are all attractive design choices. The drawbacks are also obvious: the product has not yet been released, and there are no screenshots, feature lists, model details, pricing, or roadmap. Support currently appears to be limited to a general email address, making its commercial readiness and reliability impossible to judge.

Who It’s For and Access from China

It is best suited to individuals, small teams, and research-oriented knowledge workers who care about local-first software, data sovereignty, and long-term knowledge management. For teams that need something deployable right away, it is currently better to wait and see, or consider alternatives such as Obsidian, Notion AI, Mem, Rewind, Pieces, or local RAG knowledge bases. The official website does not state anything about access from mainland China, network availability, or payment methods, so these remain unknown.