nicksolarsoul pitches a privacy gateway that scrubs names before they hit the AI
An opt-in routing layer would strip identifying details out of prompts before Mike sends them to Anthropic, Google, or OpenAI.
The idea: sit a third-party service called Hey Jude between Mike and the big AI providers. Before any prompt leaves for the model, Hey Jude swaps out identifying entities - person names, company names - so the raw details never appear in what gets sent downstream. It was wired up for all three model providers Mike uses and left switched off by default, so nothing changes for existing deployments unless a team deliberately turns it on.
There's an honest catch, and nicksolarsoul flagged it plainly: this only protects what leaves for the AI. Mike still keeps the original, un-scrubbed chat text in its own database. So it's a control over third-party exposure, not full anonymisation. Worth noting too that the proposal was closed without being merged, so it never landed in the shared codebase - it stands as a blueprint rather than a live feature.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?