sauver-org/sauver

Sauver is an extremely early-stage project (27 days old, 1 star, zero velocity) with a compelling privacy-focused narrative but critically weak defensibility signals. The project appears to be a proof-of-concept combining commodity capabilities: spam filtering (solved by Gmail, Outlook, Apple Mail for decades), tracking pixel detection (implemented in uBlock Origin, Privacy Badger, browser DevTools), and local LLM inference (standard via Ollama/llama.cpp). The README provides mission statement and motivation but lacks technical depth, implementation details, or evidence of working code. Without visible GitHub activity, open issues, or commit history, this reads as a concept document rather than a deployable product. Platform domination risk is HIGH because: (1) Email providers (Google, Microsoft, Apple) have spam ML as core competency and are already detecting AI-generated content; (2) Browser vendors (Chrome, Firefox, Safari) can add tracking detection natively; (3) Enterprise security vendors (Proofpoint, Mimecast) already solve this at scale. Market consolidation risk is MEDIUM because anti-spam/privacy software exists (Superhuman, Hey, ProtonMail, etc.) but the specific 'local-first AI agent' angle is underexplored—however, this is a feature request, not a defensible product. Displacement horizon is 6 months because a well-resourced competitor (or platform feature) could ship this capability trivially. The project has no network effects, no data gravity, no community lock-in, and no technical moat beyond executing a straightforward integration. Novelty is REIMPLEMENTATION: applying known LLM classification + pixel detection to email/web tracking—neither technique is novel. Composability is APPLICATION because it appears designed as a standalone tool, not a library or component. Implementation depth is PROTOTYPE given the project age and activity signals. This project needs: (1) working proof of concept with measurable spam detection accuracy; (2) comparison benchmarks vs. native email filters; (3) community adoption beyond GitHub presence; (4) clear technical differentiation (e.g., custom dataset, novel detection heuristic). Without these, it will be absorbed into platform features or outcompeted by privacy-focused email services within 6–12 months.