DataFrontier — Lab Notes https://datafrontier.co/blog Research-tone essays and lab notes from DataFrontier on memory as infrastructure. en DF-008 — Compliance by Architecture: DPDP, GDPR, and the End of Paperwork-as-Privacy https://datafrontier.co/blog/compliance-by-architecture https://datafrontier.co/blog/compliance-by-architecture 21 Apr 2026 00:00:00 GMT When residency and egress are structural, compliance becomes a property of the system — generated, not asserted. DF-007 — Why We Open-Sourced the Core https://datafrontier.co/blog/why-we-open-sourced-the-core https://datafrontier.co/blog/why-we-open-sourced-the-core 07 Apr 2026 00:00:00 GMT The substrate should be inspectable; the enterprise layer is the business. The strategy behind shipping the UPII core as MIT-licensed open source on PyPI. DF-006 — From UPII to UEII: Scaling Personal Memory Guarantees to the Enterprise https://datafrontier.co/blog/from-upii-to-ueii https://datafrontier.co/blog/from-upii-to-ueii 24 Mar 2026 00:00:00 GMT The same substrate, engineered for organisations. Introducing the Smritium Membrane — PII detection, egress audit, and compliance by architecture. DF-005 — What Never Leaves Cannot Leak: Local-First as a Security Model https://datafrontier.co/blog/what-never-leaves-cannot-leak https://datafrontier.co/blog/what-never-leaves-cannot-leak 10 Mar 2026 00:00:00 GMT Most privacy is a policy you are asked to trust. Local-first is a property you can verify. DF-004 — Content-Addressed Memory: Why Deterministic Ingestion Matters https://datafrontier.co/blog/content-addressed-memory https://datafrontier.co/blog/content-addressed-memory 25 Feb 2026 00:00:00 GMT Re-ingest the same corpus and reproduce 100% of chunk hashes. Why identity should be a pure function of content. DF-003 — Cite or Abstain: Designing Retrieval That Refuses to Hallucinate https://datafrontier.co/blog/cite-or-abstain https://datafrontier.co/blog/cite-or-abstain 11 Feb 2026 00:00:00 GMT A system that always answers is a system that sometimes lies. Why abstention is a feature. DF-002 — Three-Signal Retrieval Fusion: Semantic, Temporal, Relational https://datafrontier.co/blog/three-signal-retrieval-fusion https://datafrontier.co/blog/three-signal-retrieval-fusion 28 Jan 2026 00:00:00 GMT Recall is a sensor-fusion problem, not a cosine distance. How fusion reaches Recall@10 = 0.958. DF-001 — The Memory Problem: Why AI's Hardest Bottleneck Isn't Intelligence https://datafrontier.co/blog/the-memory-problem https://datafrontier.co/blog/the-memory-problem 14 Jan 2026 00:00:00 GMT Models are no longer the hard part. The hardest problem now is memory.