Cloudflare has redesigned the control plane of its SQLite-based serverless database D1, splitting transactional metadata, analytics and globally replicated routing across Postgres, ClickHouse and Workers KV so that query processing stays at the edge and is decoupled from centralized management. The new architecture adds heartbeat reconciliation, idempotent retries and eventual consistency to auto-repair inconsistent states left by failed create or delete operations, supporting an active database base that grew 18x over the past year.
ClickHouse
Open-source columnar analytics database known for high-throughput real-time queries, often used as an agent data warehouse.
Curated Intelligence
Greg Isenberg lays out a practical playbook for becoming the most AI-native company: build a personal second brain in Obsidian, capture voice notes via Wispr and meeting transcripts via Granola, then feed that private context to agents through CLAUDE.md or AGENTS.md instruction files, cross-session memory tools like claude-mem and mem0, and an Airbyte-to-ClickHouse data pipeline. His core argument: models are identical for everyone, so proprietary context is the only defensible moat.
The author recounts four attempts at building real-time sales-receipt reports on Kafka and ClickHouse, showing that at-least-once delivery duplicates and late-arriving backlog break any aggregate defined as a running sum of inserts. The fix is to treat an aggregate as a function over its document set, so duplicates and late documents trigger recomputation instead of manual correction.