TiDB, powered by PingCAP

Most agentic AI systems are not failing because of compute. They are failing because of memory. O'Reilly's latest report, produced with TiDB, breaks down exactly why memory architecture is the real bottleneck for autonomous AI systems, and what to do about it. Here is what the report covers: - Why memory, not compute, is the limiting factor for agentic AI - How to unify structured data, vectors, and history into a single retrieval layer - Trade-offs between distributed SQL, vector databases, and hybrid stacks - Architecture patterns for persistent, contextual AI workloads including RAG pipelines and long-term memory graphs - How to modernize your data foundation for intelligent, adaptive systems Built for data architects, AI infrastructure engineers, and technical leaders who are moving beyond simple RAG into production-grade agentic systems. If you are designing the data layer for your next AI system, this one belongs on your reading list. 👇 Download the free report: https://ow.ly/EF4w50YFEbp #AgenticAI #DistributedSQL #AIInfrastructure

Most engineering teams don't discover their database architecture is wrong at a planning meeting. They discover it during a traffic spike. Or when an AI feature ships and query latency triples. Or when the migration they've been putting off becomes urgent because the current system simply won't hold. By then, the cost — in time, in reliability, in credibility — is already real. TiDB SCaiLE Europe 2026 is a day designed for the teams who want to have that conversation before the incident forces it. Distributed SQL architecture, agentic AI data patterns, cloud-native design at scale covered by engineers who've already navigated it. Thursday, 4 June | Epicenter Stockholm Practitioners. Real case studies. No fluff. Register → https://ow.ly/HWpN50YFFwt

We are proud to partner with AGI House for the Agent Harness Build Day on April 18th! AGI House is the premier home for the builders and researchers pushing the boundaries of what AI can do. As the focus shifts from "cool demos" to "production reliability," TiDB is joining the community to solve the industry's biggest bottleneck: durable agent memory. We’re honored to support the AGI House ecosystem and can’t wait to see the breakthroughs that emerge from this focused day of building. When: Saturday, April 18th. Request to join the build: https://ow.ly/AeEF50YFAoE #AGIHouse x #TiDB #PingCAP #AIAgents #AgentHarness #Database #GenerativeAI

Flipkart runs 400M users. Their database became the bottleneck. 700+ MySQL clusters. Painful resharding cycles. Manual failovers during The Big Billion Days, when a few minutes of downtime means millions in lost GMV. They moved to TiDB. Here is what changed: - Horizontal scaling with no resharding, nodes added dynamically and live - Automatic failover across nodes and regions, no human intervention at peak - 25 nodes and 20 replication channels collapsed into one unified layer - Write latency under 100ms. Read latency under 20ms - Full MySQL compatibility, zero application rewrites Their Coin Manager now processes millions of SuperCoin transactions monthly without breaking a sweat, even at peak traffic. If your MySQL fleet is becoming the ceiling, not the foundation, this is worth your time. 👇 Read the full story: https://ow.ly/UHjj50YFAcu #DistributedSQL #MySQLMigration #CloudNative

Dify AI is the second most popular LLM tool on GitHub with 70,000+ stars. But their scaling challenge was not the model. It was the data layer. They were managing a massive sprawl of isolated database containers, nearly half a million, driven by multi-tenant AI workloads. Dify consolidated that into one unified system on TiDB Cloud Starter and reported: 80% reduction in infrastructure costs 90% decrease in operational overhead The big win: one storage layer for documents, vectors, chat history, and relational data, so teams can build RAG and AI workflows without stitching together separate systems. Read the case study: https://ow.ly/ibLE50YF7nR #TiDB #GenAI #RAG #VectorSearch #DistributedSQL #DatabaseEngineering

Accidentally dropping a table is one of those “everything stops” moments. In MySQL, recovery usually means backups + binlog replay and a lot of waiting. In TiDB, it can be a single SQL command. This walkthrough shows how FLASHBACK TABLE can restore a dropped table in under a second (schema + data intact), plus how to inspect historical versions with stale reads inside the GC window. Read: https://ow.ly/B8hq50YEyI4 #TiDB #DistributedSQL #MySQL #DatabaseReliability #SRE #DevOps

The future of AI isn't just about intelligence—it’s about persistence. As foundation models transition from chat interfaces to autonomous, self-evolving agents, the underlying data architecture must evolve. The "Agentic Era" requires a unified, radically scalable foundation for agent memory and runtime state. Join TiDB and MiniMax for an exclusive evening at SPIN San Francisco. We’ll be diving deep into the infrastructure powering the next generation of AI with: - Yeyi Yun (Co-founder & President, MiniMax) on scaling models to 150M+ users. - Ben Liang Yang (General Manager, TiDB) on building an economically sound foundation for agentic memory. Network with engineering executives and AI pioneers from organizations like LinkedIn, Visa, Affirm, Notion, and Stripe. When: Thursday, April 9 | 6:00 PM – 9:00 PM PT Where: SPIN San Francisco Register Now: https://ow.ly/A0Wz50YErIN Space is limited and subject to host approval. #AgenticAI #Database #TiDB #PingCAP #MiniMax #AIInfrastructure

Big news! TiDB is officially partnering with SCALE at GMI to power the next generation of AI and infrastructure startups. To fuel the journey, we’re providing $3,000 in database credits to every team in the program. High performance, distributed scale, and zero-effort operations—all are all at your fingertips. Welcome to the future of data! #TiDB #TechPartnership #SCALEbyGMI #ScaleUp #DataInfrastructure

AI apps are moving from demos to production — and the biggest challenge isn’t the model. It’s context. Agents need long-lived memory. RAG needs hybrid retrieval across vectors + metadata. Multi-agent systems need shared state. Most teams stitch together a vector DB, a relational DB, and a cache… and end up with duplication, sync gaps, and operational sprawl. On April 30, join us to learn what production AI apps actually need: a context platform — one operational data layer for vector search, relational queries, and real-time analytics. Building a Context Platform AI Applications Actually Need April 30 | 10 am PDT Register: https://ow.ly/Rg9C50YCpuu #TiDB #GenAI #AgenticAI #RAG #VectorSearch #DistributedSQL

Most observability setups fail not because teams chose the wrong tool — but because they made things too complex, too fast. In Episode 64 of Data in the Hallway, we sit down with Josh Lee, Open Source Developer Advocate at Altinity, to dig into what it actually takes to build effective data stores for observability at scale. Josh brings hard-won, practical perspective: how to think about architectural requirements, where teams consistently go wrong, and why starting simple is almost always the right call before you're buried in telemetry data you can't act on. In this episode: - What "good" observability data store architecture actually looks like - The most common traps teams fall into when scaling observability - Practical tips for getting started without overcomplicating your stack If your team is dealing with growing observability data and feels like the infrastructure is working against you — this one's worth 30 minutes. 🎧 Listen to the complete podcast: https://ow.ly/NHUZ50YC9nE #Observability #DistributedSQL #TiDB #OpenSource #DataEngineering #SRE