Introducing TigerFS - a filesystem backed by PostgreSQL, and a filesystem interface to PostgreSQL. Idea is simple: Agents don't need fancy APIs or SDKs, they love the file system. ls, cat, find, grep. Pipelined UNIX tools. So let’s make files transactional and concurrent by backing them with a real database. There are two ways to use it: File-first: Write markdown, organize into directories. Writes are atomic, everything is auto-versioned. Any tool that works with files -- Claude Code, Cursor, grep, emacs -- just works. Multi-agent task coordination is just mv'ing files between todo/doing/done directories. Data-first: Mount any Postgres database and explore it with Unix tools. For large databases, chain filters into paths that push down to SQL: .by/customer_id/123/.order/created_at/.last/10/.export/json. Bulk import/export, no SQL needed, and ships with Claude Code skills. Every file is a real PostgreSQL row. Multiple agents and humans read and write concurrently with full ACID guarantees. The filesystem /is/ the API. Mounts via FUSE on Linux and NFS on macOS, no extra dependencies. Point it at an existing Postgres database, or spin up a free one on Tiger Cloud or Ghost. I built this mostly for agent workflows, but curious what else people would use it for. It's early but the core is solid. Feedback welcome. tigerfs.io

When you've known your cofounder for >25 years.

Yes. Just use @PostgreSQL for everything! amazingcto.com/postgres-for-…

I just published a deep technical dive on Fluid Storage, our new disaggregated block storage layer for Postgres. It comes from years of running one of the largest managed Postgres fleets in the world (outside of the hyperscalers), serving thousands of customers and seeing the good, bad, and ugly of EBS. Fluid Storage is our rethink of what Postgres storage should look like in the cloud era: high performance, elastic scaling, fast forks and zero-copy snapshots, and resource efficiency with thin provisioning. Built to serve the modern needs of developers and applications, and the emerging needs of agents. Read about the architecture 👇

Today's a big day. We’re announcing Agentic Postgres, and kicking off two weeks of launches around it. 🚀 This one's special for me -- a big step toward what we think the future of data infrastructure looks like in the age of AI. Agentic Postgres is the first database built for agents. ▪️ Fluid Storage, a forkable storage layer, that lets every agent spin up its own database in seconds. And then iterate or branch instantly with fast, zero-copy snapshots. ▪️ A new Postgres MCP server that can reason and plan. Your expert DBA that helps you design, tune, and optimize production-ready apps. ▪️ Native search and retrieval built directly into the database. Hybrid keyword and vector search for applications and memory, built for real-time use. ▪️Plus a CLI and free tier, so anyone (human or agent) can spin up and start experimenting immediately. Some of these systems like Fluid Storage have been 18 months in the making. Others are smaller but no less important. Together, they redefine what it means to build with Postgres in an agentic world. For the first time, agents can build, test, and evolve applications on their own. Safely, instantly, and inside Postgres. It's a shift in computing: from databases for developers to databases for agents and developers together. I couldn't be prouder of this team. Try it now in Tiger Cloud. And let your agents cook. #AgenticPostgres #AI #Postgres #Databases #Developers #LaunchWeek @TigerDatabase @TimescaleDB

Universities with the highest net CS PhD stipends in the U.S., accounting for cost of living. @Princeton #1. 🏆 csstipendrankings.org/

Big news in the world of databases: @databricks is acquiring @Neon. The connection makes sense: Neon’s approach to serverless Postgres has really clicked with a new class of agentic and short-lived workloads, especially during development. Huge congrats to the Neon team (including some former Timescalers!) on what they’ve built. It’s also a clear signal of Postgres’ continued rise as the default database for developers – open, powerful, and now increasingly ubiquitous. But it raises a bigger question: why Postgres and Databricks, the Lakehouse + AI company? If you’re building a customer-facing application today, you don’t only need a great operational database. You also need to make your data (carefully!) accessible across your org, so solving governance and data management is key. So rather than the old OLTP vs OLAP split, I think the more useful distinction is operational databases that power customer-facing applications, and lakehouses that power internal analytics, BI, and machine learning. Startups and enterprises alike need both. Reading the Hacker News thread, one thing that stood out was the developer perception of Databricks itself as a kind of walled garden, especially given its roots in Spark. Which makes the contrast even more striking: the lakehouse is now heading in the opposite direction. We’re seeing a real shift toward open, modular lakehouse architectures. Open formats like Apache Iceberg are emerging as the standard for disaggregated storage, letting developers store massive volumes of data in S3 (or any object store) without lock-in. Separate compute engines (like DuckDB, Athena, Trino, and even Postgres extensions) are emerging that can independently read and analyze this data. And now open catalogs (S3Tables, Apache Polaris, XTable, etc) are simplifying governance. Just as Postgres has become the modern open standard for operational data – including with its plug-and-play extension architecture – are we seeing the emergence of open, plug-and-play architecture for the lakehouse as well?| Postgres + Open Lakehouse. Feels like a modern data architecture worth betting on.

🥳 Today’s the day! TimescaleDB 2.0 RC has arrived 🚀 Multi-node. Petabyte-scale. Relational. Purpose-built for time-series. 💯 free. Get started and learn what's new👇 tsdb.co/get-timescale-2-0 1/

Random weekend musing: The longer I spend in industry (startup land), the more I want my academic work to tackle more big picture / visionary problems. But peer review and top conferences highly optimize for thorough, quantifiable results (esp. performance in systems).

Fun internal engineering presentation at @TigerDatabase yesterday from TJ, the engineer leading pg_textsearch. pg_textsearch is a new Postgres extension that brings BM25 keyword search natively to Postgres -- built from the ground up using Postgres' index access method interface. You can combine it with vector search (pgvector/pgvectorscale) for hybrid retrieval, a powerful mix of semantic and keyword search and perfect for AI applications. As simple as creating a Postgres index: 💲CREATE INDEX articles_idx ON articles USING bm25(content) WITH (text_config='english'); Planned roadmap: ⟡ Released (last month): Full SQL interface with in-memory (memtable-only) storage ⟡ Nov: Out-of-core scalability: memtable spills to hierarchical disk segments (blocking spill) ⟡ Dec/Jan: Compressed disk segments, skip lists, and optimized query evaluation (block MAX WAND)

Given the continued risk and spread of #coronavirus, @PrincetonCS has proactively cancelled its in-person PhD Visit Day (Mar 26-27), and will instead hold a fully virtual event. We hope other schools will broadly follow suit for the health and well-being of our community.

Database team on fire! Who said you gotta choose between either fast point queries (indexes) or fast aggregations (vectorized columnar scans)? @TimescaleDB now gives you both. 2.18 will ship with a new table access method for our hypercore storage engine. 🔥🔥🔥

Sunday morning musing for #academics: what do you wish somebody told you / you knew when just starting as a professor.

Timescale is now TigerData. This is an exciting day for us. Over the past 8 years, we’ve grown to power a large range of modern, operational workloads. Beyond time-series and sensor data, we’re powering developer tools, SaaS applications, AI-native games, RAG applications, and more. Companies are running entire applications on us. We’ve quietly evolved from a time-series database into the modern PostgreSQL for today’s and tomorrow’s computing, built for performance, scale, and the agentic future. Super proud today that TimescaleDB is everywhere. Both in >3 million databases running in the wild, to being included in managed PostgreSQL offerings around the world: Aiven, Alibaba Cloud, Azure, Databricks Neon, DigitalOcean, Flyio, Huawei Cloud, Linode, OVHCloud, Render, Snowflake CrunchyBridge, Supabase, Vultr, and more. But the majority of our 2000+ customers today aren’t just “time-series applications,” but cover a range of transactional, analytical, and agentic workloads. So we’re changing our name: from Timescale to TigerData. To reflect who we’ve already become. This name change is not a rebrand, but a realization of who we are today, and how we’ll continue to grow. The fastest Postgres. The operational database platform built for transactional, analytical, and agentic workloads. Speed without Sacrifice. And wearing my product and engineering hat, I’m extremely excited and proud about a number of product announcements we’ll be making over the next few months: - A more agentic PostgreSQL - A deeper integration with the Lakehouse via Iceberg (as part of the BaseLake architecture I wrote about last week) - A new disaggregated storage architecture with zero-copy instant forks and replicas. Greater performance, super flexible, deep engineering. Supporting the needs of both production at scale, and our agentic future. Let’s go! 🚀🚀🚀 @TigerDatabase @TimescaleDB

Scaling Postgres with @TigerDatabase: This week, we're ingesting over 3.5 trillion metrics per day into a standard database service on Tiger Cloud -- the very same infrastructure our customers use. All to power query analytics (Insights): so you can identify, understand, and diagnose query performance with detailed visibility into planning and execution times, shared buffer hit/miss rates, cache ratios, and more. ⚡️⚡️⚡️

TIL that @Apple iMessage uses our Key Transparency work to ensure user security and privacy: E2E encryption on by default with users confident in others' identities. Unexpected academic impact can be amazing. CONIKS published @usenix Security 2015. security.apple.com/blog/imes… 🔥

1/ 💥Big news! Announcing the new Timescale Cloud, a database cloud for relational and time-series workloads, built on @PostgreSQL, architected around our vision of the "transparent box" 🐯🚀 A thread on the future of database services in the cloud 👇 tsdb.co/timescale-cloud-visi…

“Fools ignore complexity. Pragmatists suffer it. Some can avoid it. Geniuses remove it.” - Alan Perlis The new semester just started at @Princeton, and I'm teaching distributed systems again. What I tell my students in the first lecture: A central goal of software systems is to hide complexity. Do the heavy lifting, so the application developer doesn't have to. Tonight I went down the rabbit hole a bit in famous computer scientist quotations about simplicity. What a great one from Alan Perlis. First Turing Laureate in 1966.

Huge (and well deserved) congrats to @jrexnet for her appointment to the National Academy of Science! @Princeton #BragForFriends

I have stopped reviewing any non open access journal. I wonder if it's time for academics to boycott serving on any program committees whose proceedings are not open access. It's clear @TheOfficialACM isn't listening to *their* key stakeholders. cc @ACMSIGOPS @ACMSIGCOMM

Reviewer #2 strikes again! "Outdated Software: Python 2 is legacy software which will be out of support on January 1, 2020. I wonder why the implementation doesn't use Python 3." #AcademicTwitter

Princeton's endowment this year grew 46.9% (over $10B) to $37.7B. It's close to $4.6M per student 🤯 Tuition now just 6% of operating expenses. => Charging ANY tuition (and continued rate of tuition increase) is purely a choice, not necessity🤔 dailyprincetonian.com/articl…

Big shout-out to Prof. @WyattLloyd. @Princeton 's trustees officially approved his tenure! 🎉🎉🎉

Very grateful to spend some time today with Bob Kahn (@Princeton PhD) after he received the James Madison Award. Big honor to hold his endowed chair, and fascinating to hear stories of the Internet's early development.

It's a new semester! Here are the latest 150 @Princeton students super excited to learn all about linearizability, consensus, transactional semantics and other distributed system arcana!

Where I talk about decentralization maximalism, my p2p scars from 15 yrs ago, and @BlockstackOrg power of choice. blockstack.org/blog/blocksta…

Humbling honor, but would like to especially thank @dmazieres for his support and mentorship as former PhD advisor and @PrincetonCS for being a home over the past decade.

I've always wanted to be the founder & "lead blockchain" of what could only be the most amazing cryptocurrency out there! #ICOScam #AchievementUnlocked I've made it! @el33th4xor

Congrats to @Princeton's Dr. Marcela Melara (@mas0mel), PhD student #12, for defending her thesis today on "Intra-Process Least Privilege and Isolation for Emerging Applications" On to @Intel Labs for systems security research! #upupaway

Congrats to @Princeton's Dr. Amy Tai, PhD student #10, for defending her thesis today on "Leveraging Distributed Storage Redundancy in Datacenters." Now at @TimescaleDB! #upupaway

Very excited to share our design, plans, and benchmarks for the distributed version of @TimescaleDB (now in private beta). #tsdb #timeseries #PostgreSQL 🐯🐯🐯 blog.timescale.com/blog/buil…

WAAAAAT!?! @jeanqasaur ZoomBachelorette is on the front page of CNN.

Just saw amazing demo of in-progress @TimescaleDB feature: hierarchical continuous aggregates (incremental materialized views). What's 🤯 is real-time aggregates compose. Weekly aggregates roll up daily, hourly, minutely, raw, transparently. With full backfill support. 🔥🔥🔥

⚠️Warning: The latest Postgres minor releases (17.1, 16.5, 15.9, 14.14, 13.17, 12.21), released today, have an unexpected breaking ABI change that will crash existing deployments of TimescaleDB (and other extensions), unless used with a TimescaleDB binary explicitly built against those new minor PG versions. If you are using TimescaleDB, we recommend NOT upgrading to these latest minor versions at this time. We have paused our normal process of automatically upgrading customers of our managed clouds to those minor releases, as well as paused our automated community/OSS builds, as we work with the PG community how to address in upcoming versions. Thanks for your understanding! 🙏

Last week, our 2005 paper "Keyword Search and Oblivious Pseudorandom Functions" was awarded the 2021 Theory of Crypto Conference "Test of Time" award w/ @bennypinkas, Yuval Ishai, Omer Reingold cs.princeton.edu/~mfreed/doc… 🧵

Amazing part of startup founder community is how folks pay it forward to strangers. Academia doesn't feel the same. Senior folks have crazy inundation of rec letters (tenure, awards), but that's different. Let's start! Junior folks - I'm great at unsolicited advice. Ping me.

Intrigued by @zeddotdev's announcement of DeltaDB, which mentions using CRDTs to synchronize code changes. The details are thin, but it touches on a deep and fascinating frontier. For decades we’ve searched for better ways to manage concurrency. — Databases solved it with transactions and locks. — Collaboration tools explored Operational Transformation (OT): operation-based rewrites to ensure edits commute. — CRDTs pushed further: data-structure-based objects that can be merged in any order and still converge. Beautiful in theory, but code is the hard case. At the data level, CRDT merges are consistent. At the code level, the result is often nonsense: programs that don’t compile (syntax) or don’t make sense (semantics). There be dragons here. 🐉 This is why git and other version-control systems rely on human-driven conflict resolution, not CRDTs. But here’s the twist: with LLM coding agents, the calculus can change. Imagine a CRDT merge feeding into an agent loop that compiles, runs tests, reasons about correctness, and repairs what’s broken. I don’t know if that’s Zed’s direction, but it’s what the announcement made me think of. A reminder that as agents mature, ideas once seen as "beautiful but impractical" may suddenly find their moment. Exciting times ahead. ✨

Oft overlooked in crypto/web3 world: it's not solely about decentralization, it's about flexibility and choice. Good architectural layers future-proof innovation. Higher layers can evolve independently. A core principle in our @Stacks 2016 publication: cs.princeton.edu/~mfreed/doc…

Timescale’s hybrid columnar compression scales #PostgreSQL performance 20x less space, faster queries, lower costs Deep dive today unveils this journey, both optimizing query performance and improving developer UX (fully mutable compressed tables) timescale.com/blog/building-… 🧵

👀New feature in @TimescaleDB 2.9: Hierarchical continuous aggregates. Define incrementally materialized views on top of each other, where real-time queries transparently traverse down the hierarchy automatically to always get up-to-date answers. 🔥

Sometimes your research pops up in cool places years later: Looks like @deepseek_ai's distributed file system uses our "Chain Replication with Apportioned Queries" for strong consistency.

Why did the @Princeton historical house cross the street? ... 🪄

🌟🌟 @TimescaleDB hits 10,000 @github stars 🌟🌟 github.com/timescale/timesca…

Congrats to now Dr. @muneeb for successfully defending his PhD thesis at @Princeton today. Best of luck @BlockstackOrg!

🔥 Timescale Analytics: Time-Series Analytics for Postgres ⚡️ ✅ Specialized functions to make SQL the best tool for time-series analysis ✅ Developed in the open, with & for the community ✅ Built in Rust Get project details & how to contribute 👇 tsdb.co/intro-timescale-anal… 1/

💥Bottomless, low-cost object storage built on Amazon S3, now in @TimescaleDB Cloud. Scale infinitely. Pay only for what you use. Tier data across disk and S3 transparently. Query in SQL as if one continuous #PostgreSQL table. More in my post: tsdb.co/s3-tiering

Super excited to announce native compression capabilities in the latest release of @TimescaleDB, available today. Best-in-class compression algorithms with 91-96% space savings. All credit to super talented team! 🐯🐯🐯 Read more: blog.timescale.com/blog/buil…

After two years of dedicated engineering, really proud to announce that the final version of @TimescaleDB 1.0 is available today! #PostgreSQL #onwardupward #zoomzoom blog.timescale.com/1-0-enter…

For cloud services like databases, platform/fleet level status metrics are a very incomplete proxy for what users actually care about -- although we certainly pride ourselves at very high uptime here as well. What users actually care about is: Is _my_ database available/performant, not is the overall _platform_ having trouble. Because those things aren't the same. If an AWS server has a hardware failure, and the system automatically detects and recovers quickly, the "system status" of the platform is still all green, even if a user may see some disruption. So you end up building a lot of defense in depth, and various levels of mechanism to drive RTO as low as possible, within what a customer is willing to pay for (e.g., non-replicated vs. replicated databases). In fact, you want your platform to separate platform and database-level availability and engineer them accordingly. At @TigerDatabase, even if our "control plane" goes completely offline -- which it never has over 5 years of operation -- our customer databases continue to operate unscathed, serving critical customer traffic. This type of "graceful degradation" is something that customers want and need, and is the actual sign of mature production platforms. This is something that the hyperscalers like AWS are really good at. I'm not at all saying that PlanetScale isn't an operationally-mature platform, but having database companies compare blinking green lights on status pages is just not a good reflection of the hard engineering and careful design that could and should go into these platforms.

In March, we migrated our entire fleet of customer @PostgreSQL databases from ext4 to xfs. At scale and with zero downtime.🪄 Behind this and other operational improvements @TimescaleDB Cloud's new custom #kubernetes operator, PatroniSets. Deep dive: tsdb.co/replacingstatefulset…

Hi @samlambert - I honestly don't understand your interpretation of that paragraph, or why you seem so angry about it? Here's what I wrote: - We considered an architecture using local NVMe. - An example of such a database system using local NVMe is PlanetScale Metal. - Our concern is that to achieve durability with local-NVMe-backed databases, you need to always run Postgres with one or more HA replicas. - This gives durability, but from our own experience, there are many customers who don't want to pay the multiple for HA replicas. Your response:  We always run PlanetScale Metal with 3 nodes. Those two things seem very much in agreement?

Who wants to work on distributed databases? DM me...

Huge congrats to @jrexnet, my amazing colleague at @Princeton who's always served as an academic mentor, has been named the next Provost of the University. Wonderful choice of a leader! 🐯🐯🐯 princeton.edu/news/2022/11/2…

Newest capability for @TimescaleDB developer insights is backed by ... a standard Timescale service. Yes, we build product using the same product as our customers. — 100 billion recs, billions/day 🔥 — 90TB compressed to 4TB 🔥 — Data tiering across disk and S3 🔥 #production

Special thanks to all my amazing students and colleagues over the years!

Have really enjoyed working with the top-notch @blockstack engineering team to help design a new decentralized architecture for Internet applications. Read the just published 2.0 whitepaper here:

Huge congrats to @Princeton 's own @margmartonosi for taking the reins of shaping the future of computer science research as the next head of @NSF_CISE!

Does evaluating, debugging, testing distributed databases excite you? Do you like #Jepsen and #consistency models? Working w super smart but humble teams? @TimescaleDB hiring first engineer for building in-house testing infrastructure. Remote first. timescale.com/careers/454024…

Is @TimescaleDB bringing @ApacheArrow into its @PostgreSQL optimizations for further supercharging query performance? Coming soon: Faster in-memory decompression for columnar operations.

10/10 would recommend...

That feeling when the team delivers your most requested feature on @TimescaleDB Github... Fully mutable compression (eg, upserts, updates, deletes) ➡️ More scenarios it can be used in operational environments ➡️Faster query performance & greater cost savings ⚡️⚡️⚡️

Kicking off AI Launch Week at @TimescaleDB today. Announcement of Timescale Vector made front page of Hacker News, with some great technical exchanges like the following from @cevianNY... 🔥

Got an LLM? @TimescaleDB now has native vector DB support to serve as long-term model memory or provide similarity search for other vector embeddings. One click to fully managed , good from dev to production, starting at $20/month.

$0. It's long-game lead-gen for the nasdaq. They cold reach out with photos -- great reply rate on their outbound. Worked for us!

Observation: Many engineers are opting for 2015 MacBook Pros rather than 2017 models because of the hated touchbar. Wonder if #Apple cares about pretty strong product feedback. #SaveTheEscapeKey

Yes, it's Timescale AI week. But we sneaked in another little partnership announcement about how to connect @Cloudflare Workers to @TimescaleDB using Cloudflare HyperDrive. ⚡⚡⚡ developers.cloudflare.com/hy…

💥 @popsql is joining @TimescaleDB. Even before we met the team, we were impressed by the amazing product they built: A truly modern SQL editor, collaboration, and visualization tool that made working with data both easy and powerful for developers and data teams. And in getting to know @rahilsondhi and seeing how we share a vision of building the best #PostgreSQL developer experience for the cloud era, I'm even more thrilled for the future. With this announcement, we're also launching an integrated PopSQL experience for all Timescale users, available immediately. tsdb.co/popsql-ts

Big news today at @TimescaleDB – and the kickoff to an exciting *month* full of launches. It's been an exciting journey with @acoustik and the entire Timescale team – now spread across 6 continents! – and the best is yet to come. Measure everything that matters! 🐯🚀

🐯🐯🐯🦄🚀💥

The power of combining vector search with SQL-based filtering. It's almost like those logicians in the 70s knew something about the power of relational algebras.

I still believe a good-faith reading of my original text makes the intent clear: it discusses the engineering tradeoffs inherent to the general approach. That said, my goal isn't to create any misperception, so I've updated the text to make it clearer.

City that never sleeps. I ♥️ NY.

🚀 Promscale - our open-source analytical platform & long-term data store for @PrometheusIO - just got some 💯 new features: ✅ Support for multi-node @TimescaleDB ✅ Performance tuning ✅ Data migration tool ✅ JSON push tsdb.co/promscale 1/

"Exactly what we need. It's like you read our minds." ☝️Customer feedback on this new major @TimescaleDB feature we're working on. 100% of customers interviewed have joined the private beta. The future is 😎

New in-depth post where I compare @TimescaleDB and #InfluxDB and how they are purpose built differently for time-series data blog.timescale.com/timescale… #PostgreSQL #timeseries #SQL #NoSQL #tsbs

Pro-tip for junior CS colleagues on academic job market: Talk to chair / engineering dean during second visit. If they start asking about journal publication plans, RUN.

Announcing pgvectorscale: faster and cheaper than Pinecone, 100% open source! We just released two new open source extensions (PostgreSQL licensed) to make Postgres the best AI database. Let's make Postgres a better database for AI, together!

We're building a new product growth team here at @TimescaleDB! 📈 If you are equally passionate about both the tech and business side of SaaS, come join as our first growth PM, working closely with me & senior product leadership. DMs open. 🐯🦄🚀 timescale.com/careers/611025…

I’m hiring a PM who will report directly to me (CTO/cofounder) at @TigerDatabase. This is a founder’s PM role: ⚡️ Entrepreneurial, fast-moving 🤖 Deep in AI + infra 🤝 Works directly with partners 🔥 Builds, ships, and makes things happen Remote co, but I want you in SF. Help define the future of Postgres + AI-native infra: tigerdata.com/careers/542068…

Sneak peak at some performance improvements coming to @TimescaleDB using vectorized aggregates on its columnar storage. Some 30x SELECT and 10x DELETE improvements compared to latest 2.16.1. 😋😋😋

And as somebody mentioned: forcing academics to pay the @TheOfficialACM something like $2000 for open access doesn't count. That's revenue replacement, not costs. And if anything is firmly against broadening participation and inclusion in CS, one of ACM's stated goals.

WhatsApp launched "Key Transparency" today, which uses a crypto-verifiable log to strengthen/simplify PK security. Exciting - this approach was first introduced by our CONIKS work in 2014. Congrats @mas0mel @AaronBlankstein @josephbonneau @EdFelten engineering.fb.com/2023/04/1…

Academia has got to get serious about reference checks for culture and behavior. We don't do it nearly enough. And especially given the rate at which we hire (glacial), there is little excuse.

(6) Don't push your way in to get your name on papers where you don't contribute, don't demand that all your students' papers list you as co-author for same. Beyond the ethics, the practical: Word will get out and it'll call your other contributions into question.

(4) PhD students pay much larger opportunity cost going to grad school than you do advising/funding them. Consider that. Path through grad school / what makes a successful thesis can vary based on their long-term plans (eg, academia, industry). Think of those goals throughout.

We started with ‘boring is awesome.’ Now we’ve leveled up to ‘deeply unoriginal.’ Consistency matters. @jjacky quoting TJ Green, PhD at @DeepLearningAI NYC conference today.

When one of your database engineering teams is quietly implementing all these cool probabilistic data structures in #rustlang 🦀 for even MORE efficient time-series analysis in @TimescaleDB & @PostgreSQL ... 🔥🔥🔥

(3) Write grants on research you plan to do /already starting. Don't let others' large grant applications direct your focus (waste of time and tail wagging dog). Most of time (CS) it's same amount anyway: $500K for one person, $1M for two, or $5M split 10 ways, it's all the same.

Super proud of what the team launched today for production use. Bottomless storage for @PostgreSQL built on Amazon S3 for transparent data tiering. Never run out of space. Pay for only what you use. 🚀🚀🚀

What a week! 🐯🚀 With Agentic Postgres, we introduced a new foundation for how databases work in the age of AI: forkable databases, reasoning MCP servers, and native search and retrieval, all built directly into Postgres. The part I'm most excited about is Fluid Storage, the technology that makes it all possible. It's built on a transactional distributed block store, lineage-aware storage proxies, and a user-space block device driver, with versioned copy-on-write and application-consistent snapshots. Together, these give us fast, zero-copy forks on real production data, enabling entirely new developer and agentic workflows. Next week I'll be publishing a deep dive into the architecture and design choices behind it: why we built it this way, and what it unlocks for agentic systems. Until then, the full Agentic Postgres stack is live: — Fluid Storage for forkable databases — Reasoning MCP Server that understands Postgres — Native BM25 + vector search for retrieval and memory — A great CLI to live in your terminal — A Free Tier to explore it all Agents are the new developers, and they need infrastructure that matches how they work. This week, we started building that future. $ curl -fsSL cli.tigerdata.com | sh Let's go!✨

Where I talk with @OReillyMedia about building @timescaledb: time-series DB on #PostgreSQL. Full SQL at 100B+ rows. oreilly.com/ideas/a-scalable…

🚀 @TimescaleDB 2.23 is out! 14 feature improvements + 16 bug fixes, including: 🔹 Full PostgreSQL 18 support 🔹 UUIDv7 compression now default (~2× faster, 30% smaller) 🔹 Hypertables can be set to UNLOGGED 🔹 Continuous aggregates now support set-returning functions Rolling out to Cloud in the coming weeks. OSS users can grab it now from GitHub or our pre-built binaries.

Very honored Ethane won @ACMSIGCOMM Test-of-Time Award for "ushering in age of SDN" @martin_casado @nickmckeown1 sigcomm.org/awards/test-of-t…

I personally don't love any of the more conceptual books in this space, but here's the course I sometimes teach in this area with slides, which you might find useful: cs.princeton.edu/courses/arc…

Putting native compression in @TimescaleDB through its paces on ARM @Raspberry_Pi before v1.5 release. Especially now with 95% storage savings, they're good DBs brent. 14/10

🐯🔥 NEW: Timescale Forge 🔥🐯 Excited to unveil this public preview of the easiest way to get fully-managed @TimescaleDB. 100% free for 30 days, plans starting $49/month. tsdb.co/forge-preview Independently scale compute & storage, instantly pause & resume, more coming!

Benchmarks in...🤯🤯🤯 ➡️ Promscale on @TimescaleDB Cloud up to 94% cheaper than other managed Prometheus Offerings! Full @PrometheusIO, @JaegerTracing, @opentelemetry support for metrics & traces in scalable platform. (Say hi at #KubeCon this week!) timescale.com/promscale

🧟‍♂️@TimescaleDB Vector docs are aliiiive! Simplify your operational stack with a single database for your business, analytical, and AI vector data. Built on #PostgreSQL, ready for production. With performance and reliability superior to many bespoke vector databases, which are still in their embryonic stages of development.

All these amazing new capabilities hiding behind production feature flags, just ready to come forth... 🐯🔥

Read the technical blog post here: tigerdata.com/blog/fluid-sto… 🚀

Hello, Timescale Insights. Your observability detective for your #Postgres service, at a level of granularity that is unmatched. 🔎📈 It's Cloud Launch Week here at @TimescaleDB! timescale.com/blog/database-…