@rawert
Account based inGermany
About this account
- Account based in
- Germany
- Connected via
- Web
Account-level information from X, not a live location or the device used for a specific post.
Berlin
Joined February 2013
- Tweets1.1K
- Following1.3K
- Followers1.1K
- Likes8.1K
Pinned Tweet
I'll be speaking at @ClickHouseDB user conference Open House in London at the end of September. Registrations are now open & I'm very much looking forward to being in 🇬🇧
- clickhouse.com/openhouse/lon…
Jesus christ
Some data we recently assembled on entrepreneurship/compute in Europe: eudata.vercel.app.
We hope that one of the useful roles that Stripe can play is in collecting and publishing empirical data pertaining to entrepreneurship and industry in Europe. There's growing appetite to get Europe on a better footing, and cross-sectional comparisons can often shine light on where opportunities lie. If you're interested in this kind of thing, we publish more at stripeeconomics.substack.com.
Clemens Rawert retweeted
how awesome would it be if every new table created, every column name change, every column TYPE change is replicated from Postgres to @ClickHouseDB ?
WalShadow does all of these, crazy to see users experience this magic. checkout: clickhouse.com/blog/introduc…
Clemens Rawert retweeted
Based on our experiments, Jev bears big potential for evals
✔️ It is up to 400x cheaper than frontier models, making full scoring of production traces feasible, no more sampling needed, even at huge scale
✔️ It allows for concurrent questions on the same state, where LLMs previously struggled with interaction effects and common wisdom was to not mix evals in the same requests. This makes Jev even cheaper.
✔️ The structure of Jev's output forces thinking in distinct categories during setup, improving the engagement with defining the exact evaluator categories
Play around with this demo here: langfuse.com/blog/2026-09-18…
Clemens Rawert retweeted
Finally got to record a 15 min overview of what @langfuse is based on an end-to-end example
- how it helps build great AI agents/apps
- how to continuously improve them with Lanfguse
- how the integration works
- why langfuse and no closed-source solutions
- how agents use it
@marcklingen sat down to demo all langfuse features in a single video.
Agent observability, monitoring, evals, prompt management, and what is special about Langfuse.
Great place to start if you are new, or catch up on latest changes.
langfuse.com/watch-demo
Fresh walkthrough of @langfuse
@marcklingen sat down to demo all langfuse features in a single video.
Agent observability, monitoring, evals, prompt management, and what is special about Langfuse.
Great place to start if you are new, or catch up on latest changes.
langfuse.com/watch-demo
On demand compute on @ClickHouseDB is neat. @peignonmel dives in on Oct 7: clickhouse.com/company/event…
Clemens Rawert retweeted
Why Jev is interesting for evals
> Every question is evaluated in parallel and in isolation against the same state. So adding a fourth question, or a fourteenth, barely changes response time, costs only the tokens of the question itself, and cannot degrade the answers to the others. You can ask speculatively and throw away what you do not need.
Fits very nicely to how good evals should work. Let's see what competition will do to this segment (make it even better, cheaper, faster).
🇮🇳 Bangalore - my co-founder @maxdeichmann is bringing @langfuse to India for the first time at @ClickHouseDB Open House conference. He's going to be there sunday-tuesday & would be keen to connect with folks locally, please tag here!
Just boarded the plane to Bengalore to meet customers at @ClickHouseDB openhouse conference. Anyone in town who i should be meeting?
Good read! Most often your users will leave implicit feedback signals (do they accept an edit, copy an output etc) that you can use to infer whether your app provides a good/bad answer.
Scalable, free, mostly unbiased
Clemens Rawert retweeted
Germans really get around. You'll be in the most isolated place hiking up a mountain most of the locals don't even know about and a German on their 5th vacation of the year will pop out of nowhere
Clemens Rawert retweeted
Props to @langfuse for their excellent docs and learning materials! 👏
If you’re trying to wrap your head around concepts like LLM observability, monitoring, and evaluations, these resources are well worth checking out:
🎓 Academy: langfuse.com/academy
🛠️ Workshop: langfuse.com/workshop
📚 Guides: langfuse.com/guides
Clemens Rawert retweeted
listen man sometimes you gotta put the manhattan project on hold to sell b2b SaaS
🇫🇷 Speaking at @aiDotEngineer Paris by @MistralAI next week!
talk slightly from left field: What Can Economic History Teach Us About the AI Moment?
industrial Revolution, printing press, railways, electricity, internet and so on!
Wednesday, 17:40 Main Stage
Clemens Rawert retweeted
@Waymo is preparing to launch in Berlin?! Hell yessss
same thing here this weekend, german version -- give check24 access, instinct gets car insurance. so bullish
My instinct just saved me 24% on car insurance:
- gave it access to Geico
- pulled my coverage
- shopped around for same
- sent me a ready-to-pay quote
Will save $1005 for the same coverage.
How can you not be BULLISH?? 🥹
Clemens Rawert retweeted
. Today, we’re taking Postgres replication to the next level.
With WalShadow, you can now replicate Postgres to ClickHouse directly from the physical write-ahead log (WAL), without the hassles of logical replication, and with sub-second replication latency.
This one is especially personal and honestly, a little bittersweet.
We benchmark WalShadow against @PeerDBInc, something @iskakaushik and I founded years ago. PeerDB eventually became the foundation of Postgres CDC for thousands of customers.
Seeing a benchmark showing up to 50x lower replication latency than PeerDB creates mixed emotions :)It also makes me incredibly proud. Because this is exactly what we always wanted: to keep pushing Postgres CDC/replication forward.
WalShadow takes a different approach. Instead of logical replication, it works directly with the physical WAL, getting us as close to Postgres as possible and enabling replication into ClickHouse in less than a second.
Some of our best engineers have spent months working through some very hard problems to make this possible, and I’m incredibly grateful to the team for getting us here.
The ambition is the same one we had: build something that thousands of customers can reliably depend on. And, make Postgres + ClickHouse feel less like two databases connected together and more like one deeply integrated stack.
Very excited to finally share WalShadow with the world. Give it a try we’d love your feedback! 🚀 clickhouse.com/cloud/postgre…
Introducing WalShadow: sub-second Postgres replication to ClickHouse from physical WAL.
No logical replication. WalShadow decodes the same physical WAL stream Postgres replicas use and writes ClickHouse-native blocks directly into ClickHouse. In our benchmarks: ~200 ms commit-to-visible latency and 289K rows/sec sustained, keeping pace with the source Postgres.
Open source and available today: buff.ly/wzD3ZKh Blog: buff.ly/piXeM2z
