building https://xitter.cf/t.co/YQbF3jGv6Q (YC W23)
Berlin
Joined January 2022
- Tweets228
- Following541
- Followers633
- Likes160
Discussing right now how we will launch features over the next weeks. It's great to see how our marketing team understands engineers.
kicking off the 2nd all @ClickHouseDB offsite this yr in berlin! @langfuse team ofc also fully assembled.
Revamping our careers page, I noticed something I say on every hiring call that wasnt explicit enough.
At Langfuse, engineers write RFCs and customers sometimes review them. The bar here is to think through features by yourself and beeing able to communicate the idea and caveats clearly with customers to have a targeted discussion.
langfuse.com/careers
Max Deichmann retweeted
در وصف خفن بودن تکنیکال Langfuse قبلا نوشته بودم. مثل اینکه در زمینه مارکتینگ و تبلیغات هم دستی بر آتش داره، خیلی برلینیه این آگهیکار :)
[این شرکت دوتا شعبه داره، یکی توی برلین یکی توی سانفرانسیسکو، هاب مهندسیش برلینه]
Design and engineering at Langfuse now work from the same components.
Our design system lives in Storybook. A daily Claude routine syncs every component into Claude Design, along with its TypeScript types and a short usage note written for the model.
1/ Ben's been building our design system since May. 50+ components in Storybook.
But what actually gets coding agents to use it is lint rules. Lots of them.
2/ No className or style on components, only typed variants. Design-system code can't import feature code. Icons only from @langfuse/icons. Stories can't assert on CSS classes.
Some of the errors even tell the agent what to use instead.
Sequence diagrams are the future of RFCs. They are easy to create with LLMs and are the perfect abstraction level to discuss implementation detail.
Amazing UX quality and speed of this ship, @hassiebpakzad
It was a blast in India 🇮🇳 . The customers i met there ship great products at an incredible pace.
I love to present product demos over slides. It shows the progress the team made over the last months and sometimes I am even surprised by new changes made by the team during presentations.
One of the best parts of @ClickHouseDB Open House Bangalore was that the @langfuse session wasn’t just slides - it had a proper demo too.
The biggest takeaway for me:
AI failures don’t just show up in the output but they hide in the path.
Seeing the trace flow broken down step by step made that super clear:
retrieval, context, response generation, evals, monitoring : all visible instead of guesswork.
Really liked how @maxdeichmann showed the full loop:
trace → evaluate → improve → ship
And the live demo made it click even more - this is what reliable AI/agent infra actually looks like in practice. ⚡
@langfuse + @ClickHouseDB is a seriously strong combo for building and improving AI agents.
#ClickHouse #Langfuse #LLMOps #AIObservability #AI #OpenHouseBangalore
Just boarded the plane to Bengalore to meet customers at @ClickHouseDB openhouse conference. Anyone in town who i should be meeting?
I'm speaking at @ClickHouseDB Open House in APJ next month
sep 22 bangalore: workshop on observing and improving AI agents with langfuse
sep 24 singapore: talk on shipping reliable AI agents
say hi if you're around 👋
clickhouse.com/openhouse/ban…
clickhouse.com/openhouse/sin…
Max Deichmann retweeted
👊I've joined the crew at @ClickHouseDB to establish the *ClickHouse Labs* research group. We'll be working side-by-side with ClickHouse devs to cook up new ways to make the system even faster. I'll also be exploring new topics related to PostgreSQL. clickhouse.com/blog/andy-pav…
If you want to work on AI, open source, dev tools and want to sit on a rocket: join Langfuse engineering in Europe
This quoted post is unavailable.
I wanted this so long
any table is a chart.
toggle the observations table into a time series over the exact same query. same filters, same time range. p95 latency by model, count by environment, cost by name. one click adds it to a dashboard.
langfuse.com/changelog/2026-…
Max Deichmann retweeted
any table is a chart.
toggle the observations table into a time series over the exact same query. same filters, same time range. p95 latency by model, count by environment, cost by name. one click adds it to a dashboard.
langfuse.com/changelog/2026-…