Ivan Burmistrov

ClickStack: A High-Performance OSS Observability Stack on ClickHouse We're delighted to announce ClickStack: the open-source observability stack built on ClickHouse - logs, metrics, traces, and session replay in one blazing-fast, developer-friendly platform.

Wow, this looks quite great. And this announcement has a link to my post about wide events 🤯
clickhouse.com/blog/clickst...

From this description, it looks amazing, honestly. Obviously, the devil can be in details... so some real experience is needed, but it definitely feels as the right product. Also happy for ClickHouse to look into this direction, it's such a natural use case for them.

The core part of AGI will be a single unsharded Postgres, too

GitHub - ezyang/scubaduck Contribute to ezyang/scubaduck development by creating an account on GitHub.

A famous Meta's Scuba UI reimolenented from Scratch with DuckDB as a backend!
github.com/ezyang/scuba...
Scuba is a really awesome internal system at Meta. It's UI is simple but intuitive and powerful, I wrote about it some time ago: isburmistrov.substack.com/p/all-you-ne...

The Harsh Reality of Building a Real-time ML Feature Platform Ivan Burmistrov shares how ShareChat built their own Real-time Feature Platform serving more than 1 billion features per second, and how they managed to make it cost efficient.

My talk on QCon London 2024 has been opened to everyone. I discussed the challenges of building a realtime feature engineering platform.
www.infoq.com/presentation...

The Harsh Reality of Building a Real-time ML Feature Platform Ivan Burmistrov shares how ShareChat built their own Real-time Feature Platform serving more than 1 billion features per second, and how they managed to make it cost efficient.

My talk on QCon London 2024 has been opened to everyone. I discussed the challenges of building a realtime feature engineering platform.
www.infoq.com/presentation...

My @scylladb.com cluster.

@cynthiadunlop.bsky.social it's growing!

Thanks for the great talk!

David is @whoisdavid.bsky.social :)

Andreas Saudemont, @sarna.dev, and @isburmistrov.bsky.social have a long track record of achieving AI/ML performance at scale. There is still time to join and learn from them at Monster Scale Summit on March 11-12. RSVP for free here > www.scylladb.com/monster-scal...

#ScyllaDB #techtips

I mean, in general, being sick is not advisable

The Scylla team’s conferences—P99 CONF and Monster Scale Summit— are top-tier. The hardest part is choosing a track when every session looks great. And the swag is in its own league :)

"Using OTel to instrument your c̶o̶d̶e̶ path to promotion"

That's actually a pretty good analogy. And bicycle is healthier, btw 😏

Yep, this is fair

I mean, they also discussed hunting and how Zuck's daughters learn that sometimes you have to kill animals 😆 For me, it felt more weird than this masculine energy thing, for instance. Just saying this part of the interview didn't feel like a super serious discussion, more like random chatting.

Did you watch the interview? This masculine energy thing is being mocked, but my feeling is it was just two ppl discussing martial arts and some thoughts around that. It was a very abstract comment, IMO, not "I'm going to improve this at my company" kind of thing.

I don't have problems with technical topics, but I do have them when I want to explain some events / controversies circulating in the news. E.g.: UK grooming gang scandal, Meta moderation changes,...
I ask AI to summarise these news, clarify different positions, etc. It does it pretty nicely!

"So, we’re going to continue to focus these systems on tackling illegal and high-severity violations, like terrorism, child sexual exploitation, drugs, fraud and scams" - this?

It doesn't seem like a complete list, I see these as examples.

Where has this list been published?

Where is the information about harmful content? It's not like content moderation is no longer happening, as I understand it, so these things should be considered as serious stuff, I suppose?

Battle human vs. AI happened today. My 4 y.o. daughter wanted to re-watch a Peppa Pig episode where kids did an animation. I asked ChatGPT and Perplexity, and they both failed. Then I started showing my daughter random Peppa Pig episode titles, and she recognised the right one! Score 1 for humans 🏆

HNY to the team of HNY!

In private conversations :)
Also, "plex" / "plexing" seems to be a known word on Twitter, see e.g. x.com/annbordetsky...

Naming matters. Started hearing the phrase "let me plex it" (ask Perplexity). There is no similar sticky alternative with ChatGPT

To understand how software engineering will change when anyone (technical or not) can hire an AI agent to spit out code:

How did filmmaking and the movie industry change now that everyone has a professional-camera (previously unattainable to non-pros) in their pocket?

Opened Twitter. The whole feed is celebrating the achievement of AGI. Good that don't see it here, AGI-free feed feels refreshing.

Yep exactly, would also be happy to read something about it. Especially curious about how folks are going with accuracy problems when sampling can lead to inaccurate picture - do they have some more data for specific sensitive places or what. Interesting!

IMO sampling is one of the most undercovered topics in the whole o11y area. There are either claims "sampling is not needed" (confusing, untrue), or this one you mentioned. There are no proper guidances, examples of how to do it properly.

Not sure if this is a trend, though. All-senior teams are not without specific challenges, and so successful examples come with a deliverate investment from leadership to the right culture, hiring process, etc.