Product @ ClickHouse Thinking about agentic experience & identity

Joined April 2015
Al Brown retweeted
We are live with the 26.10 ClickHouse Release call! Join us, on youtube or LinkedIn youtube.com/watch?v=k30ikPXN… (a thread)
1
2
9
1,604
ClickHouse 26.10 release call today. Lots of JOIN work in this one! I’ll share my highlights after the call. c.house/2610ReleaseCall
41
Al Brown retweeted
Replying to @alexatallah
Last week I was on a sales call with TiDB. We were thinking of using it for an upcoming product launch. After they ran through their pitch I asked them specifically about day-to-day security measures they take. They couldn't really answer. Went with ClickHouse.
1
1
321
Anthropic enters the personal assistant race with Claude Duet, built on Opus 5.5. The best model right now, for me, so I’m here for it. A couple of weeks with OpenAI’s dot has been reasonably useful, but no game-changing use cases yet. claude.com/resources/article…
35
1/6 ClickHouse is a launch partner for @Cloudflare OS, with an integration for working with live ClickHouse data. I've been digging into the wider app model, especially how people can share an app and personalise their own copies with AI.
1
2
105
5/6 There's an interesting attempt to solve the maintenance problem too. Personalised software means taking responsibility for the features you added and everything that came with the app, including bugs and security fixes.
1
14
4/5 Restarting clears the memory cache. Results on disk keep their original expiry time. Upgrading or downgrading invalidates them. Disk results can still be evicted before TTL expires, and serving them requires reconstructing the result in RAM.
1
17
5/5 First configure a named filesystem cache. For one called 'query_results', use use_query_cache = 1 query_cache_on_disk_cache_name = 'query_results' query_cache_ttl = 60 The TTL is in seconds.
16
1/5 ClickHouse is getting a disk backend for its optional query cache, so saved results can survive a server restart. Useful when lots of people open the same expensive report and can tolerate a short delay before new data appears.
2
1
3
219
3/5 With both backends enabled, ClickHouse checks memory first, then disk. Eligible results can be written to both. Disk stores compressed Native chunks that preserve compact columns. A constant column can stay one value plus a row count.
1
39
2/5 Caching stays a choice. New inserts don't invalidate a saved answer, so you set a TTL for each report. Queries that need current data can leave the result cache off and read fresh data directly.
1
33