Ben
@bengamble7.bsky.social
490 followers 540 following 190 posts
Tech sommelier @ Ververica Keeper of ducks, Builder of games Mostly blind🦯 (its not a hiking stick), He/Him, All opinions my own
Posts Media Videos Starter Packs
As much as I prefer writing other languages, it's annoying how much more productive I can be writing c++.
There's a 1 million request challenge to win!
As much as I prefer writing other languages, it's annoying how productive I can be writing c++.
There's a 1 million request challenge to win!
The @KuzuDB repo has been archived on github. But the entitlement never stops...
Opensource doesn't mean free.
Open source does not mean maintained. Things still take time and money to build and test.
If you're making money from it, maybe contribute / pay?
www.theregister.com/2025/10/14/k...
KuzuDB graph database abandoned, community mulls options
: Users left wondering whether to fork it or forget it as another FOSS project bites the dust
www.theregister.com
Flight delays on the way to flinkward means time to mess about with code!!

Turns out that if you spend ages hacking it, you can run @duckdb with more than one node!😁

With only minimal slowdown! 😭
The amount of guard rails I've had to add to my environment to stop @cursor_ai just faking stuff is extreme....

There's literally rules against mocks, and I still needed to add an mcp service that checks code paths are being called...

Seriously, I'm not in a hurry stop trying to rush.
I didn't think I'd find myself writing a cypher parser at almost 11pm on a Saturday just before flink forward, but that's life 🤣
Now to give the optimizer another poke...
Turns out that you really can build AND hot reload stuff with TinyCC so fast that it makes cython Dev workflows as fast as #python....

Shout out to @charshenton for the idea.
Continuing my experiments in cython, I built github.com/Ugbot/CyRedis because redis / valkey are cool but we end up not getting all the juice out of it in python land.
It's wildly experimental, but I let claude code write the readme 😅
I think my Next experiment will be "fasterAPI" 😎
GitHub - Ugbot/CyRedis: a smarter redis client implmented in Cython for python
a smarter redis client implmented in Cython for python - Ugbot/CyRedis
github.com
Today's interesting diversion was from building an operator of other operators in k8s.
Firstly, as much as I like the
operator pattern, the implementation leaves much to be desired. Not sure there's a better way but the current one is annoying...
Especially Cascading roles through systems
I'd entirely forgotten about that 🤣😭🤣
Hehe maybe 🤣but at the moment it's just an experiment
In a moment of true AI hallucination, #claude decided that the best way forward was to mix macros with template metaprogramming in C++

There are a few things I really didn't want to see again in code, and this was on the list...
In a moment of true AI hallucination, #claude decided that the best way forward was to mix macros with template metaprogramming in C++

There are a few things I really didn't want to do see again in code, and this was on the list...
Well, I'm in the deep stuff now....
I've got a #postgres plugin called wal2arrow and it's very very fast.
Now to see if I can make it work from Pgoutput directly...
Time to channel @gunnarmorling.dev ...
Gunnar Morling
Read More...
www.morling.dev
I may have fallen down the rabbit hole here but...
#apacheArrow had my curiosity but now had my attention...
And I'm building wal2arrow for #postgres and #mysql...
It's kind of awesome not to have json /avro around in the hot paths.
That's basically where I'm going with this 🤣 except I'm mostly working in Cpp and cython.
Processing 200m Rows a second is proving fun!
The vibe of today's claude code is an epilepsy warning!
There's no escaping the fact that python is slow. Slow enough that it's got a whole ecosystem of tools like numpy, numba and cython to fix that.
Currently I'm working on a way to apply numba dynamically across python code via ast and byte code parsing. Wierdly, it's working 🤣
A quick PSA:
The number of AI written "seo" articles is getting rather crazy, but if you have to, at least run the article through a fact check so you don't end up with content that makes you look stupid....
Also use an up to date LLM, the new ones are better...
We're having a family gathering for my dad's 70th birthday, and when I asked my nephew (16) what he was watching, he seemed surprised I knew what naruto was 🤣...
Then I felt old when he didn't know full metal alchemist...
But now we can watch cowboy bebop together!
I spent some time last week Playing with Apache Arrow.
Sometimes the numbers are frankly crazy....
I have a bunch of machines at home, but I've never felt as in control of them than after spending 30 mins with #claude setting up some ssh and rSync magic.
It's all the hacker vibes of Tmux without the muscle memory required
Reposted by Ben
The Trump administration ordered several National Park Service sites to take down materials related to slavery and Native Americans, including an 1863 photograph of a formerly enslaved man with scars on his back, one of the most powerful images of the Civil War era. nyti.ms/4nyyyaQ
This week I was on a podcast:
Get Kafka-Nated Episode 6: Kafka Meets Flink