The claim of "built for teams that measure latency in microseconds" is conflicted by the pseudo demo showing a simple query that takes 4 seconds. I could assume that this is slowed down for effect, but the glowing green label[1] tells me this is a "real engine run" so now I'm confused.
My core question is why would I use this instead of duckdb and writing SQL as God intended[2]?
It's a little weird to truncate the history [0] of your open source project and lock the v1 commits in a separate repo with limited access [1]. Are the authors embarrassed of the commit history?
I just did this recently to reduce the noise I got for using 'superpower's skill for an internal library. Personally trying to find the right git ops balance of observability/transparancy of agent work while keeping human readablity. I haven't seen any accepted standards or workflows, its really interesting to watch dev culture evolve with this
You start with a hunch and then dive into their github repo to confirm that Claude is a top contributor. At least, that's my strategy. Honestly though, I don't care about the website being AI generated. That's not what's important for a project like this.
I was just thinking yesterday that some of what these AI generated sites have exposed is just how vacuous and low information content a lot of websites have become. The difference is previously you could at least say, "hey, someone at least spent a little money to create this" now that cost is pretty much zero. I think the answer is to simply demand more from people who have real products. If you have a product, I want to see a screenshot. Even better a demo and I'm not hitting the "schedule a demo" button. Either you step up and demonstrate capability or you're going to get lost in the noise.
afaiu this is a kdb-adjacent columnar runtime and query engine
think SIMD, morsels and so on ~ Clickhouse, DuckDB …
but more like a lisp-like vector language and execution engine at its core + allowing for numeric and analytics adjacent use-cases instead of the other way round
The claim of "built for teams that measure latency in microseconds" is conflicted by the pseudo demo showing a simple query that takes 4 seconds. I could assume that this is slowed down for effect, but the glowing green label[1] tells me this is a "real engine run" so now I'm confused.
My core question is why would I use this instead of duckdb and writing SQL as God intended[2]?
---
[1] A clear indication of AI generation
[2] This is a joke.
FWIW I liked your joke :D
I think God did intend SQL for us. There is no going back. It's too deeply intertwined.
[2] duckdb indexes suck
Remember duckdb is for OLAP, not OLTP!
olap doesn't mean no indexes. Prev gen olap (cubes, star schemas) were index heavy, you can't achieve subsecond analytics without indexes.
It's a little weird to truncate the history [0] of your open source project and lock the v1 commits in a separate repo with limited access [1]. Are the authors embarrassed of the commit history?
[0] https://github.com/RayforceDB/rayforce/pull/166
[1] https://github.com/RayforceDB/rayforce2/blob/master/README.m...
I just did this recently to reduce the noise I got for using 'superpower's skill for an internal library. Personally trying to find the right git ops balance of observability/transparancy of agent work while keeping human readablity. I haven't seen any accepted standards or workflows, its really interesting to watch dev culture evolve with this
At this point, I see an AI generated site I quit.
this might be good stuff, but I'm not even interested anymore!
How can you tell it's AI generated?
It has the look: fonts, style, layout, too many em-dashes, and weird phrasing like "Every surface reaches the same optimizer and execution core."
Just like you can recognize Hemingway's prose voice from a few sentences. After dozens-to-hundreds of hours with Claude, you can tell.
You start with a hunch and then dive into their github repo to confirm that Claude is a top contributor. At least, that's my strategy. Honestly though, I don't care about the website being AI generated. That's not what's important for a project like this.
I was just thinking yesterday that some of what these AI generated sites have exposed is just how vacuous and low information content a lot of websites have become. The difference is previously you could at least say, "hey, someone at least spent a little money to create this" now that cost is pretty much zero. I think the answer is to simply demand more from people who have real products. If you have a product, I want to see a screenshot. Even better a demo and I'm not hitting the "schedule a demo" button. Either you step up and demonstrate capability or you're going to get lost in the noise.
Let's add it to ClickBench: https://github.com/ClickHouse/ClickBench/
afaiu this is a kdb-adjacent columnar runtime and query engine
think SIMD, morsels and so on ~ Clickhouse, DuckDB …
but more like a lisp-like vector language and execution engine at its core + allowing for numeric and analytics adjacent use-cases instead of the other way round
The honesty is what makes it good
A graph analytics database? Could you talk about the motivation, and competition?