7
points
gkapur
Joined 94 karma
I work in venture capital.
- Congratulations to their team. SQLGlot is a really powerful tool that a lot of companies use so a huge contribution to the OSS community so hopefully it continues to be supported and gets better and better!
- There was also Wing cloud (fka Monada) and there’s Mojo by Modular (https://www.modular.com/mojo.)
Feels like two types of companies raised money: - Companies trying to couple the cloud with a programming language. - More recently, companies trying to couple GPUs with a programming language/alternative to CUDA.
Will be curious how this generation goes.
- If you are running things locally (I would think especially on the edge, whether on not the LLM is local or in the cloud) this would matter. Or if you are running some sort of agent orchestration where the output of LLMs is streaming it could possibly matter?
- I’m convinced I get more “deals” (temporary discounts) from Uber without Uber One/after canceling it, which offsets the benefits from Uber One.
I don’t see those deals on Uber Eats so it feels like the real value of Uber One is for heavy Uber Eats users.
PS. Worth going through the cancellation flow when you are up for renewal as they will probably offer you 50% off Uber One.
- Today there are so many other solutions: Stytch, Descope, PropelAuth (For B2B companies), and others.
VCs went a bit ham on this category when Auth0 got bought. I sense that the general thought process was: Auth0 multi-billion dollar company -> Auth0 will become crappy under Okta > replace Auth0 and build a bit company.
- Basically people are constantly calculating metrics based on existing tables. Think something as simple as a moving average or the sum of two separate columns in a table. Once upon a time you would set up a cronjob and populate these every day as a SQL query in some python or Perl script.
Dbt introduced a language for managing these “metrics” at scale including the ability to use variables and more complex templates (Jinja.)
Then you do dbt run (https://docs.getdbt.com/reference/commands/run) and kapow the metric is populated in your database.
More broadly dbt did two other things: 1. It pushed the paradigm from ETL to ELT (so stick all the data in your warehouse and then transform it rather than transform it at extraction time.) 2. It created the concept of an “analytics engineer” (previously know as guy who knows SQL or business analyst.)
- Thanks for the transparency and thoughts!
- What’s interesting is how much it contrasts with TechCrunch’s story: ‘Most of Command AI’s 30-person, San Francisco-based team will be joining Amplitude. Command AI’s co-founder and CEO James Evans wouldn’t reveal the terms of the deal, but said candidly that an acquisition wasn’t something he’d been planning on. “Our growth was great and we had plenty of runway,” Evans told TechCrunch. “We weren’t out shopping ourselves or anything. But when Amplitude reached out a little while ago — this summer — we got really excited about the combination and became convinced that we could grow faster and reach more users together.”’
- Interestingly, according to Axios, the price was pretty limited: "Amplitude (Nasdaq: AMPL) acquired CommandAI, an SF-based software user experience startup, for $20m (net of cash). CommandAI (fka CommandBar) had raised around $23m from Insight Partners, Itai Tsiddon, Thrive Capital, and BoxGroup."
I would be curious to learn more about the rationale to sell the business as I understand the strategic value to Amplitude. Interestingly, these next-generation digital adoption platforms have generally been pretty challenged.
- Not an expert in the space at all and it does seem like people are exploring new file and table formats so that is really cool!
How does this compare to Lance (https://lancedb.github.io/lance/)?
What do you think the key applied use case for Vortex is?
- Been watching this episode unfold on Twitter and has read about Matt’s domain hijacking of thesis, etc.
It seems to me like Matt is the type of person who likes to hide behind character and other ad hominem attacks rather than addressing actual issues at hand. Perhaps this is because of a psychological issue but I can’t really know. Normally I would think the community would be repulsed by this and would find an alternative.
What is remarkable is that there has not been enough community animus to fork relative to Terraform and OpenTofu. It shows the power of the underlying GPL license and “plugin” approach. Something to think about for other companies that may be thinking about relicensing versus just building a durable ecosystem around their proprietary brand and assets.
- Not really. This is more like SQLite or DuckDB for vector databases (on disk.) Chroma is more like redis for vector databases (in memory.)
We have seen similar products in the olap space, as well, ie. Clickhouse local.
- This whole thing comes off as tone-deaf and deceptive even to me (who is all for COSS monetizing.) Warpstream was sponsoring Benthos, it sounds like they didn't get a great heads up of this happening, which makes the project owner sound self-serving. Then you renamed the repo and relicensed some connectors all in one go without giving anyone from the community a chance to opine or think about how this affects them.
Finally, Redpanda did some partnerships with vendors nobody cares about whose businesses are at risk to show how you are opening up the ecosystem.
It actually comes off as somewhat malicious and Ashley's note where he notes he didn't read the article also comes off as not caring about developers (even insofar as he has facts wrong -- if the plug APIs remain compatible this creates more choice for users.)
- I empathize for the co-founder who was CTO and became CEO. I imagine some of the challenges come from the fact that there was a big chunk of equity owned by the original founding CEO. As the remaining co-founder, I can imagine feeling like I was climbing a grueling uphill battle to just get back to the most recent $300 million (!) valuation and a large amount of the upside of that battle was going to someone who abandoned me/the team. That said, I'm not privy to any equity adjustment that happened after the original founding CEO left.
That said, the after-the-fact communication feels lacking. It sounds like your CEO could have at least explained things in more detail after the fact (ie. Why he felt a pivot would be necessary to get to where the business needed to go.)
Such is life, I guess. I'm wishing you and the team the best for the future!
- Do what it’s worth supabase definitely feels slow to me. Neon, in contrast, feels lightning fast for my workloads.
- Very cool! Function calling seems to be a new paradigm that is really taking off.
Does anyone know how the LLM vendors are actually implementing function calling? Is it just a thoughtful prompt and a loop where they parse outputs and check if it corresponds to the arguments of the function? Or something else?
- 2 points
- Cool product! A few questions mostly out of curiosity:
- What underlying data flow technology is this based on? Is it Timely dataflow a la Materialize, something else a la RisingWave, etc.? - What can the materialized views not have? Ie. Does it support window functions? Unions? User-defined functions? - Finally I’m curious about how you think of the landscape versus RisingWave, Readyset, Materialize, etc.!
The TLDR/key (from my understanding) is that verifying N tokens can be faster than generating N tokens.