mjirv

It basically already has since they started developing dbt Fusion, so in that respect this probably doesn’t change much.

I expect they’ll keep developing Fusion but possibly as even more of a commercial-only offering than it already was.

sure, why not?

Is OpenAI having trouble acquiring enterprise customers?

Yep. Tool calling is terrible across all Gemini models. I’m not sure why, when the model itself is so good.

Claude 3.7 Sonnet for function calling, and it’s not particularly close in my experience.

Not sure about multimodal as it’s not what I work on.

Do you have results for gpt-4? I’d be very interested in seeing the lift here from their last “big one”.

thanks. guess I forgot to paste the URL (it was that one)

unless you know what you’d be getting paid at silicon valley tech companies, don’t say you’re “overpaid”

your comps are with “working in San Francisco”, not with your peers

Out of curiosity, I googled several of the pubs he mentioned. All but one* was still around.

*I found a pub called the Ranelagh, but it’s not in Pimlico, so I assume it’s a different one. It was the one he described as “really terrible,” so no big loss, I suppose.

Addendum: the other interesting thing I noticed was the ones he derided as having been “modernized” in the 1960s were also newly renovated today, with airy, Scandinavian, 2020s aesthetics. Presumably because unlike the traditional pubs, the 60s style became dated pretty quickly.

re: dbt, it looks like Hashquery is more for running analyses than doing in-data-warehouse transformations.

Unless I'm mistaken there's no native way in Hashquery right now to materialize the results back to your warehouse like you would with dbt.

Current OpenAI job postings top out at $385,000 for full stack engineers which is 98th percentile in the US (and less for San Francisco).

The €80k is the start of the Mistral salary range. €100k is the top, which would put you in the top 1% of earners in France.

Three out of every four desk workers report working in the 3 to 6pm timeframe, but of those, only one in four consider these hours highly productive.

It’s nice to see some data on this. A lot of people talk about a 4-day workweek, but personally I think a 6-hour work day would be better.

At least in white-collar jobs, I think most people could fit the same amount of productivity into fewer (focused) hours.

Plus a day that’s roughly 9-3 instead of 9-5 would have the benefit of being aligned with school schedules (at least here in the US) and make things easier for working parents.

Unlikely. Non-poaching agreements are illegal under federal antitrust law.

The key line:

“Microsoft has assured us that there are positions for all OpenAl employees at this new subsidiary should we choose to join.”

No. As one of the other commenters mentioned, Sam (and possibly Greg) probably recused himself and didn't vote (likely forced to by the board's bylaws).

Nice! We could be in the market for something like this soon.

What’s your quick pitch for why someone should use Revert vs competitors like Merge.dev and Nango?

Interesting that they don't have the watermark on the "DALL-E 3" images.

Sure, I understand why you’d want real-time. I guess the OLAP part is the one I’m not sure about here.

Most of these use cases seem more “pull all data for a specific user/post/company” and less “do analytical queries on millions+ of column values.”

Nice. We (Delphi[0] - enterprise AI data assistant) do something similar internally to power our app.

We’re all typescript under the hood so I’ll give this a look and see if we can use it.

Symphony wouldn’t support other LLMs currently, right? Only GPT-4?

[0] https://delphihq.com

Potentially dumb question, but why does “who viewed my profile” (the feature mentioned in the post as the original use case at LinkedIn) require a realtime OLAP datastore anyway?

I stumbled upon OpenRouter[0] a few days ago. Easiest I’ve seen by far (if you want SaaS, not hosting it yourself).

[0] https://openrouter.ai

Cube (https://cube.dev) is a good one.

Others include AtScale[0], dbt's MetricFlow[1], Google's Looker[2] (also a BI tool but powered by a semantic layer), and Propel[3].

[0] https://atscale.com

[1] https://www.getdbt.com/product/semantic-layer

[2] https://cloud.google.com/blog/products/data-analytics/introd...

[3] https://www.propeldata.com

They're kind of an updated version of OLAP cubes if you're familiar with those.

Typically semantic layers sit on top of a data warehouse, let you define metrics using code or a UI, and provide APIs or SQL connectors so that you can query them.

Yeah, similar to what you and the other commenter from Definite said, we (Delphi)[0] find semantic layers way better for this kind of work than just going straight to a database/data warehouse.

One thing you really need with LLMs is consistency. Text-to-SQL kind of lets the LLM do whatever it wants - join tables that shouldn't be joined, define aggregates one way in one query and another way in the next.

Because semantic layers define how tables should join, measure definitions, etc., they mean people get consistent results from one query to the next, which builds trust in the LLM.

Cube (which was mentioned in another comment and has a great open-source semantic layer) has a good article about that here: https://cube.dev/blog/semantic-layer-the-backbone-of-ai-powe....

[0] https://delphihq.com

The average human cannot program, not even a little.

The average human cannot summarize (for example) Anna Karenina from memory.

It just needs to be better than the average person, not the average programmer

OpenAI arguably already has AGI. GPT-3/ChatGPT is much better than the average human at coding, text summarization, etc. (“system 2” cognitive tasks).

Recording: https://twitter.com/winglets747/status/1622054756666540033?s...

Not the OP, but:

Oftentimes data teams have those data sources in the warehouse already (or a process for ETLing new ones in).

Copying data only once means lower costs (maybe) and fewer inconsistencies.

Plus if they’ve done work to join/clean/transform the data in the warehouse they can take advantage of that for product analytics without having to reinvent the wheel in ChartBrew or Mixpanel.

What? Noah’s point about America’s inability to build things has nothing to do with swapping out Gavin Newsom.

Cool! I'd be interested in a more detailed writeup of how you did this. It's not perfect but looks like very solid work for someone trying to land their first SWE role, so well done.

Maybe re-submit it as a Show HN?

Also, it would be interesting if the answer linked to the relevant episode.