ML is just applied stats. Stats is just applied algebra. LLM is just ML backward and with an extra L.

The obvious reaction here is to shift at least some of the hiring out of the country to get access to the talent. The obvious counter reaction is to tax payments and wages to foreign employees and contractors. Which will also provoke a reaction. And none of this makes the US stronger or smarter.

About a decade late with this, but: Someone should have started a social media ad agency called Twaddle.

It's tempting to take shortcuts that give you speed today by mortgaging speed tomorrow. Trouble is, today is yesterday's tomorrow.

I tried 2 different english to insights sql llm agents from reputable vendors in the past week. Data analyst jobs remain safe. Firmly in the toy category for now.

Happened to be by the Cloudera building in the south bay earlier. Checked LinkedIn and discovered I have literally 0 1st degree connections who work there now. Not unexpected, I guess, but, man... betwen hnwx and cldr I used to know like 100s of folks there

Heck of a vote of confidence from LinkedIn for Flyte: www.linkedin.com/blog/enginee...

About once every two years I have cause to re-learn a very important data lesson: never, ever, trust analysis based on ratio metrics.

Trying and failing to make the page edits look right? Is offline access lackluster? Tired of AI upsell as a replacement for poor search quality? You might be suffering from Notion sickness.

Phone book, noun: an ebook you read on your phone.

Looking up latin phrases on Google results in an AI response in French a good % of the time. Fortunately, my French is slightly better than my latin.

The scariest thing about dinosaurs is that they were huge, absolutely dominant, *and humans had nothing to do with them dying out*

Andor is a very boolean show.

Just 45 minutes north of SF, and AI means something completely different on a dairy farm in Sonoma, in the context of breeding livestock. I seriously wasn't tracking for a few minutes there.

A Random Walk In Question Space: A particularly ineffective, but popular, investigation method commonly found when the investigator does not have direct access to the data, and does not directly incur the cost of asking for irrelevant analysis that does not get them any closer to useful answers.

Still thinking about it. Serious "retired samurai" energy. Man's just standing there, pouring beers, drying glasses, listening to young coders in this city of startups go through their ups and downs. Dispensing wisdom about code and life.

A friend once said that experienced software folks get a weird superpower: they can diagnose and even predict bugs / seams in systems they've never seen the code for, just from knowing how stuff is put together. It's a thrill every time one pulls this off successfully :).

Just saw a LinkedIn profile that is a real career to aspire to: * CS Professor * Google Director of Eng * Distinguished Eng at Microsoft * Brewer and bar owner.

DuckLake makes tons of sense, as all things DuckDB does. How long till they implement CMETA style optimizations right in the query planner? vldb.org/pvldb/vol14/...

I am just a boy Standing in front of an agent Asking it to try again and think harder.

Sure, an english to sql agent on your db won't give you an answer as accurate as a data analyst would, but it will give you incorrect answers so much faster!

Seeing com.google.hadoop is still a mind bender. Don't think any of us saw this coming back in the day... or at least I didn't.

Surprising how much people talk about Iceberg providing "separation of data and compute" given that it, you know, doesn't.

Wow, Informatica still has enough of a book of business to be worth 8B?!

They say "give away your toys" but then you go and explain to someone how to chase down expensive queries and optimize the bejesus out of them, and now you have to stop cause you asked them to do it, and all the joy is drained out of life.

If Morel gets traction, I can almost guarantee someone will write a SQL-Morel transpiler and call it Squirel

The downside of darkmode-for-everything is that all the web apps start looking the same and I end up searching for my Notion pages in ChatGPT.

If you are an aspiring founder out there, struggling to find the perfect name for your company, consider that having a name as stupid as "Supabase" didn't keep a good product from achieving a $2B valuation.

When Windsurf can't make dependency resolution work out and decides to just "implement our own lightweight version of dbt"

So BigQuery pushed a change to how they allocate slots this morning, huh 😡

There was a post going around a week or 2 ago about how someone uses coding assistants, which got a lot of kudos. Google is filled with superficial crap on the subject and bsky search is a baby... anyone got the link?

I just love what the bitdrift team is doing. Even if sometimes it feels like they are building this cause they were so annoyed at the observability tooling they were getting from me 12 years ago 🤣

Who's working on the BigQuery Autoscaler @ Google these days? I have questions! 🤓

Pandering, cowardly, egotistical dimwits.

If your pages take over 10 seconds to load, I'm literally not interested in any feature updates not having to do with speed.

Typoed "functions" as "functinos" and tbh I like the second way better. Sounds more... fun.

If you make a browser plugin that basically stores 1MB of text, max, and Task Manager shows your plugin is eating 3GB worth of RAM, you need to fix some code. That optimization is no longer premature.

Got tripped up by an INFO message that preceded an ERROR one, not realizing they were about two different conditions - one of them benign. Should've read my own book and remembered what the log levels meant. Save yourselves, buy and read The Missing README.

When I was a baby VP, my startup CEO told me to hire half a dozen data scientists to create a DS team, so I did. Rookie mistake. Take these instructions seriously, not literally. He wanted a DS function, not 6 DS's. I should've hired 4 DE, 2 DS.

A comes before B, which makes abbrevating "before" and "after" rather inconvenient.

It really is surprising how often, given a sufficiently complex sql/dbt setup, you can find 80% efficiency improvements for heavy queries with a couple days' work.

"Codeium, optimize this massive sql query." "Ok, here are 5 ways to improve it!" "Why would 3 and 4 be faster?" "Lol, they wouldn't. Should I just try 5, then?"

Proxy measurement? More like a poxy measurement, amirite?

I'm still thinking about this discussion, and if I'm honest I'm a bit afraid I'm looking for what a great engineer looked like in the 2010s, not what they look like today. I don't think it's wrong, exactly, but what does excellence look like for folks who grew up as engineers after 2020?

You wouldn't need all that monitoring and alerting if your code worked right to begin with, you know. What a defeatist attitude. Believe in yourself! Ship blindly with pride!

My life is pretty calm and boring, which is why I use both PyLance and MyPy in the same project. The adventure of them being mostly but not totally in agreement brings that sizzle and flavor into my day.

"Full stack DW platform using Scala and Spark, processing 50GB per day" 😵‍💫 We really did a number on the industry with the big data hype. I'm sorry. Glad the duckdb folks are swinging that pendulum back.

That feeling when you fix a None dereference in a line that's had a `# type: ignore` on it since 2022 and finally tripped on what the type checker was trying to warn you about in 2024.

Slack Driven Development (SDD). You heard it here first.

Comments? Suggestions? Call us! 1-800-DEV-NULL