tdfirth

Great post! I've worked on a few query generating systems of varying degrees of complexity, and SQL is pretty rough as a compilation target. I've spent more time than I'd have liked wrestling with queries as comparably ungodly as the beauty you have presented.

One trick that I always find helpful is to unroll the subqueries into a long sequence of CTEs. You can do a depth first traversal of whatever data structure produces your subqueries and build a list of CTEs as you go (and track of all the CTE names of course). The ordering 'just works'!

Slightly nicer SQL doesn't seem important at first for stuff like this, but it's actually a godsend for debugging. Every operation in the query gets a name that's accessible in the outermost scope, so you can slap a `select * from cochoice_abc123` on the end and see what's going on.

Figuring out what's going on in those subqueries was no doubt hellish at times.

No rename symbol? What am I missing? It seems like a no brainer.

Ha! I had not intended to imply that one is better than the other, but I am glad that it made you feel good :).

I also came "from above".

I don’t think this is confusing to the vast majority of people writing Go.

In my experience, the average programmer isn’t even aware of the stack vs heap distinction these days. If you learned to write code in something like Python then coming at Go from “above” this will just work the way you expect.

If you come at Go from “below” then yeah it’s a bit weird.

American hegemony, and all that.

I believe they change color to express emotion.

You're right. How naive of me.

Google kills Gemini cloud services is the best one. I can't believe I haven't seen that joke until today.

In my experience (I've put hundreds of billions of tokens through structured outputs over the last 18 months), I think the answer is yes, but only in edge cases.

It generally happens when the grammar is highly constrained, for example if a boolean is expected next.

If the model assigns a low probability to both true and false coming next, then the sampling strategy will pick whichever one happens to score highest. Most tokens have very similar probabilities close to 0 most of the time, and if you're picking between two of these then the result will often feel random.

It's always the result of a bad prompt though, if you improve the prompt so that the model understands the task better, then there will then be a clear difference in the scores the tokens get, and so it seems less random.

It should have always worked this way. Without this feature you take the algebra out of relational algebra. That's the root of most of the composition issues in SQL.

Sadly it's a few decades too late though, and sadly this just fragments the "ecosystem" further.

The best strategy is to write a helper function in the console to click for you. Then invest heavily in the DVDs, DVD bounce rate, stimulation per bounce, and general SPS increases.

I reached several quintillion stimulation, at which point I was offered to purchase "go to the beach" for 2 million. This ends the game and plays a relaxing beach video.

You too can get to the beach in just 5 minutes.

This is a beautiful comment, and I couldn't agree more.

There is only one standard of accomplishment and it's set by people like Mozart.

Accepting that is humbling, but it's required to know yourself and grow. My contributions probably won't amount to much, but Mozart (et al) have shown us what good looks like and it's fun to strive.