Igor Cotruta

Business Intelligence developer. Tweeting about #PowerBI and #DataViz

[Igor Cotruta @igocrite]

Forget about writing your own DAX parser.
You can pass a valid DAX query to👇

$SYSTEM.DISCOVER_CALC_DEPENDENCY

And get all the underlying dependencies in under 100ms.

Peeps parsing the Log Analytics traces know what I'm talking about😉

[Igor Cotruta @igocrite]

@TobiM This the evolution of comparison slop?

[Igor Cotruta @igocrite]

@theo Comments
The models think the devs are stupid for asking them to do the work.

[Igor Cotruta @igocrite]

@matsonj @DSJayatillake The number is true and false at the same time but it’s wave collapses when you look at it 👀

[Bad Pote @bad_pote]

La vérité, c’est comme ça maintenant.

[Igor Cotruta @igocrite]

@SQLGene Seen a lot of meat designers delivering this type of artefacts as well 😉

[Mo @atmoio]

I’ve spoken to over 50 people 1:1 over the last several months to better understand how they code with AI.

It’s an absolute mess.

Here are the top 8 mistakes I see.

0:00 - Intro
1:19 - Mistake 1: Talking to 2026 models like they’re 2025 models
2:43 - Mistake 2: Devising funky metaphorical factories to orchestrate your agents
3:59 - Mistake 3: Treating context as sacred
6:02 - Mistake 4: Treating previous conversations as sacred
7:59 - Mistake 5: Having a todo list
10:00 - Mistake 6: Using funky tools
13:47 - Mistake 7: Stressing over code quality
24:27 - Mistake 8: Using branches and worktrees

[Igor Cotruta @igocrite]

@SQLGene Too many things going on in this screenshot 😅

[Igor Cotruta @igocrite]

Hey, Jev promoters, have you ever called the OpenAI API with logit_bias? It's been there from the very start.
You can also count it as free output tokens.

[Igor Cotruta @igocrite]

@kurtbuhler Compare a screenshot of PBI Desktop from 11 years ago with the latest releases. They look about the same.
And we're talking about MS, which has an unlimited talent pool

[Alex Monahan @__AlexMonahan__]

Change a few DuckLake defaults and you get a 30% speed boost. Partition and cluster well and the gains can be bigger.

Matt Martin and I just finished the performance optimization chapter of the @OReillyMedia Definitive Guide, and we're giving it away!

https://t.co/UxOOpyfyue

[Igor Cotruta @igocrite]

@GergelyOrosz FTSE500 much?

[Igor Cotruta @igocrite]

@mim_djo There were quite a few good bots back in the day that would notify you about what ppl you follow talk about. If they engage in a conversation that you can contribute to.

[Igor Cotruta @igocrite]

@jezell Tried OPFS about a year ago. The blocker was extensions that do their own I/O instead of going through DuckDB's filesystem layer, so they never see it. Still on my radar. But I'd only ever thought of it as storage. Spill is a much better fit.

[Igor Cotruta @igocrite]

I put DuckDB-WASM in a Google Sheets sidebar.

Attach Parquet, Avro, PBIX, SQLite, GeoJSON and a bunch of other formats, query them with DuckDB, then write the result straight into the workbook.

It also works with httpfs/S3, DuckLake and Iceberg.

[Igor Cotruta @igocrite]

@jezell Fair point.
The result is capped, but the joins and GROUP BY on the way there aren't.
No spill today. Good idea. OPFS spill files are scratch anyway, so a sidebar that's wiped on close doesn't get in the way.

[Igor Cotruta @igocrite]

@snappercayt QUERY works on data that's already in the sheet, one range at a time, with no joins. This works on data that isn't: Parquet, SQLite, S3/Iceberg.
Full DuckDB SQL runs on those, and only the result comes back.

[Igor Cotruta @igocrite]

@jezell No WASM64 yet.

The sidebar is wiped on every close, so it was never meant to hold data.

DuckDB is the compute. Parquet/Iceberg stay in S3, get read over range requests (only the columns and row groups the query needs), and just the result lands in the sheet.

[Igor Cotruta @igocrite]

@__AlexMonahan__ Here is the setup that works at the moment
https://t.co/6YujIjdr6y

[Igor Cotruta @igocrite]

@__AlexMonahan__ I'll do a write-up eventually, since it's quite interesting.

I couldn't load motherduck in the sandboxed iframe that GSheets provides. It's worked since v1.5.2, when MotherDuck removed the cross-origin isolation requirement for its client.

[Igor Cotruta @igocrite]

@oleks01 DuckLabs joined AWS this month, so not sure if the fun names will last.