Pinned Tweet
5 questions that Data Engineers should know answers for:
1. Why are we doing it?
2. How do we know what we have and what we know?
3. How do They know we are doing it correctly?
4. What will or can happen if it goes bad?
5. How do we get rid of it?
dear imgui 1.93.0 released! (123rd release)
github.com/ocornut/imgui/rel…
- ImDrawList rework w/ many improvements and fixes (e.g. inward/outward stroke pos, better thick strokes, AA ends, baked rounded corners & more)
- mixed values, and 50+ other fixes/features.
Štefan Urbánek 巫思蔚 retweeted
Replying to @signulll
Better architecture?
No, not at all.
All architecture is design, but not all design is architecture. Architecture represents the set of significant design decisions that shape the form and function of a system where significant as measured by cost of change.
I would argue that code agents are not even good at design: they are simply choosing implementations based upon pattern matching, and even that being done at the most superficial of levels.
Architecture and design to a lesser degree are things that humans do, not coding agents.
Diagramming is safe from them, for now. "Make me an entity-relationship diagram of the draft": Polishing layout with 4 attempts of "image understanding" taking ~5 minutes each on a diagram the agent itself generated using graphviz dot to png. He came up with this. Do you see it?
1
2
124
Beautiful saying:
「人工智能都是人工,沒有什麼智能」
The beauty is lost in translation: "Artificial Intelligence [人工智能] is all human labour [人工], and no intelligence [智能]"
Supply chain now comes with a demand shock.
Supply Chain Stock Flow demo – clanker made with Poietic.
Notice funny quirk (bottom-right): he got really annoyed by unpolished SVG output (not seen here), it spent most of time fine-tuning node positions, gave up with a spacer dummy node.
Cheers to @soobrosa.
5
148
Supply Chain Stock Flow demo – clanker made with Poietic.
Notice funny quirk (bottom-right): he got really annoyed by unpolished SVG output (not seen here), it spent most of time fine-tuning node positions, gave up with a spacer dummy node.
Cheers to @soobrosa.
Štefan Urbánek 巫思蔚 retweeted
# Naming in CLAUDE.md
I am somewhat of a collector of modern low IQ lexical slop. Help us out with the collection!
"Transfer Funds" -> "Move Money"
"Registration Complete" -> "You're in!"
"Restore" -> "Put back"
"Error"-> "Oops!"
"System Failure"->"Something is broken"
"Return"->"Go back"
"OK" -> "Got it!"
"Download" -> "Get it now!"
"Cancel" -> "Never mind"
"Retry" -> "Try again"
"Access Denied" -> "Whoops! Not allowed"
"Confirm" -> "Yes"
"Disable" -> "Turn off"
"Tweet" -> "Post"
Epidemic with vaccination – A simple stock-and-flow system-dynamics model in Poietic Playground. Clanker made it. With help of @soobrosa . Stay tuned.
Fascinating how LLMs struggle with the name Poietic. 50% of sessions they sooner or later degrade by adding random characters. Surprisingly most common is "pocitetic", unrecoverable. From "poetic" they tend to recover quickly. Sometimes they funnily give up by globbing everything
1
3
116
Štefan Urbánek 巫思蔚 retweeted
Replying to @rahulj51
"The issue needed to be resolved is that this function which is supposed to do a thing is actually doing exactly the thing it is supposed to do, it is even documented and therefore it is not an issue to be resolved and everything is just fine as it is." – wasting my attention.
Štefan Urbánek 巫思蔚 retweeted
Replying to @Stiivi
I got my first agent generated, three(!) pull-requests to my not even officially announced open-source project.
Any recommendations for a reasonable "Only humans welcome" OSS project policy?
1
1
121
Current state of the Poietic Playground.
Pleasantly surprised that porting some features from now abandoned prototype in Godot to the current DearImGui-based was easier than expected.
Štefan Urbánek 巫思蔚 retweeted
From what I see, there are 3 types of people in Tech today.
1. Rapidly hurtling towards a massive burn out. They just don't know it yet.
2. Just arrived at the Burn-out station. Going through denial, resistance or slow acceptance.
3. Long term burn out survivors who have figured out how to stay burnt out but continue functioning.
Štefan Urbánek 巫思蔚 retweeted
The people who say AI is killing SQL can you please join the long line of people that Joe Reis and I will cackle about in 2030.
NoSQL didn’t kill SQL
Hadoop didn’t kill SQL
Tableau didn’t kill SQL
Spark didn’t kill SQL
dbt didn’t kill SQL
AI won’t kill SQL either
Every major shift in data has made SQL more valuable, not less.
The tools change
The interfaces change
The abstraction layer changes
Asking questions of data and expressing that logic in SQL has been durable for basically the entire existence of modern computers.
SQL isn’t disappearing. You’re kind of a silly goose if you think it is.
Data modeling (especially with SQL and complex data types) is king in 2026 because data modeling is just “context” and almost all analytical context is expressed in SQL.




























