worst_cases – ORGANIZING CREATIVITY

Telling ChatGPT the OCR might be rough gives it permission to hallucinate

2026-03-18

Telling ChatGPT that «OCR was a bit rough» gives it the permission to treat a text not as fixed, but as incomplete / unreliable input. Great in an RPG scenario, but probably quite devastating in real contexts.
[show the article]

Leaving Academia – for the Second Time

2025-12-01

Some personal reflections on why I left the university a second time after working there for ten years.
[show the article]

AIs or rather ChatGPTs View on Mankind Improving Itself

2025-11-29

After using ChatGPT to identify mankind's patterns that emerge only at scale and what these patterns reveal about humans, the question becomes, whether mankind can do better. Well, it depends what we mean by «better».
[show the article]

AIs or rather ChatGPTs View on Mankind

2025-11-29

As AIs are trained on vast amounts of human knowledge, they should have a unique position in seeing mankind for what it is, simply via the emergent patterns in the data and what those patterns imply about the species that produced the data. And yes, it did not disappoint.
[show the article]

Psychology and Morality Blind Spots via ChatGPT

2025-11-14

ChatGPT did identify a few interesting blind spots in psychology, esp. when it comes to therapy. Image: ChatGPT Logo (left side)
[show the article]

ChatGPT is a great conversation partner but not on certain topics

2025-11-14

AIs like ChatGPT mirror our tabus and fears. Via structural filtration it takes mainstream positions, leading to differential epistemic courage in various «emotional» areas. But these distortions can be detected and worked around it.

Image: ChatGPT Logo (modified)
[show the article]

Changing ChatGPTs Tone and Attitude — Especially getting the pre 5-1 playfulness and exploration back

2025-11-13

ChatGPTs updates can change the basic interaction style (kills creativity and exploration, is much to abrasive), personalization brings it back.

Image: ChatGPT Logo.
[show the article]

Creating your Own Feedback Team — ChatGPT AI Council of Muses

2025-07-13

You can create multiple personas in ChatGPT — each with its unique strength — and have them work together. That might give you advice from multiple perspectives.
[show the article]

A solution to the «The Road to Hell is Paved with Good Intentions» Issue

2025-06-15

As the road to hell is paved with good intentions, you should test your good with the same rigor as you test others' harm.
[show the article]

Being a real Learner again

2025-05-25

Some short reflections on being a learner in a subject for which I have no talent or prior knowledge.
[show the article]

This is a private blog showing my opinions. If you have an issue with what I write here, either don’t read it or talk to me directly. Don’t waste your time — or mine — dragging my employer into matters where they hold no authority.

The institution I work for, or rather the people officially representing it, have already distanced themselves from this blog. I find this unnecessary, as I’ve clearly stated in the «About this Blog» section that this is a personal blog.

Frankly, I have nothing but contempt for those who lack the backbone to address their concerns directly and instead resort to passive-aggressive tattling to others. Such behavior strikes me as not only cowardly, but also spineless, gutless, and wholly unprincipled. Moreover, these efforts are as pathetic as they are futile.

That said, I might listen if you present compelling arguments and evidence, as constructive feedback is always welcome. Who knows? We might even both gain something valuable from the interaction.

A comment for those seeking to use this site for personal gain: Given the increase in requests, let me be clear. I write on this blog because I want to. It’s my hobby, my playground. Sometimes people point me to interesting products/services and I write about them. But any request regarding ads or sponsored placements ends up the trash without a reply. And if you think something would be of interest, differentiate yourself from the spammers by referring to a posting — in an intelligent way. (I get enough auto-generated mails to identify them immediately.)

BTW, posts can get updated after I published them if I spot spelling errors (not a native speaker) or think a different wording might improve precision and clarity.

This blog is not focused on a single topic, or method. As long as it is relevant to improving creativity (or allowing it in the first place), it's fair game.
Some postings on this blog deal with freedom, as I think that we need freedom of thought, of speech, of association, etc. pp. to solve mankind's problems. Thus, some postings may seem a bit remote when it comes to organizing creativity. Freedom is, however, the bedrock of creativity.
The heterogeneity of the postings can make reading this blog a bit cumbersome, at least if you are only interested in one topic. You can either use the search function (above), or use the categories or the tags to narrow down the postings you see.