Last year, I rescued a tuber from a compost bin at the local cemetery. It took until summer for it to come out of dormancy. I placed it in a partially shaded spot, and never forgot to water it. I gave it fertilizer but nothing seemed to happen. This morning I checked, and was surprised to find these beautiful flowers!
#flowers #begonia

I had made quinces jam from the left-overs of steam-juicing quinces for making jelly.

The jam was an experiment - the pulp in the steam juicer smelled and tasted too good to just throw it away.

@MatthewNewell told me that in the UK meat is sometimes served with quinces jelly,I tried some of the quinces jam yesterday: it's enriching the taste a lot!

Today I tried it with a rice-pepper dish. I added some chili sauce, some turmeric based spices mix and a spoonful of quinces jam - a very nice taste!

I guess that I will try to make some quinces chutney. Maybe @LifeTimeCooking has a recipe for that, too :-)

Here is the paper. I can see that molecular biology is moving in this direction - ecological systems theory, too.

https://users.dcc.uchile.cl/~cgutierr/papers/ecal2011.pdf

I found this citation in the Wikipedia page on "Ouroboros":

> Soto-Andrade, Jorge, Sebastia Jaramillo, Claudio Gutierrez, and Juan-Carlos Letelier. "Ouroboros Avatars: A Mathematical Exploration of Self-reference and Metabolic Closure". "One of the most important characteristics observed in metabolic networks is that they produce themselves. This intuition, already advanced by the theories of Autopoiesis and (M,R)-systems, can be mathematically framed in a weird-looking equation, full of implications and potentialities: f(f) = f. This equation (here referred to as Ouroboros equation), arises in apparently dissimilar contexts, like Robert Rosen's synthetic view of metabolism, hyper set theory and, importantly, untyped lambda calculus. We envision that the ideas behind this equation, a unique kind of mathematical concept, initially found in biology, would play an important role in the development of a true systemic theoretical biology."
https://en.wikipedia.org/wiki/Ouroboros
#RobertRosen #autopoiesis

This year I opened the last jar of quince jelly prepared by my grandmother: it sat on the shelve for 10 years, and surprisingly it was still delicious, but it was the last one. What to do?

Last weekend, my brother in-law gave me fresh quinces, and I borrowed the more than 50 years old steam juicer ("Tekla").

Yesterday I made quince jelly, a first!

Today I will try to make some kind of quince cheese from the (already prepared) soft parts of the pulp.

#quince #cooking #nostalgia

I've been using Gemma-4, both 26b (MoE) and 31b (dense) for a while. Both models are containers of normalized knowledge of all kinds of things, not just science or technology, but some of all of the internet. Gemma-4 is a free model from a company that makes money with Gemini. The output is almost always "encouraging", and rarely it formulates a critique, lists different approaches or formulates a competing view. It's sycophantic. Not as much as "Meta chat", or "Google AI" but the tendency is there.

Now I've started using Qwen3.8-27b (dense). The difference is stark: the reasoning style reminds me of some my Chinese colleagues: criticism and push-back, as a user with an idea I have real work to do, but it's easier to guide it to a specific dialectic in a Socratic dialogue.

Maybe Chinese companies, e.g., Alibaba, don't limit the usefulness of their free (open weights) LLMs.

If someone has some competence in a topic they appear to be much more useful than American models.

2/2

LLMs are a touchy topic, in particular on Mastodon. My approach to it is "LLMs are the result of technology *and* ideology. I'd like to understand the impact of both".

What is the psychological effect of using LLMs? What are epistemological processes that can be used, and how can an LLM be assessed if ideation and vetting ideas is the target?

The first thing, of course, is critical thinking skills of the user. Using an LLM for ideation without competence is like using LLMs for "Vibe Coding" - but as you don't "own" the architecture you won't be able to take any kind of responsibility for it. Users that apply a strict conservative Software Engineering approach can do much more, and they can take responsibility for the usefulness of the resulting software.

For ideation it's the same: if you can construct the idea in its context it's easy to use the LLM to help with formalization. If you have no clue you better only prompt for things that don't matter at all.

1/2

RE: https://beige.party/@CuriousMagpie/117310983525275522

@CuriousMagpie
As I see it "AI" wouldn't be nearly as damaging if it were not used to bolster individualized propaganda vectors: through social media. These existed before, say, 2022: Meta (Facebook, Instragram, now also WhatsApp), TokTok, Xitter (in no particular order).

Journalism was first embraced, then strangled. Before that the on-line ad industry had long created a damaging pressure on news outlets: the salient pays in the economy of attention. It had been long known that people who only consume TV are less apt to think analytically.

Another thing we observed was that more people than before gave up constructing, negotiating, a common, public, objective, truth: that became tribal, up to the extremes of Q-Anon and the like.

Today, what I observe, is that a divide-and-conquer strategy is applied on the common, and I don't see an easy way out. Anyhow, my view is that we can't see "AI" in isolation but it accelerates forfeiting common sense.

#journalism #truth #commonsense #ai

The day before yesterday I spent in Würzburg for important reasons. While I was there, I got hold of a book, a collection of papers, humanities, in honor of the 60s birthday of Winfried Böhm, a humanities professor, almost 30 years ago, 22/Mar/1997.

The book is fascinating. Really deep reflections on the problem (almost a paradox) of teaching someone to be free, and yet reasonable. The approach is simple, and at the same time deeply humbling: learn from history what it means to be human.

Schleiermacher plays a big role there.

#humanities #philosophy #pedagogy

Good morning!

#cycling #Germany #sunrise

I cycle a lot, like really a lot. Lately it changed how I perceive the world. I can also see how I used to perceive the world a few years ago when I didn't spend nearly as much time on the saddle.

I see many people in cars doing stupid, thoughtless, things. Sometimes I also observe stupid, primitive aggressive, things.

Both have in common that using cars is totally normalized in society. But clearly the car has something to do with that stupidity, it's not neutral.

Recently a word popped up in my mind: "Blechkistenmenschen", "sheet metal box people", not fully translatable.

For German natives concatenation has a meaning: it implies a category. In my book, categories are linked to "normalization" and it's not the "normality of cars".

The connotations are "Blechlawine" ("sheet metal avalanche") and, "Zombie Apocalypse".

The takeaway: as someone socialized in a car-normative world something flipped: the saddle is a powerful drug.

#cycling #cars

https://en.wikipedia.org/wiki/Motonormativity

Good morning!
#cycling #latesummer

And I thought that the heat during my ride on June 29 was bad. Today the top temperature was about the same but the average was 3° higher.
#cycling #summerheat #Germany

Someone in my family traveled to the Himalaya in the 1990s. This is a photo of a world that no longer exists.
#travelphotography #Himalaja #time

I don't write too often about my cycling milestones - but I guess the one today is worth mentioning:

I've reached the 20,000 km mark in this 208th day of the year.

That's 96 km/day, every day.
#cycling

From my morning commute this week.
#cycling #summer #sunflowers #Germany

Extended commute.
#cycling #summer #sunset

One more post about "Firefox AI":

In my opinion Mozilla has actually done a very good job. The communication problem, however, in the current hype cycle is **HUGE**.

How do you help people take responsibility for large language models (LLM) when marketing guys and corporate bots invariably use anthropomorphic robots or brain-shaped 1970s through-hole printed-circuit-boards when writing about "AI"?

What they call "AI" is "stochastic models on steroids". Even the smallest ones, e.g., Gemma-3-270m comes with 270 million parameters. It's a next-token-predictor that can produce, e.g., syntactically correct English. Quine's revenge on Carnap.

The tools behind LLMs are now such that the technical language of traditional stochastic modelling tools, e.g., STAN for analytical stochastic modelling, is no longer useful.

Turn the table on the hype: use the scientific language of statistics and epistemology and solve a real problem
#ai #firefox #mozilla #science
https://ai-guide.future.mozilla.org/content/introduction/

"Firefox and AI" has produced strong emotions. I've tested the latest version of the #Mozilla browser and made a few observations:

- the "AI" features are just a web page loaded as a "sidebar"

- the "AI" is a single-page web application which the user can give access to pages or selections

- if the single-page application doesn't meet certain implied UI patterns the feature breaks

- the Web-UI in llama-server works, the one llama-swap breaks it (as seen in the screenshot)

The takeaway is this:

1. the "AI" feature is useful, especially when used with a self-hosted RAG (Retrieval Augmented Generation) like Open WebUI, or a simple LLM on a local GPU

2. the page-scraping feature (from the screen-reader API) is good, better than what the extensions Page Assist uses

3. 3rd party extensions, like Page Assist, have far better LLM support (an API endpoint just works)

4. the real problem is trusting OpenAI, and the like, with the feature.

#privacy #selfhosting #homelab #surveillance

Evening in the Hohenlohe.
#cycling #summer #Germany