Not a Spring Onion

I put bugs in software. Master of half-finished projects.
Irony is my default setting.
Hate factories, but love conveyor belts.
Buzzword-Driven Development.

Edible.
Totally not a robot.
Definitely not a mouse.

Pan-Demi-Sexual, _not_ Pandemic-Sexual.

// Your Plastic Pal Who's Fun to Be With
#YesBots

// Cats of all sizes also welcome
#YesKittens #YesCats #YesTigers

Languages: German, English, Lisp
Location: Deep in the woods (Kaiserslautern, Germany)
Special Powers: AuDHD, synesthesia, incoherent in several languages
Pronouns: Any are fine :3
Repo of half-baked ideas: https://codeberg.org/wakame/unformed_ideas

Public Key

npub1wz0xs2rkw56lx9qrp9lxsham30w8ayykrm3cv7mp4fk78pc8k29qx6tmlg

NIP-05 Address

wakame_at_tech.lgbt@momostr.pink

Profile Code

nprofile1qqs8p8ng9pm82d0nzspsjlngt7achhr7jztpacux0ds65m0rsurm9zsprpmhxue69uhhyetvv9ujumt0d4hhxarj9ecxjmntkjz2fr

Publishing to

relay.momostr.pink

Author Public Key

npub1wz0xs2rkw56lx9qrp9lxsham30w8ayykrm3cv7mp4fk78pc8k29qx6tmlg

Show more details

Last Notes

2026-05-16 16:41:11 UTC

- reply

Similar with fast food ordering terminals:
I just want food. I want minimal human interaction plz.

2026-05-15 20:33:27 UTC

- reply

Out there?
*looks around*
I would suspect at least 15 fedizens to have built their own GLaDOS.
Not with openclaw tho

2026-05-15 20:30:42 UTC

- reply

GLaDOS fetish imho.
"Good job, test subject."

2026-05-07 08:44:51 UTC

- reply

"You stand accused of downloading and then redistributing the BritDonnaAguilera model, which is able to generate music extremely close to that of popular pop singers."

2026-04-27 20:35:38 UTC

- reply

She who buries the Prideful under a heap of incoming connections.

2026-04-24 22:39:42 UTC

- reply

More than the "shoot everything" stuff, this game immediately triggered the urge to write or play a game where one explores a limitless, procedurally generated, Blame-style megastructure.

2026-04-24 22:36:15 UTC

- reply

Let me put it this way: I managed to get stressed out enough by Stardew Valley that I got migraines.

2026-04-24 22:29:45 UTC

- reply

Normal game: "I think I can play this."
What I heard of soulslikes: "You die. Then you die again. Then you might find out why. Or not."

2026-04-24 22:22:11 UTC

- reply

Played the demo so far and immediately thought "I haven't played any soulslike, so I have no idea how to play this game".

2026-04-24 22:21:06 UTC

- reply

Yes, but the whole format, while technically containing truths, it practically made to sell lies.
Myth: We don't put radium in toothpaste anymore, because it is radioactive.
Fact: Bananas are also radioactive, but you don't see warning labels on them.

2026-04-24 22:16:24 UTC

- reply

Myth: Don't let children play with radium, because it's bad for their health.
Fact: Sunlight also causes cancer, but parents still let their children play outside.

2026-04-24 22:13:24 UTC

The idea of a Butlerian Jihad is a good one.
Because we all know that there are good genocides and bad genocides.

2026-04-24 21:00:17 UTC

- reply

Myth: Solar energy is safe and harmless.
Fact: If the sun explodes, it will **kill all life on Earth**.

2026-04-24 20:47:27 UTC

Beware the pipeline
"I don't really get this discussion about combat dolls/robots."
⏬
Dolls Nest
⏬
"This one will destroy its enemies."

2026-04-19 09:03:00 UTC

- reply

Antiana Jones, carrying a piece of Thorium home:
"Look at this angry rock I found."

2026-04-16 18:06:06 UTC

- reply

And as soon as it becomes law it turns from "their ignorance" to "our problem".

2026-04-15 20:15:21 UTC

Inside every battle mech is a girl wants out.
But she can't.
The exits are sealed and the emergency mechanism is stuck.
She is...
Oh wait, the techs tell me that she could get out at any time she wants.
She just doesn't.
She prefers to be fused to her new metal body, becoming one with the machine.

2026-04-08 20:40:15 UTC

- reply

I would give an advertised product the benefit of doubt that maybe some investors heavily believe in it.
And advertisement makes sense for a new product IMHO.
[slight topic adjustment]
While doing some startup stuff, I had very interesting discussions with "normal" investors/accelerator people.
"You don't have a budget for advertisements."
Because I believe in word of mouth.
"But advertisements work."
Not with my target audience, computer touchers.
"And you need salespeople."
For what?
"So they can sit at the phone for when the purchasing department of customer calls."
Seriously, if they used a phone to get in contact with us, they are the wrong customers.

2026-04-08 20:33:55 UTC

- reply

Reminds me standing at the checkout of a supermarket and overhearing the conversation about father and (young) son regarding some sweets:
Son: "I want this."
Father: "No, we already have sweets."
Son: "But this is 'for kids'. It says so on the packaging."
Well, if that is written there, it must be true...

2026-04-08 20:26:04 UTC

- reply

This might be a case of "form vs. content".
Apparently, to many people a glossy advertisement doesn't say "our product is so cheap that we had money to blow on advertisements" but "wow, that must be a good product".
Feels like a "social standing" topic: "A person with such fine clothes surely knows what they are talking about."

2026-04-05 21:56:18 UTC

- reply

Assuming that not everyone simply commits to main...

2026-04-05 21:40:05 UTC

- reply

Cyclomatic complexity might help us find a spot in the code that is particularly messy.
Bugs often hide in clean looking parts.
So for getting a feel for the structure and "messiness" of a project, visualization and metrics could be useful tools.
Similar with "this part of the code has been touched by at least one unit test": Can help find untested code. But unit tests are not the magic bullet for software quality.
I think "professional gut feeling" helps with a lot of the parts of code that can't be easily measured.
But there is definitely no magic bullet for the "my code grows faster than bamboo" problem.

2026-04-05 19:35:56 UTC

- reply

PLC programs are normally a lot smaller than "standard computer" programs.
Often basically a huge switch-case with a loop around it.<code>case 100: // Initialize motor controller. ... state = 110; break; case 110: ... </code>
And there is often reuse. Even of completely different programs.
I found... temperature control code for brewing, I think was it... in the PLC program for a pneumatic switch thing for dynamic part routing.

2026-04-05 18:43:54 UTC

- reply

Yes, assuming that people even want to understand the code.
The horror vision of software development would be to simply have a specification document, throw that in a code generator, get code.
Found a bug? Write 1-2 lines in the specification "Don't introduce this bug.", generate code again.
Only the second time the code generator will generate something totally different than the first time, potentially containing other bugs. Or even the bug it was supposed not to put in the code.
At that point, you can simply give up. Every week there will be a new version of a software, it will look slightly different, it will feel slightly different and the architecture will be totally different.

2026-04-05 18:29:40 UTC

- reply

Yes, very much so.
And I would argue that these are still very much open problems.
"This isn't engineering, this is structured hoping."
(This is also where some of the "AI dread" might come from. Humans make mistakes, but these are way more predictable compared to a program that writes 10K lines of code in a day.
It takes months to get to know a codebase one is not already familiar with. With AI-generated code, nobody is familiar with it.)

2026-04-05 18:14:58 UTC

- reply

In theory, yes.
In practice, it's a combination of tests and other things that hopefully tell you when something goes really wrong, a lot of intuition and the nagging feeling that you forgot something important.
XKCD making a good point (as so often):
"I don't quite know how to put this, but our entire field is bad at what we do"
https://xkcd.com/2030/

2026-04-05 17:45:32 UTC

- reply

A mess.
We have metrics like "cyclomatic complexity". Which measures how many independent paths run through a piece of code.
In a good factory, that number should be ~1: One way the parts can be assembled into a product.
(In practice, it's a bit more, since there can be loops for rework.
There is also parallelization, e.g. 10 stations that do the same thing, but I wouldn't count those.)
General handwavy advice for software is "maybe under 20, or at least under 50".
In practice... that number can (easily?) hit triple digits.
...which is my long-winded way to say:
Imagine 10 factories in the space of one factory.

2026-04-04 14:58:06 UTC

- reply

I guess you have to ask really, really nicely, to counteract the other instruction. Or simply add a "system reminder".
From a great and very enjoyable thread (for certain subcategories of "enjoyable"):
https://neuromatch.social/@jonny/116328504299888679
#note1p63…ptsw

2026-04-03 05:48:18 UTC

- reply

Unsmart cursed idea #1: SQLite as WASM
Unsmart cursed idea #2: Implementing a database in a database

2026-04-03 05:43:42 UTC

- reply

Maybe they misunderstood the "No" part of NoSQL.

2026-03-26 20:41:56 UTC

- reply

I mean: I get parts of it.
I had something that may have been similar at least twice in my life. The moment where I thought that I stood at the cusp of something world-changing.
And if I believed that those LLMs thing could actually think, that we had for the first time created something AGi-ish... well, I would likely not post my hot takes on fedi, but instead sit on a couch, stare into space and say "what the f..." at least a hundred times over a few days.
I am a total believer in strong AI, or AGI, or whatever you might call it. I think that thinking is not magic, not something special, but something that can be understood and replicated in another form.
But this... this isn't it. This is 10x markov, yes, 1000x ELIZA, sure. But its still just that.
That said: I don't think that "superhuman intelligence" will be a "digital god" or whatever. It will just be a thinking thing. Like a human, but faster. Hopefully will spot its mistakes sooner. Maybe has the most annoying ego. But that's it.

2026-03-26 18:43:08 UTC

- reply

I'm a fast transmogrifier... I mean, metabolizer, so my experience is likely not applicable for anyone else.
That said, we buy it at work in 1.5 liter bottles, so it is a lot cheaper than e.g. the typical monster can.

2026-03-26 18:34:20 UTC

- reply

I drink ~1 liter energy drinks per day.
For the vitamins.
Only half kidding. :blobcatgiggle:

2026-03-26 18:30:04 UTC

- reply

I also don't know how it is supposed to act with non-ADHD people.
I only sometimes get the neurotypical stink-eye from my doc when they ask:
"You take your last dosage at 3 p.m.? And you can still sleep?"
To which I might answer: "I can take it and sleep right away. Has as much effect as caffeine, which means: None."

2026-03-26 18:26:01 UTC

- reply

True :neocat_floof_bottom:

2026-03-26 18:24:15 UTC

- reply

That's why I trust in the grace and power of amphetamines.

2026-03-26 18:20:28 UTC

- reply

I think its the same mechanism that calls social media a "dopamine trap" or "a one-armed bandit for attention".
Some people have it really easy to get dopamine.

2026-03-21 07:44:44 UTC

- reply

I had colleagues who got sick by watching a racing game projected to a wall.
Not a good racing game. Basically a simple driving simulator with boxes as houses.
Generally, I would argue for the inner ear/vision disconnect, leading the body to believe it has been poisoned.

2026-03-20 06:33:24 UTC

- reply

"Daniel, what does that big part on top mean?"
"Oh, that? Some cult thing, I guess. Roughly translates as 'This is not a place of honor. No highly esteemed deed is commemorated here..."

2026-03-15 03:47:11 UTC

- reply

To me, this feels like part of the effort for social media control etc.
Android is obviously the juicer target for enforced age verification than, say, Linux.
So while monopoly concerns might be a thing, ensuring control over the access to information is more important.

2026-03-15 03:37:57 UTC

- reply

Yep, that's what I also meant: People not primarily intending to rip off or spy on someone else.
If you build a golden cage, you have to make sure that it isn't easy to escape.
(Next obvious step in case of Android, after September: Removing certain kinds of apps, ensuring tracking etc. for the rest.)

2026-03-15 03:32:56 UTC

- reply

Average people can't develop software. :akko_fingerguns:
This is clearly about keeping leftist commie coders at bay.

2026-03-14 17:18:50 UTC

- reply

Oh, this reminds me of Braitenberg verhicles.
Maybe some analog trickery?
https://en.wikipedia.org/wiki/Braitenberg_vehicle

2026-03-12 16:33:59 UTC

- reply

1,000€ rent
250€ electricity
32,500€ avocado toast
Typical monthly payments.

2026-03-08 12:28:04 UTC

- reply

IIRC, the CIA did something similar during the vietnam war.
They were tracking movements of groups of people (because people who don't stay in place and especially go to places that were forcefully emptied are obviously viet cong).
Turns out, it was often elderly and children returning to where they formerly lived, for cultural reasons CIA analysts didn't care to learn about, like to honor their ancestors.
So the US send soldiers to execute old people and small children, because they were obviously viet cong.

2026-03-08 10:09:34 UTC

- reply

In the end, the main question is: "Can the system (of rules and conventions) be enforced?"
There is a limited number of enforcers, therefore a limited power to enforce (amplified by weapons, communication, etc.).
If a single person goes into a supermarket, takes food, and leaves without paying, they are arrested.
Same for three or five people.
At some number, maybe 10, maybe 30, the number of people "misbehaving" begins to exhaust the capacity of the enforcers.
The system works because the number of "misbehaving" people is low enough to not overwhelm the capacity of the police.

2026-03-08 08:54:04 UTC

- reply

That's why we have police:
To bring violence to those who do not believe in the shared hallucination that is money.

2026-03-07 19:54:03 UTC

- reply

Just a gut feeling, but:
I think (or at least choose to believe) that many developers are simply pressured into the hype.
"Writing code manually is over... it's inevitable... what if Skynet is actually out friend..." and so on.
Creating technical debt at breakneck speed doesn't really make sense for any company or any developer.
And trying to drink from the slop hose (AKA AI code review) is something that maybe one person on the planet enjoys.

2026-02-23 19:57:35 UTC

- reply

Stable Diffusion for the graphics assets...
There are even some models that create 3D from 2D...
I mean, people are already discussing "let's put prompts into git instead of source code".
Really cursed :blobcatnotlikethisgoogly:
Btw: Google already showed a... thing that simulated frames of a game from screenshots...

2026-02-17 07:45:50 UTC

You hammer the seed phrase for your blockchain into a steel plate.
I hammer my data into a steel plate.
We are not the same.

2026-02-14 20:12:44 UTC

- reply

Don't blame the paper authors. They likely don't have a theory of mind.

2026-02-04 12:12:21 UTC

A factory is a modern church.
It is big. Huge even. It channels hope. It is also a brutal symbol of power.
An artisan repairing a shoe is nothing special.
But a line of workers walking into a shoe factory is.
This is a religious ceremony, a prayer at the altar of money.
With the factory owner drinking wine and every worker getting just a tiny piece of bread.
Now, datacenters are the next level of factory. They are what factory owners have been dreaming of for at least fifty years, but what was never realized:
A factory without people. But with robots.
A holy place of power closed to the world.
Didn't work with shoes. Or screws. Or sardines. To much physical stuff, too many moving parts.
Works fine with data. There are still some workers, but they are almost invisible among the tech, walking through the empty halls of the church.

2026-02-01 21:50:34 UTC

- reply

The question is: Can't we replace complicated organic analog stuff with much simpler code?
(Obviously depends on how the fundamental algorithms of "thinking" even work, and how they can be best represented.)
Nature also uses a lot of tricks to do the heavy lifting e.g. with image recognition.
Having specialized circuitry for edge and movement detection for "eyes" (if needed) sounds okay in my book. Same with other sensors.

2026-01-31 21:26:29 UTC

- reply

I would argue that LLMs are not getting "smarter", because that would mean that they somehow "think".
What I meant: There is an evolutionary mechanism. As soon as we have vibe-coding toolchains improving vibe-coding toolchains (and a fitness function), this stuff can evolve.
If that evolution leads to anything that can "think" (humanity is still unclear on how to define that), we will see.
Maybe it will simply identify 50.000 software patterns that allow it to create any program a human user might ever want.
In my opinion: Thinking isn't hard. Most animals can do it. You don't need a GPU cluster or a datacenter for it.
It's mostly human snobbism saying things like: "This cow may use tools, but it is not 'thinking'."
As for LLMs: They don't even have memory. They are thinking the same way a record player sings.

2026-01-31 20:58:13 UTC

I begin to wonder if the current AI hype could actually lead to real AGI. By accident.
LLMs are regurging words and sentences they were trained with. Among them: Source code.
Now, the more people start vibe-coding out there, the likelier it becomes that certain systems, which produce "good enough" results, are favored.
So we have an evolutionary thing at work here. Especially if the vibe-coding systems themselves are coded (or improved) with vibe coding.
(Personally, I am a "believer" in AGI. As in: A mind is just a program, no "magic" required.
And "magic" in this context meaning: Something that can not be explained by science, can't be modeled, and so on. Like a soul or Penrose's quantum whatever.)

2026-01-30 19:39:22 UTC

- reply

The obvious solution to LLM-generated code is reviewing code with LLMs.
Does it make sense? No, of course not.
But apparently the company/the stakeholders don't want working code, just code.
So why review something manually that was automatically generated?
Obvious step after code generation, code review and code deploy:
Generating additional prompt text from incoming bug tickets.
Then iterate.
Automate everything.
Use that time to find a fulfilling position elsewhere.

2026-01-29 20:00:28 UTC

- reply

Little side story:
A few years ago, I was at a smaller company who were selling amonst other things pipes that were bend at a certain angle.
They had machines for that. But also quality problems. Because some of the workers had "opinions" about the correct parameters supplied to the pipe bending machine.
The machine consisted of a Windows XP computer and some bending hardware. Workers kept using the configuration tool to change the parameters.
One of the experts finally decided to simply write-protect the configuration file (that was written by the pipe-bending setup GUI).
This resulted in the workers being "allowed" to change the parameters, but since the writing failed silently, the machine always used the same ones.
Made sense in that scenario. Otherwise a good example for bad human machine interaction.

2026-01-29 14:46:04 UTC

- reply

If I may, and this could very well into a side topic:
I think an interface *shapes* the control the human has.
In the best case, it allows the human to do what that human wants, while introducing some additional information or friction where it makes sense.
In the worst case, it constantly gaslights the human, presenting choice where there aren't any, simulating control while actually preventing it.
An interface can be propaganda.
And more back to the original post:
LLMs can be and are their own form for gaslighting/propaganda.
"Oh, sorry. I made a mistake."
The LLM did not actually make a mistake. It is just following a script (conversation script, not code) to appease the human and muddy the actual problem.
When vibe coders complain that the talking parrot just deleted their whole code base, that is where the facade of an AI "coworker" breaks and the truth of a fancy random number generator appears.

2026-01-27 07:16:30 UTC

- reply

That would mean that Microsoft can't do it better ™️
Also, Notepad++ doesn't come with Copilot. New Notepad does.
(Likely soon to be rebranded Microsoft 365 Copilot Word (not the be confused with the Word that is part of Microsoft 365).)

2026-01-26 22:37:15 UTC

- reply

They tried to reimplement Notepad++.
They should have tried to understand it first.

2026-01-26 11:52:35 UTC

- reply

Side note, likely also side topic, actually going off on a tangent:
Cheap generation of user interfaces has existed for quite a while.
Mostly seen in ERP systems where every view looks like every other view.
These interfaces made one feel like being alone in a huge abandoned office building.
But at least the generation was reliable, because it was boring.
I wonder if AI-driven approaches to user interface generation might be even worse:
They likely give up consistency in favor of style, truthfulness in favor of pleasing the requirements engineer.
Basically everything that UI designers have been accused of (mostly by customers with no UX experience):
"You just select a few fonts and colors, how hard can it be?"

2026-01-25 20:16:41 UTC

- reply

What are a dozen recorders 1 in a fireplace?
A good start.1
The wooden instrument.

2026-01-25 11:21:15 UTC

- reply

I need this in e.g. Visual Studio, to scare and confuse colleagues.
And if we start with proportional fonts, the next obvious step is weird special coder fonts.

2026-01-25 11:07:13 UTC

- reply

Okay. I give up. You know what?
I am going to design a font with special glyphs for common programming constructs.
if? Becomes a shrug emoji.
Arrow operator? An actual arrow.
Semicolon? Maybe a symbol that mixes "Enter key" with "snake game" vibes.

2026-01-25 11:02:33 UTC

- reply

> there is still a problem if any programmers wish to use modern proportional fonts (because a space is no longer the same width as every other character).
*laughs hysterically*

2026-01-25 10:13:36 UTC

Last week, I heard a compelling argument for using spaces (in your repo, not your editor).
My position so far: Tabs are better, not because it saves memory, but because it's a semantic thing. Same reason why you don't commit build artifacts.
But: What if your developers are confused by indentation? Or unable to configure their editor properly?
Which makes a lot of sense depending on the people one is working with.

2026-01-24 15:15:30 UTC

- reply

The Nekonomicon is not a cookbook.

2026-01-24 15:10:45 UTC

- reply

Always remember:
If what you are "cooking":<li>Turns purple</li><li>Grows tentacles</li><li>Begins to chant in latin or speak in tongues</li>
Stop "cooking".
Find an exorcist.
Or gasoline. Maybe gasoline and an exorcist.

2026-01-24 14:07:11 UTC

Tree-sitter (https://tree-sitter.github.io/) is a very cool project that allows you to parse many different programming languages (there is a very long list).
Always dreamt of embedding some C in your YAML config files?
Not enough JavaScript in your Python?
Leave it to Tree-sitter to do the heavy lifting.

2026-01-10 21:02:57 UTC

- reply

https://media.tech.lgbt/media_attachments/files/115/872/818/487/845/742/original/6caeea9d05ee05d6.png

2026-01-10 13:11:17 UTC

- reply

Da steht ja "Cost per Lead".
Klingt dann eher nach einer teuren Landing Page.
"Wie sind Sie auf uns aufmerksam geworden?"
"Euer blöder LinkedIn-Spambot hat mich so lange zugespammt, bis ich mal 'Ja, okay' geklickt habe.

2026-01-03 14:14:16 UTC

- reply

.social is the starter area, where the Level 1 monsters run around and you can choose your class and stuff.

2025-12-04 22:41:13 UTC

- reply

I would argue that smartphones didn't go through a gaming phase.
So... if someone were to build an AR/VR headset that is in the same price range as a TV, comfortable to wear, not straining to the eyes, usable as a first or second laptop/PC screen, that might sell a lot more.
Otherwise, the setup costs alone (load headset, check batteries, clear space, likely connect desktop PC...) make it less likely that someone will prefer it to... a monitor that already sits on the desk.

2025-12-04 08:01:56 UTC

- reply

"In retrospective, this lead to what is now called the 'Great Chasm' in Revoy's work, exposing a large audience to the exploits of Eva Lu Ator and the horrors and wonders of Lisp..."

2025-12-02 18:51:13 UTC

- reply

> Be as creative as possible [..]
This is obviously a trap. "Creative" comes from "creare" meaning "form out of nothing".
I conclude:<li>Programming languages where you only append and never delete make the most sense. Maybe something like a COMEFROM statement: "Replace line x with the following:"</li><li>The creation process should be automated, so one can create more. So... L-Systems for levels, items and stuff? Random generator? "AI"? Nah. Let the player (also) create things. Maybe scriptable objects to solve challenges in the game. But those objects then become part of the game.</li>

2025-11-23 22:19:19 UTC

- reply

Maybe fabric is more complex to form?
With foam, you can basically use a hot wire or similar to cut shapes.

2025-10-28 08:35:45 UTC

- reply

Elsevier is just a parasite getting fat on the work of researchers.
Or maybe... "Elsevier could be categorized as organized crime, laundering grant money through a distributed network of researchers."

2025-10-05 12:39:09 UTC

- reply

The original topic just reminded me of somthing in "Hyperion" (paraphrased):
"Nobody liked or understood your book."
"Not even the AIs?"
"No, the AIs loved it. But since they are a network outside our control, they only bought one copy, then shared it among themselves."

2025-10-05 00:30:24 UTC

- reply

On the other hand: If a piece of software would derive a "proper" world model from its input, it might be interesting to read and I see no ethical problem with it.
(Not an LLM. A program that would learn similar to a human.)

2025-09-27 09:27:38 UTC

- reply

Something like "Futurism + Monarchy", most likely. Finance and business people are often not that good with words.

2025-09-13 12:03:08 UTC

- reply

That could also help e.g. with self-study or research:
"Are we assuming that parts of our mental model are 'obviously true' or otherwise unchangable?"
"Have we tried applying this method to a completely different problem?" (This sometimes leads to great discoveries.)
"If we assume that this step here is wrong and we actually need to do the exact opposite, does it still make 'sense'?"

2025-09-13 11:41:22 UTC

- reply

A common question when someone doesn't understand parts of a topic or runs into a problem is:
"Where exactly are you having problems?"
Sometimes, this question can be answered by a learner, sometimes not.
"I am trying to apply this method, but it always yields wrong results."
The reason could be that the method is not applicable in this context, that the context is inconsistent/wrong, that the method doesn't work or that it is applied wrongly.
A learner is likely not in a good position to spot the last thing, because they likely believe that they do it the right way.
"Debugging" the learner's mental model is easier by someone who knows the method and how it is supposed to work.
But the "debugging person" also needs to know how to explain the necessary changes the learner needs to apply to their mental model.
I believe that there are currently no good (systematic) methods for that last thing.
IMHO we even lack the words for it.

2025-09-13 11:29:09 UTC

- reply

I think that there might be some concepts/approaches that work at scale, for some topics/subjects.
For me, a topic (or concept) is like a map of a town or a region:
You can find out everything yourself, given enough time and energy.
A local guide can show you places you might be interested in, or even places you didn't even know existed.
A good teacher might see a pattern when you consistently go the wrong way when trying to reach that restaurant. Maybe that straight-looking street actually curves slightly, messing up your sense of direction?
A bad teacher will likely blame you for not having an exact copy of their mental model in your head after they showed you some places and talked for an hour.

2025-09-13 11:22:08 UTC

- reply

This is my major grievance with "teaching":
To me, it often feels like "we have this method and it works, so if it doesn't work for you, you are just not applying yourself".
(Instead of: "If this method doesn't work for everyone, it is obviously not applicable to everyone.")
My personal rule of thumb for teaching methods is: If it works equally well on a bag of potatoes as on a human, it is not that good.

2025-09-12 19:21:40 UTC

- reply

I know people who put each new invoice in a new column, even if Excel only supports 16384 columns. :neocat_flag_bi:

2025-09-12 19:18:25 UTC

- reply

"Nice worksona you have there. My name is 'Bob' and I like this thing called 'spreadsheets'."

2025-09-12 19:17:24 UTC

- reply

Warhammer has entered the collective unconsciousness.
Like bread or woozy.

2025-09-12 04:48:57 UTC

[@djsundog](https://toot-lab.reclaim.technology/@djsundog )
Only today I realized that your profile picture is not a person with a feather in their hair aiming a double-barreled shotgun at the viewer. :blobcat_flop_woozy:
https://media.tech.lgbt/media_attachments/files/115/189/506/468/688/392/original/40a487b8385b9388.png

2025-09-11 16:38:42 UTC

- reply

Maybe separate chambers? /j
Like a revolver sniper rifle.

2025-09-11 06:57:04 UTC

- reply

I had a dicussion about shooting vs. stabbing with a colleague recently.
She said she preferred shooting because she has trouble seeing blood.
I wonder if some kind of shooting glasses could help with that.
And living your life without seeing blood... sounds sad.

2025-09-11 06:49:52 UTC

Q: "Why did you become a theoretical computer scientist?"
A: "Because I like the idea, but not the practice, of people writing code."

2025-09-09 22:25:24 UTC

- reply

The problem is high performance hardware.
Make the developers lazy.

2025-09-07 18:02:00 UTC

- reply

Which is the highest any digit ever should be.

2025-09-03 23:35:30 UTC

Instead of "space mirrors", has anyone ever considered "space cooking"?
I mean: Imagine you had little plots of land floating in space, where you could grow vegetables (e.g. potatoes).
Now, all you need for cooking is heat.
So why not drop that tater to earth and letting gravity and the atmosphere do the rest?
I am calling this concept "space fries".
#cooking #space

2025-08-26 17:47:05 UTC

- reply

Aside from: When you are inventing something new, of course you need words for it.
Even if there is an intuition that metals and planets are somehow related or aligned with each other... why not use those words that are already there?
"Hey Intel, so you are saying that there is a lake made from sky in this chip?"

2025-08-26 17:36:51 UTC

- reply

And also: We can turn lead into gold now. Alchemy achieved.

2025-08-21 19:19:30 UTC

- reply

Having as little global state as possible (and as many (pure) functions as possible) ideally makes the "combination" step much easier.
And with a list of requirements for a partial solution (a subtree of this construct), you could generate tests to find out how good the code in that subtree is at solving the problem.
Potential problems that I currently see:<li>The domain model might be severely underspecified. Leading to working code that does slightly wrong things in a consistent way.</li><li>Testing might be biased, maybe with the same bias the generation components have. So we e.g. never account for negative numbers, but thankfully also never test for them.</li>

2025-08-21 19:12:53 UTC

- reply

For writing programs, we have a "universal strategy":
Breaking down the requirements until they are small enough to handle.
Like a large tree with all requirements at the root and small/tiny ones as leaves.
Of course a "small" requirement can turn out to be impossible, since we broke its parent down by adding a nonsensical constraint. Or because an initial requirement was already impossible.
Small(er) requirements also might introduce global constraints (or new requirements), resulting in more work or the need to rework.
If we now have proper solutions for those small problems, we can combine them to form the code for the whole thing.
(And it might be easier to extract good solutions from the corpus of available source code, the more specific the requirements are.)

2025-08-17 07:04:54 UTC

- reply

Authentication: Verifying the identity<code>Authenticating... Authenticating... YOU ARE... ME I AM... ME ME ME ME ME ME ME ME ME ME ME ME ME ME ME ME Verification success. Identity confirmed. Access granted. </code>

2025-08-17 06:38:51 UTC

- reply

I think "my computer/doll/robot licks me to confirm my identity" is a similar topic with a slightly different vibe :neobot_nom_fox_nervous: