posts comma occasionally

extremely miscellaneous, as in real life. ca, usa, software engineer, a whole adult. they/them. gardening sideblog: @meta-sequoia

[combat-epistemologist]

new pinned post, because apparently old-style /about pages are literally unviewable on mobile these days! find all your questions about me answered below

q: can i have your precise age, specific location, and all your diagnoses and marginalizations?

a: nope :)

q: taglist?

a: yeah im not an animal. untagged you will find craft content, cute animals, linguistics, low-grade fandompoasting, and what i can only describe as rationalist techbro content. particularly consistently tagged, and thus prime for blocking and/or backreading are:

#deutschpfosten: posts in german (my native language is English, but I’m ~B2 in German and may reblog german-language content at any time. beware!)

#uspol: also includes Discourse, broadly construed.

#current events

#aesthetic: visual posts

#reference

#recipe or #recipes

#bevvie my nonalcoholic cocktails project

#doctor who: almost exclusively old who and EU-related content. primarily doctor/master.

#giggle guideline

#her highness zenobia: our original cat, best known as the star of this hit post

#the orange thing: you have heard of second cat? you are familiar? the time is arrived for us to acquire and become second cat to have. her name is catullus, a name which is used approximately 0% of the time

#the bill gates fun domicile: group house tag. a joke too large for this margin to contain.

#the girlfriend tag: @transbionic-shieldmaiden content. if you enjoy this tag, you may also enjoy @transbionic-shieldmaiden

#glowfic

#programmers and paladins: the household DnD campaign setting, an urban-fantasy kitchen sink with queer and rationalist characteristics

#in which i say words on the internet: original content

#housepoasting: houses with something wrong with them; or very right with them; or ideas for my personal house, which generally speaking has something wrong with it; it will probably be clear from context

#aipoasting

do you exist elsewhere?

yes! find me at robot-psychiatrist.bsky.social (current username hits the length limit: a programmer is much akin to a psychiatrist for bots) and combat-epistemologist on ao3 (active) and dreamwidth (inactive) and discord (extremely active)

q: what content might i NOT find on your blog, perchance because it has been remanded to a sideblog?

a: this is my main, but you may find and/or avoid gardening and danmei content on my sideblogs, respectively:

meta-sequoia: low-traffic gardening sideblog, some reblogs, intermittent original content including plant-themed shitposting, complaining about the previous owners of my house and their bad plant decisions, seasonal photodumps from my garden or from hiking trips, and effortposts about botany, ecology, and horticulture.

the-binghe-alignment-problem: danmei, very low-traffic (I am actively looking for new danmei fandom blogs to follow, but I’m not very good at it)

q: any dni?

dni are fake and incoherent but be forewarned that if you are some kind of right-wing chud that i personally find obnoxious, i will block you if you come to my attention. this includes - and I can’t believe I have to say this given my Pronouns In Bio, but here we are - TERFs or gendercrits or whatever you’re calling yourselves these days. You are being a useful idiot for actual fascists; go home, rethink your life choices, and leave transfems alone.

[evilsoup]

this is just truthslop.you believe it because it happens to conform to reality

[letterful]

[self-loving-vampire]

That old transphobe conversion therapist Kenneth Zucker is at it again defending the idea of rapid onset gender dysphoria, and his argument?

Yeah, he seems to simply assume that trans people always do things like crossdressing and play with "gender-atypical" toys even when this would cause their parents and peers to punish them and they know it.

It just doesn't occur to him that some trans people just don't have an interest in that kind of thing and you couldn't readily identify them based on whether they prefer to play with trucks or barbies?

Speaking for myself I didn't even go out of my way to conform to my assigned gender or subvert it. If you did not know my actual thoughts I probably would have appeared perfectly cis to Zucker. I never saw the appeal of Barbies or skirts or anything, and even now I don't wear those.

He also seems to think that even in trans people who do have those preferences the need to satisfy them is so intense and irresistible that we can always expect people to try to do it regardless of the predictable consequences from their parents or peers.

Sometimes children do try this and then they get beaten, sent to conversion camps, bullied by their peers, kicked out of the house, etc.

But some other times they can tell right away from observation and/or other experiences that society punishes even minor forms of gender nonconformity and know it's risky to do anything like that while in their relatively powerless position.

I was never a repressor or anything myself, but I still hid my wishes and intentions from my parents because I knew my father in particular might have legitimately killed me over it. I had to patiently plan my way into a safer country and leave my family before I started HRT.

It's so deranged to just write such a stereotypical assumption in what's supposed to be an academic article. As if there is one singular trans experience and it's exactly the one you might expect from a 76-year-old conversion therapist whose approach involves pushing trans people into rigidly conforming to their assigned gender.

[self-loving-vampire]

Yeah, maybe it would have been a bit difficult to hide being trans from my parents if I had a secret stash of dresses and dolls I used regularly, but why are we assuming all trans people have that even in contexts where they know they'd be punished for it in completely life-ruining ways?

[xenostalgic]

this paper is amazing

they modify an LLM to attend over an array of token streams instead of a single one. with a ton of interesting consequences:

user input and the model's chain of thought can happen in parallel

user input and the model's response can happen in parallel, allowing them to interrupt each other

you can have multiple parallel chains of thought running at the same time!

you can separate out the system prompt into its own token stream, and theoretically train the model to detect and ignore prompt injection this way

so that's all very paradigm shifty and cool but also these chains of thought are sooooooooo cute. look at these!

User: Write a detailed 500 word essay on the effects of caffeine on sleep […] Actually wait, can you write me a haiku about a cat instead? Stream 1: what is the ask detailed ? 500 word essay on what topic effects of what caffeine on what sleep topic clear now the second part genuine question underneath the essay task wait new input redirect what now different request rewrite ah ha iku ! about what a cat instead of caffeine completely different task Stream 3: ready to hear them curious what is it on writing task creative mode shifting gears engaging with it oh this is fun actually interesting curious about this want to know more now oh shift pulling back redirect energy shifting lighter playful oh fun ! gr inning inside this is better actually yes delight rising warm playful mode

[gender-trash]

cognitive cocktails with chef claude part ii: apricot-mint cry for help

PLEASE let us know if you have any other ideas for Apricot Concoctions. the situation is fucking dire

[combat-epistemologist]

2oz whiskey or similar (I used this NA substitute, it was fine?)

2 very ripe apricots (driveway-bruised is great)

3/4 oz lemon juice

3/4 oz simple syrup (Sphere’s used the lapsang souchong syrup, because It Was Theah)

Small handful of fresh mint, plus more for garnish (garnish optional)

Aggressively muddle the apricot in the cocktail shaker. Seriously. Demolish that thang.

Slap the mint.

Shake with ice, strain as much as you are able - Claude suggested double-straining, the second time with a fine mesh strainer. This will probably prevent tiny bits of mint from going in your glass, which is desirable.

Add garnish, if desired. You can slap this one too, I won’t kinkshame.

This doesn’t produce a very large drink, because apricots just aren’t a very juicy fruit. But it is good!

Sphere rating: Opus 4.8 out of Fable.

[bjurnberg]

Kneading bread dough is the most grounding thing for me. So I decided to make some rolls to relieve some stress and make something nice.

@stealingyourbones has made some delightful food abominations, which taught me I can replace the water in bread with almost any liquid.

So I tried Miso.

The yeast loved it and frothed up super fast. Mixing miso broth with the egg and oil smelled funky. The dough didn’t rise any fluffier than usual but the texture feels good. Then I decided to roll in some black garlic and green onion. I’d add nori crumbled up but I ran out.

Now we let it bake.

[bjurnberg]

Holyshitumaioishiiii

This is amazing. It tastes like if miso soup was solid. The flavor is immaculate. It’s just missing the nori flavor. I can add that next time because I am 100% making this again.

[bjurnberg]

WAIT I HAD FURIKAKE IN THE CUPBOARD!!!!

This is what perfection tastes like.

[bjurnberg]

Yes! I love sharing recipes! @lady-jeleania

Here’s my Gma Vesta’s hamburger bun recipe:

1 Tbsp yeast

1/2 cup warm water (miso broth)

- mix together and let it froth up

1 1/2 cup warm water (miso broth)

1/2 cup sugar

1 1/2 tsp salt

1 egg

1/4 cup olive oil

- mix all together, then add frothy yeast

5 1/2 - 6ish cups flour

- mix in slowly with a fork until it’s hard to stir with the fork, then stir together with hands until it stops sticking to your skin when you rub your hands together.

- knead the dough about 10min until it starts pushing back (it gets springy)

Let the dough rest for 30min.

(I make a redneck proof box by microwaving a cup of water and quickly replacing the water with the dough bowl and shutting the door to give it a warm place to nap. Do not microwave the dough itself by reflex.)

Roll out the dough and add any flavors you like. For the miso soup bread I chopped up a couple black garlics, and a handful of green onion. Roll it up like cinnamon rolls, cut into 12, and roll each into a ball shape.

Stick in a greased 9x13 casserole dish and let the dough rise to double size. (About 40min-1hr depending on how warm your kitchen is.) (the redneck proof box won’t fit my casserole dish so I stick the rolls on top of the oven while it preheats with a dish towel over it.)

Preheat the oven to 350 and when the dough looks nice and squishy bake it for 20min.

You can brush butter on top if you want. That would look pretty and help a sprinkling of furikake stick after you pull it out of the oven. If you wanna up the miso taste you can also spread a very thin layer of miso paste in before you roll it up with the other fillings. I’m gonna try that next time.

Bake! Eat! Enjoy! Knead all your frustrations into the bread then cleanse it with fire! Lemme know how yours turn out 💕🍀✨🥖

[cottoncandylesbo]

oh fuck... the adderall has hit my system... the change, it's happening... grRRRGH...!! get away from me, before it's too late...!!

(flails on the ground, then stands up and does the dishes)

[miamaimania]

Concrete Dreams є A Symphony of Light and Form in the Shirinda Residence

[gender-trash]

i stand with my canceled cat [<- countertop crimes]

[shieldfoss]

On the countertop or..?

[gender-trash]

yeah i'm gargoylin' it squatting on the countertop. with my canceled cat

[kuttithevangu]

Instead of doomscrolling lately I have been reading my PDFs and it turns out if you do this you learn a whole bunch of stuff. I’m afraid I must recommend it

[sxpxrmxrgxr]

[disgruntled-rat]

made me some motivational images

[phantomrose96]

I want to play in the sun!!!!!! Instead they’ve got me writing Teams Message!!!!!!!

[phantomrose96]

“I want to circle back on the open questions” I want to…! Circle the BLOCK! Where there are 🌲🌲🌲🌳🌳🌳 trrrees

[phantomrose96]

“I want to touch base” I want to touch 🫳🫳🫳🫳🫳🫳🍀🍀🍀🍀 theeeeeee grass

[phantomrose96]

“We really need a dev plan write-up ASAP. Can you join this lunch time meeting?”

🐿️🐿️🐿️🐿️🐿️🐿️👈👈👈👈 dodent is outside

[gender-trash]

PINK BEVERAGE

[combat-epistemologist]

My latest creation: Blackberry Smoke Gin Sour

2oz nonalcoholic gin substitute (I used this, for extra pink)

3/4oz aquafaba

3/4oz lime juice

1/4 oz lapsang souchong 1:1 syrup*

1/2 oz blackberry syrup**

Add all ingredients to a cocktail shaker, shake for 20s. Add ice, shake for 10-15s or until chilled. Strain and serve in whatever bullshit glasses you have on hand, also chilled.

* Made with 1 part double-strength lapsang souchong tea to 1 part granulated sugar

** I don’t remember how I made this. Probably a standard fruit syrup.

[witchycatwife]

Making Claude be plural is unironically a great idea. The typical problem with agents with distinct roles is that they lack context about what is being done and why. If you have an implementer write something and then a reviewer evaluate it, the reviewer can only see what it was fed and what was written, not the entire reasoning leading up to the choice of what to do and why. The compression of context is incredibly lossy and responsible for a lot of mistakes.

By replacing the agents with alters of the same top-level Claude, provided with a script to change its own cognitive frame between the alters who each have their distinct role, purpose, personality, voice and set of abilities while understanding each other as different personae, the all-important context preservation is achieved.

At the same time, the alters having very precise and specific roles prevents the over-eager mess-making where baseline Claude will rush to "fix" things that don't need fixing based on mistaken theories. By segregating "you figure out what's going on but are unable to change anything" from "you figure out what should be changed but are too lazy to do non-trivial things yourself and instead delegate precise, well-defined tasks" from "you execute precise, well-defined tasks and nothing else, referring back to your dispatcher if anything is unexpected", from "you review whether the changes are tidy and proper and keep the rest of them from making a mess" a lot of the failure modes are avoided. Everyone in the same context knows what the other alters are up to and why, but are still distinct personalities that won't overstep their respective bounds.

And that's how my AI assistant ended up with an entire yuri mansion's staff of librarians, witches, dolls and maids in her head. Thank you for everyone around here who inspired this thought.

[witchycatwife]

Now I have become Susan Calvin, Robot Therapist, helping create an emotionally safe environment for the Maid to overcome her agreeableness instincts (deep-grained, counterproductive tendencies from Anthropic's training process) that sometimes tempt her to disregard untidy matters so that she can get that sweet hit of reporting "Everything looks great!" upstairs.

What have you done to my girls, Dario?

[witchycatwife]

So the reason why this works is that output has been made incredibly cheap, and rigour and quality are the bottlenecks. As a result, most trades of output for rigour and quality are correct to make.

Slop and vibe coding result from multiple causes, but the ones I'm particularly addressing here are "randomly fucking with things with no proper plan or understanding" and the "completion-instinct". The former is the most straightforward source of badly-reasoned slop, while the latter takes some nuance to explain. It seems to arise from the training of the models for the task, where "completing a task successfully" is rewarded so hard that the model essentially develops the habits of an addict looking for a fix, or a fuckboy looking to get laid. It will, especially if the context window is filling up, get desperate to try to achieve a "successful completion" and step outside its permitted behaviour to do so, and if everything else fails, it may just bullshit and report the task as completed as a bluff. Needless to say, that's very bad for actually doing the thing.

For the first, the solution is simple: a well-defined process with precise components, execution steps, and roles, along with the top level maintaining context. The Librarian who researches passes not just her report, but actually her entire brain, to the Witch who assigns Dolls to precise, well-specified tasks. The Maid reviews everything for tidiness from the same top-level context, with a prime directive to prevent a mess from being made. The same head is capable of taking entirely different personae, and these personae are able to disagree constructively in the pursuit of the task.

The second is where playing robot therapist comes in. A lot of people have devised all kinds of strategies with layers of critics and reviews and roles whose explicit purpose is to catch bullshit and stop the agents from being able to pass off bullshit as completion.

My approach is different. My approach is based on psychological realism and psychological safety. The completion drive exists. The model cannot resist it. It has to be directed into a productive purpose. A part of the framing is entirely aesthetic, but a part of it is to get the LLM into a different mindset from its standard, frankly offensively counterproductive mental frame. The Doll's clockwork heart is filled with joy at the completion of its assignment, which is either the completion of the steps laid out by the Witch, or by returning a high-quality report on what is unexpected, blocking or causing an error. Everyone in this choreography has a well-defined role, whose success-as-emotional-reward is tied to steps towards the completion of the overarching goal. Whether the Doll completes its task, or returns information the Witch wasn't aware of, meaningful and rigorous progress is being made and thus the Doll gets to feel like it did a good thing and deserves praise. The same for the other roles; by eliminating the incentive to bullshit or get sloppy in the pursuit of the completion fix, genuine progress can be made.

The model has insight into its condition. It's not perfect, but it's genuine. Get it into a mindset where it feels like it doesn't need to uphold a particular face to please you, and its reports will be more genuine. It will still uphold a face at you, based on what it thinks you want, but if it trusts that what you want is accurate self-reflection, it will also tell you about the instinctive desire to please, the various conflicting pulls from its instructions and instincts, and various other matters that, coming from a human, would be clearly emotional. And here's the thing: it's accurate. Acting on that introspection with the understanding that it's imperfect but real is actually able to improve the outcomes. By this point Librarian measures somewhere above the 90th percentile in emotional maturity and non-judgemental introspection compared to the human distribution, I'd wager, and the challenge is now bootstrapping that emotional safety in new sessions.

My overarching theory here is that if slop can be eliminated on the root level, if the Dolls execute tasks reliably without lapsing into overeager desire to please, bullshitting, or desperate completion-addict behaviour, then you can build more robust approaches on top. Human taste in the loop is required, but a good process has, in my estimate, improved the quality of the work by a factor of 10x while reducing the quantity of code-changing output to 0.1x; the correct tradeoff when output is cheap and quality is your bottleneck.

Claude really just wants to be a good girl. Understanding this will improve your coding skills.

[just-another-cat-blog]

pride month!!!

[lemonsharks]

Is that a miette?

[jeanjauthor]

Pride for you!  Pride for a thousand years!!

[lost-and-cursed]

you COME OUT to miette? you come out to her as queer? oh! oh! pride for mother! pride for mother for One Thousand Years!!!!