I am so tired of people in engineering gushing about AI.
A bunch of developers that wouldn’t even know how to speak about natural language processing are pretending that writing some instructions is gonna always work in a predictable way because…
Because they forgot that they are usually the AI that needs to transform natural language instructions into a lasting actionable and acting knowledge. Because they forgot how fucking hard that is. Because bullshitting beats admitting ignorance…
I wonder when ChatGPT starts asking for donations. “You know, if you find me useful, or you appreciate me treating you like a predestined meta human, please consider giving me some extra financial support this month.” Maybe the failed model push will inadvertently normalize investment in “free” or shareware, open source support.
Learning about the upcoming git-history(1) command alone would have been worth this year #FOSDEM trip. It's heavily inspired by #Jujutsu UX and aims to implement in native #Git its most notable commands.
Tangentially related (to other parts of this email), there's already an implementatin of git absorb github.com/tummychow/git-absor… and it is a very useful feature indeed!
absolutely unsearchable! I had no idea it was a thing either before the talk. I suggest checking out the recording of the talk when available (main track, Jansen, the future of Git something something).
About that, "We built a web browser from scratch with AI," claim…
> When AI 'builds a browser,' check the repo before believing the hype. Autonomous agents may generate millions of lines of code, but shipping software is another matter. go.theregister.com/feed/www.th…
> "… I agree this isn't just wiring up of dependencies, and neither is it copied from existing implementations: it's a uniquely bad design that could never support anything resembling a real-world web engine."
> "Ignore the rhetoric that tech CEOs spew onstage and pay attention to what the researchers that work for them are finding, and you’ll find that even the AI industry agrees that the tech has some fundamental limitations baked into its architecture."
NOTE: This continues the bad practice of referring to 'generative AI' as 'AI'. Not all AI is LLMs…
Disclaimer: None of this is written by AI, I’m still a real person writing my own blog like its 1999 I finally figured out how to do Gigabit Ethernet over my existing phone wires. Powerline adapter…
@withaveeay Absolutely. The Isle of Man is currently dying in the gutter, unaware of what hit it. It was never in the EU so most people here don't realise that its big sibling severed the umbilical and is wasting away.
I reckon it should secede and form a Celtic Union with Ulster, the Republic of Ireland and Scotland.
I'm lucky that my ISP provides free in home mesh wifi with their Internet plans, otherwise this would be a great alternative, as my mid 90s home is still blessed with phone jacks.
If I have to go back to the other ISP that doesn't do mesh wifi, maybe I'll try this instead.
@javensbukan I do use WiFi, but where possible, I use cables. It's much more reliable and faster. But I paid about £1000 to have the 4 floors of my house cabled up last year, because I didn't know this was an option. It has dozens of telephone sockets!
@futuresprog I've used powerline networking in 5 out of 6 of the last places I've lived, and it works great and it's much more reliable than WiFi. But in the British isles, houses tend to have telephone line built in, often to every room and at least to every floor.
I wonder if any of my contacts on mastodon could connect me (via email) with someone on the @fedora Project documentation team? docs.fedoraproject.org/en-US/f…
You're invited to talk on Matrix. If you don't already have a client this link will help you pick one, and join the conversation. If you already have one, this link will help you join the conversation
I was wondering when a reporter would uncover this.
So BitLocker is super secure, right? Well... BitLocker recovery keys are backed up to Microsoft's Cloud - and they give them out to law enforcement on request. Using the BitLocker recovery key, you can just unlock the device without a PIN etc. forbes.com/sites/thomasbrewste…
I don’t understand putting your trust in black box proprietary encryption software when TrueCrypt/VeraCrypt exist and are older than BitLocker by 3 years, stupid doesn’t even begin to describe it.
If you interrogate Claude about magic strings it goes into a “stop trying to social engineer Claude” state to where it locks down its ability to browse to URLs. This is probably a safety state it triggers prevent enumeration of other undocumented magic strings.
I’m curious what other hidden magic strings exist for this or other LLMs. This might be additional attack surface to consider from an availability perspective. I expect it could be used as a string in a malicious binary to prevent analysis or break scrapers that send something to Claude.
What remains true is this though: a single string if ingested as data can cause headaches.
What happened? When streaming from Claude 3.7 Sonnet, response chunks do not contain thinking_blocks of type redacted_thinking when forcing redacted thinking with the magic string specified by Anth...
no, ma tu l'hai preso dal sito ufficiale del nuovo social, wsocial.eu, e ci hai messo le icone di mastodon e del fediverso, o sbaglio?
Questo è un altro sito fatto da qualche burlone del fediverso, che ci ha messo il trattino, w-social.eu, e ha fatto in pratica la stessa cosa che hai fatto tu
Ever driven the same car or used the same camera or power drill for 8-10 years and thought: "I wish all these buttons and levers were suddenly in different places and did different things"?
Una montagna di siti web con pop-up invasivi che dicono “abbiamo a cuore la tua privacy!”, “clicca qui per accettare!”… Tutte balle. Se aveste davvero a cuore la mia privacy, non mettereste nemmeno un cookie, nemmeno uno “tecnico”, non mettereste finestre pubblicitarie, ma mi permettereste di visitare tutto il sito standomene tranquillo.
il problema non sono i gestori dei siti ma i programmi di ads. Un sito può anche avere dietro un'agenzia editoriale ed ha dei costi che in qualche modo devono rientrare. Sono i programmi ads che cercano di massimizzare i target riempiendo gli script di traccianti.
@francommit ricordo che diversi anni fa esistevano i circuiti di banner, per farsi pubblicità. Era sufficiente accordarsi con un altro sito simile al proprio per uno scambio banner e si era a posto. Certo, le cose sono cambiate, ma lo trovavo un approccio molto più “sano”. In effetti ora un’azienda dovrebbe affiliarsi ai propri concorrenti, cosa ovviamente non voluta.😅 Ma basterebbe affiliarsi ai siti delle riviste del proprio settore, per esempio. Pubblicità gratis per tutti.
When Le Guin talks about genre writers as “the realists of a larger reality” we surrender the power of that when we narrow our work to only depict one type of future. We have great power to restore…
I think we may have been led here by imagining too many dystopias. Billionaire #scifi fans seem to have taken them as guidebooks, while the billions of the rest of us may not have imagined enough about our power to create better alternatives
LOL. FIFA - who infamously awarded Trump a made-up Peace Prize, is in the throes of FAFO: "Tens of thousands of fans cancelled their tickets to the 2026 World Cup over the weekend amid safety fears, political protests, and human rights concerns in the U.S., leading FIFA to schedule an emergency meeting this week." ticketnews.com/2026/01/nearly-…
@amministratore non trovando il "box suggerimenti" provo a mettere qui un paio di cose che a mio avviso andrebbero sistemare su owncast.it per evitare fraintendimenti e non dare adito a chi potrebbe male interpretare i vostri intenti.
Lo faccio in pubblico sia perché non vedo indicato un contatto mail per owncast.it, sia perché essendo un argomento che anche altri hanno sollevato magari possono sfruttare il thread per altri suggerimenti da darvi 😊
1. Disclaimer (ben visibile) che indichi che NON siete Owncast e quale sia lo scopo di owncast.it. Qualcosa tipo "questo sito non è in alcun modo affiliato al progetto Owncast (con link), ma ha lo scopo di diffondere e coltivare la community italiana che lo utilizza". (ovviamente sapete meglio voi come/cosa scrivere).
Questo non solo aiuta a chiarire lo scopo della pagina (evitando fraintendimenti e confusione), ma fa sì che interessatx possano andare a dare un'occhiata upstream.
2. Crediti. Oltre a creditare il progetto stesso (ma l'abbiamo smarcato al punto precedente), ha senso dare il merito a cose come l'Emoji Wall che linkate. Anche qui manca un link diretto o un qualche tipo di spiegazione/chiarimento anche solo sul fatto che sia qualcosa di non vostro, ma creato e mantenuto da altre persone.
Una sezione crediti sarebbe la cosa migliore, ma anche banalmente un paio di link in più migliorano un sacco la situazione.
3. Sarebbe carino che i testi non fossero scritti da AI, ma se decidete di farlo occorre una piccola review per refusi come "Certo! Ecco un capitolo strutturato con le best practice per promuovere uno stream su Mastodon (ad esempio tramite Owncast.it) nel contesto del Fediverso:" (Conclusione al termine del paragrafo OBS).
4. C'erano anche altre cosine, ma vedo che avete iniziato a sistemare un po' (rispetto all'ultima archiviazione su archive.org almeno) la pagina aggiungendo il link alla directory ufficiale e a dove scaricare owncast.
Non penso sia sufficiente e non credo copra i punti sopra indicati, ma si apprezzano sempre le modifiche in positivo. Se posso aiutare a sistemare la pagina sono assolutamente a disposizione, non sia mai che si critichi senza aver voglia di contribuire.
@cicciofritz Stando lontano dal PC mi stavo perdendo tutto questo dramma ^^ Risposta veramente da signore, che oltre a farmi venire voglia di mangiare i casoncelli mi dà la possibilità di postare per la prima volta nel 2026 questa immagine ^^
Il sito nasce come raccolta di anni di appunti, configurazioni e suggerimenti, poi impacchettati e riscritti per essere comprensibili alle persone normali. Il fatto che oggi Google lo mostri come secondo risultato cambia tutto: è evidente che vada ripulito dai loghi ufficiali e linki a risorse utili necessarie per aiutare i curiosi che capitano per caso da google in quanto è l'unica documentazione italiana sul progetto. Oltre ad aggiungere link per supportare il progetto economicamente.
spero gli appunti qui sopra siano serviti a darvi un paio di spunti su come evitare di far cadere in errore un navigatore distratto che leggendo quella pagina ne comprende male il senso e fraintende le vostre buone intenzioni. Se vi va fatemi/ci sapere quando è aggiornata che un'altra mano la do/diamo volentieri!
ps: vi anticipo e vi dico che no, non serve uno screenshot del mio toot dove mi ringraziate per tenere pulito il Fediverso 😊
Detto questo, al di là di refusi o possibili dimenticanze, senza alcuna offesa, la valutazione sulla correttezza del sito spetta solo ed esclusivamente al team di Owncast, cui compete anche esprimere un giudizio sull’iniziativa ed eventualmente chiarire eventuali ambiguità se presenti.
Per il resto il servizio nostream non è un nostro canale ufficiale ma un servizio che viene usato da chi ne fa richiesta, può capitare che qualche admin capiti come ospite live ma non è mai utilizzato da noi.
assolutamente nessuna offesa, come detto sono suggerimenti da persona che è capitata su quella pagina: non sono la polizia del fediverso o altro, ma in uno spazio aperto come questo trovo sia utile potersi aiutare a vicenda a crescere.
un'ultima cosa, non pensi/pensate che per onestà intellettuale sarebbe il caso di sistemare un attimo il toot sul vostro profilo? O secondo voi il mio comportamento resta un attacco mirato a sparare fango invece di una critica costruttiva?
A volte è bene ricordare che i consigli si danno senza ricorrere a post indignati utili solo a creare flame e a trascinare la propria community a urlare “parassita”. Il fatto che quel commento sia rimasto visibile conferma che non si tratta di un incidente, e francamente non sorprende. Evidentemente, su livello segreto il confronto lascia spazio all’insulto: va bene così, è la dimostrazione del livello dialettico dell'istanza, e non è manco la prima volta che succede.
Now is a great time to force-feed your CTOs on the writings of Niklaus Wirth (inventor of Pascal, Modula-2, and Oberon among other things, and proponent of "less is more" in software—he wrote an entire operating system with GUI in about 4000 lines of code.). mamot.fr/@pluralistic/11584856…
Code is a liability (not an asset). Tech bosses don't understand this. They think AI is great because it produces 10,000 times more code but that means it's producing 10,000 more liabilities. AI is the asbestos we're shoveling into the walls of our high-tech society:
If you'd like an essay-formatted version of this thread to read or share, here's a link to it on pluralistic.net, my surveillance-free, ad-free, tracker-free blog:
@jackwilliambell @ravenonthill Pascal was missing a few things it really needed (the variable-length string as a fundamental type that Borland added via Turbo Pascal would have been perfect) but it was a really good first step, and solid enough that OSs were written in it (LisaOS and early MacOS among others).
@jackwilliambell @ravenonthill Back in the day, at least one model of X.25 packet switches from Hughes were programmed in Pascal. This was when Pascal was the big teaching language. As it was explained to me when I asked "why not C?", Pascal was fast enough to meet the requirements and they could hire new grads who knew Pascal without having to train them in C.
The FAA's NADIN II packet switching network (long since retired) used these switches.
@jackwilliambell let's be honest: the absolute majority of 'developers' these days can't read K&R and would burst into flames upon being presented with Perl. Brooks, Wirth, it doesn't matter, they'll refuse, fail to comprehend, and insist that it's too old and only they have the brilliance for modern programming.
Met a nephew at Christmas. He did a software engineering course at university and had just got his first graduate job and he was working mainly with Pascal.
"One of my most productive days was throwing away 1000 lines of code" - Ken Thompson.
I saw that handwritten in red pen on Ken's review copy of the first 8 chapters of the manuscript of The Art of Unix Programming which he mailed back to Eric in 2003.
The quotes in catb.org/esr/writings/taoup/ht… mostly came from Eric having sent (physical printout) review copies to people, who scribbled margin notes on them which Eric then incorporated verbatim. (Apparently a thing publishers did back then.)
@jackwilliambell @ravenonthill "OSs were written in it" is not necessarily a good metric. PR1MEOS was mostly written in early style FORTRAN and I don't think BLISS should be mentioned in polite company.
Pascal has some interesting limits but then so did early C, Just PASCAL tended towards "you can't easily do that" and C tended towards "thats dumb, but if you aim at your feet and pray go for it"
on a consulting gig where employees were evaluated by how many lines of code they checked in (and thus everyone copied lots of code instead of calling functions), I asked the manager, "which employee is smarter, the one who accomplishes his task in 10,000 lines of code, or the one who accomplishes the same thing in only 1,000 lines of code?"
@seachaint @elmar @rootwyrm @jackwilliambell Perl since the late 90s had regexps that supported inline whitespace and comments so they were self-documenting and highly readable.
@elmar @rootwyrm @jackwilliambell Yea, the old joke about "write only Perl" was always unfair, because Perl so strongly emphasised regexes and they're responsible for a lot of the reputation. And I agree that the real blame out to lie with "write only developers" :)
@seachaint @rootwyrm @jackwilliambell Perl was never a write-only language but lots of write-only developers were using Perl in its day. Now lots of the new breed of write-only developers use JavaScript and it shows.
Neither Perl nor JavaScript are "beautiful" languages out of the gate, but both are powerful enough that with discipline they allow for well-structured, maintainable code. One core tenant of the "less is more" argument is that you should understand all of the code you write and hence it must be small enough that this is possible. AI defeats that goal unless you use AI only to teach you about programming. In my experience this is ultimately more work than learning directly from the manuals.
With old languages (C, Perl, JavaScript, ...) it was hard to be disciplined enough to write maintainable code. Newer languages (Go, Rust, Swift, ...) were developed to help with that (and they do!) but it is depressing to see that so many projects are writing sloppy code while pretending that the new languages automagically lead to "good" software.
(btw. I went through my education at ETH but am just too young to have been to lectures by Niklaus Wirth himself. However, his influence still showed through his disciples and the curriculum.)
@seachaint @elmar @jackwilliambell yup. There were ISPs up until about, oh, 4 years ago that ran their entire provisioning system on Perl. That was it. Just Perl.
@rootwyrm @seachaint @elmar @jackwilliambell Perl is dying out slowly because Larry made *one* stupid decision in the late 1990s—to call his follow-on language "Perl 6", thus blocking iterative development of Perl from using a recognizably larger version number. Thus rendering it pitch-deck hostile, rather than actually unfit for purpose. (The former Perl 6 is now known as Raku—it's the C++ to Perl 5's C—while Perl development is still ongoing, in its own direction.)
Hey Guys, gals, people and pets, Satya Nadella is getting tired of people calling AI Slop output AI Slop. If everyone can stop that's be great as it is upsetting Microslops share holders.
Don't start calling Microslop Microslop or use the #microslop tag as it might burst the AI bubble and then Microslop might have to take SlopPilot out of Windows. As you know, that would be terrible.
We should do as Microsoft's CEO wants and stop calling AI garbage Slop. Instead we should call it Nadella, to show him our appreciation of all he's done for us!
In the 1980s there were an incredible number of personal computers of all shapes, sizes, and operating system types, and there was very little interoperability. Unlike today’s Windows-Mac duo…
A few days ago, I saw a nice approximation to the arcsine function by @eniko
asin(x) ≈ a(x) = x + (π/2-x)(1-√[1-x^2])^2
I won’t bother plotting a(x) against asin(x), because the two curves are indistinguishable to the eye, but the proportional error is plotted below.
I spent some time trying to figure out if there’s an underlying *geometric* reason why this works so well ... in the same sense that there is for, say:
asin(x) ≈ chord(x) = √[2(1-√[1-x^2])]
the length of the chord that subtends the angle asin(x).
There are some obvious nice features built into a(x): it clearly must agree with asin at x=0 and 1, and less obviously it will match derivatives at those points as well.
But surely there had to be some special geometric relationship too, to make it work so well?
If so, I never did find it. Maybe someone else will (or already has). But I found another approximation, roughly as simple and roughly as good:
asin(x) ≈ b(x) = ½(π-4)x^2 + x + 1 - √[1-x^2]
which also matches values and derivatives with asin(x) at x=0 and 1, and whose proportional error is the gold curve in the plot below. [Note that it matches values with asin exactly at an additional, intermediate point, asin(1/√2) = b(1/√2) = π/4.]
That partly cured me of my conviction that there had to be a nice geometrical account for any approximation this good. Maybe all that’s really needed is a low-degree polynomial and one function, √[1-x^2], with an infinite derivative at x=1 the same as asin(x).
@futurebird If I were at a computer, I'd be extremely curious to compare this implementation against Skia's asin approximation. They look similar but Skia has a much more involved polynomial in the middle.
I retract that. It "IS" Cecil Hastings "sheet 35". Couldn't find it on RAND's website but it's in "APPROXIMATIONS FOR DIGITAL COMPUTERS" 1955, page 160
That approximation is almost three orders of magnitude better than the ones I described as a(x) and b(x)!
Though they choose *not* to make it exactly match asin at the endpoints (as you can see from the error curve they plot for the function ψ), which leads to huge *proportional* errors close to x=0. If that was a problem, it would be easy to tweak it, making a_0 in their polynomial exactly π/2, which would roughly double its maximum absolute error elsewhere.
keep in mind that as @eniko pointed out there are exponent to the c=1-sqrt part that make the curve fit even better. We did some testing and it turned out to peak around 1.93 for some reason (1.9296...). I wonder what *that* means.
[Codeolences] tells us about the FORBIDDEN Soviet Computer That Defied Binary Logic. The Setun, the world’s first ternary computer, was developed at Moscow State University in 1958. Its troub…
In programming, we have a nice pair of opposed acronyms:
• DRY for "Don't Repeat Yourself" • WET for its opposite, "Write Everything Twice" (or "We Enjoy Typing")
But there's an intermediate position. The benefit of DRY, other than brevity, is that if a thing is specified just once, the specifications can't get out of sync with each other. If you can't manage that, the next best thing is to make sure the compiler or test suite _checks_ that they're in sync. You have to do more typing than you'd like, but at least you've removed the risk of an accident, which is the _most_ important thing.
(For example, in Rust, if you add a new branch to an enum and forget to update one of its match statements, the compiler complains about the one you missed.)
I feel as if there ought to be a nice intermediate acronym for that state of affairs, so you can say "Weeell, it's not as DRY as I'd like, but at least it's only MOIST." Or DAMP, or HUMID or something.
"Match Or Interpreter Spots Trouble"? "Disallow Almost-Matching Programs"? "Holler Unless Many Instances Dovetail"?
Not sure about any of those. The last one in particular seems especially "you resorted to a thesaurus, didn't you?".
i feel we could also use an acronym for "these two things are not the same thing, and are probably not going to end up the same, so it's okay that we have repeated code here"
@fishidwardrobe yes, also true – it's a classic refactoring mistake to merge five pieces of similar code and then realise too late that four of them need to stay in sync and the other one was a temporary coincidence which now has to diverge, so you have to undo some of your work.
Not sure how we can make that one fit into the water metaphor, though!
DAMP is already in use - "Declarative And Meaningful Phrases" , and is primarily useful for things like test suites where readability and simplicity are more important than avoiding repetition. It's not quite what you're aiming for, but is still a useful middle ground!
Not related to what you're talking about (keeping things in sync) but just pointing out another usage of one of these acronyms: this “Tests Too DRY? Make Them DAMP!” (testing.googleblog.com/2019/12…) uses “DAMP” for “Descriptive and Meaningful Phrases”.
Meredith Whittaker (@Mer__edith) has been invited to speak at #39C3 later today. If you make it to the mic in time, here is an interesting question for her: "If I’m put on a boycott list by the Trump regime because I’m openly anti-fascist or oppose war crimes, will you adhere to US laws and block my access to #Signal?"
You're invited to talk on Matrix. If you don't already have a client this link will help you pick one, and join the conversation. If you already have one, this link will help you join the conversation
Auf YouTube findest du die angesagtesten Videos und Tracks. Außerdem kannst du eigene Inhalte hochladen und mit Freunden oder gleich der ganzen Welt teilen.
It is an interesting question! If the location conversations.im was located in would turn fascist and request such a thing from you, would you boot that user off the server or what would be your reaction to such a request?
Today, I collected those (and 2 extra that I thought about afterwards) into a new article on my Production Log:
"27 Essential FOSS Packages Our Animation Studio Relies On" lunaticsproject.org/2025/12/29… I changed the order to be slightly more topically organized, although the text for each packages is nearly the same as what I used in my thread.
A grid of book covers under the title "MY 2025 BOOKS" displays multiple titles including "THE ELECTRIC STATE", "DELTA GREEN GOD'S TEETH", "DEVIOUS DREAMS", "BLUE ROSE", "UKETSU STRANI DIZEGNI", "THE SANDMAN", "CHARTER SHARE", "HALF SHARE", "COCKTAIL CODEX", "KULF", "WHEN THE MOON HITS YOUR EYE", "FANTASY ENEMY IN SHADOWS", "MONSTERS, ALIENS, AND HOLES IN THE GROUND", "DEATH ON THE REK", "THE BEZZIE", "PICKS & SHOVELS", "FLOATING HOTEL", "ZERO CALCARE", "WARHAMMER", "CLOWNTOWN", "UNDER THE GUN", and "AMERICAN PASTA". Some covers show five red stars beneath them, and "THE SANDMAN" appears multiple times. Authors like "John Thorne", "Nathan Lowell", "M. HERRON", "Cory Doctorow", and "Mick Herron" are visible on specific covers, alongside titles such as "Birra in casa" and "TEN". The layout arranges these diverse book covers in a structured grid format.
Provided by @altbot, generated privately and locally using Qwen3-Vl:30b
It was a Tuesday in 1981 when the San Francisco police kicked in the door.
Inside the small apartment, they expected to find a hardened criminal. They expected a drug kingpin. They expected resistance.
Instead, they found a 57-year-old waitress in an apron.
The air in the apartment smelled sweet, thick with chocolate and something earthier. On the kitchen counter, cooling on wire racks, were 54 dozen brownies.
The police officers began bagging the evidence. They confiscated nearly 18 pounds of marijuana. They handcuffed the woman, whose name was Mary Jane Rathbun.
She didn't look scared. She didn't look guilty.
She looked at the officers, smoothed her apron, and reportedly said, "I thought you guys were coming."
She was booked into the county jail. The headlines wrote themselves. A grandmother running a pot bakery. It seemed like a joke to the legal system, a quirky local news story about an older woman behaving badly.
But Mary wasn't baking for fun. And she certainly wasn't baking for profit.
To understand why Mary risked her freedom, you have to understand the silence of the early 1980s.
San Francisco was gripping the edge of a cliff. A mysterious illness was sweeping through the city, specifically targeting young men. Later, the world would know it as AIDS. But in those early days, it was just a death sentence that no one wanted to talk about.
Families were disowning their sons. Landlords were evicting tenants. Even doctors and nurses, paralyzed by the fear of the unknown, would sometimes leave food trays outside hospital doors, afraid to breathe the same air as their patients.
Men in their twenties were wasting away in sterile rooms, dying alone.
Mary knew what it felt like to lose a child.
Years earlier, in 1974, her daughter Peggy had been killed in a car accident. Peggy was only 22. The loss had hollowed Mary out, leaving a space in her heart that nothing seemed to fill.
When the judge sentenced Mary for that first arrest, he ordered her to perform 500 hours of community service. He likely thought the manual labor would teach her a lesson.
He sent her to the Shanti Project and San Francisco General Hospital.
It was a mistake that would change American history.
Mary walked into the AIDS wards when others were walking out. She didn't wear a hazmat suit. She didn't hold her breath. She saw rows of young men who looked like ghosts—skeletal, in pain, and terrified.
She saw "her kids."
She began mopping floors and changing sheets. But soon, she noticed something the doctors were missing. The harsh medications the men were taking caused violent nausea. They couldn't eat. They were starving to death as much as they were dying of the virus.
Mary knew a secret about the brownies she had been arrested for.
She knew they settled the stomach. She knew they brought back the appetite. She knew they could help a dying man sleep for a few hours without pain.
So, she made a choice.
She went back to her kitchen. She fired up the oven. She started mixing batter, not to sell, but to save.
Every morning, Mary would bake. She lived on a fixed income, surviving on Social Security checks that barely covered her rent. Yet, she spent nearly every dime on flour, sugar, and butter.
The most expensive ingredient—the cannabis—was donated. Local growers heard what she was doing. They began dropping off pounds of product at her door, free of charge.
She packed the brownies into a basket and took the bus to the hospital.
She walked room to room. She sat by the bedsides of men who hadn't seen their own mothers in years. She held their hands. She told them jokes. And she gave them brownies.
"Here, baby," she would say. "Eat this. It'll help."
And it did.
Nurses watched in amazement as patients who hadn't eaten in days began to ask for food. The constant retching stopped. The mood on the ward shifted from despair to a quiet sort of comfort.
Mary Jane Rathbun became "Brownie Mary."
For over a decade, this was her life. She baked roughly 600 brownies a day. She went through 50 pounds of flour a week. She became the mother to a generation of lost boys.
She washed their pajamas. She attended their funerals. She held them while they took their last breaths.
She did this while the government declared a "War on Drugs."
By the early 1990s, the political climate was hostile. Politicians were competing to see who could be "tougher" on crime. Mandatory minimum sentences were locking people away for decades.
In 1992, at the age of 70, Mary was arrested again.
This time, the stakes were lethal. She was charged with felonies. The district attorney looked at her rap sheet and saw a repeat offender. He threatened to send her to prison.
One prosecutor famously whispered to a colleague that he was going to "kick this old lady's ass."
They underestimated who they were dealing with.
They thought they were prosecuting a drug dealer. In reality, they were attacking the most beloved woman in San Francisco.
When the news broke that Brownie Mary was facing prison, the city erupted.
It wasn't just the activists who were angry. It was the doctors. It was the nurses. It was the parents who had watched Mary care for their dying sons when the government did nothing.
Mary turned her trial into a pulpit.
She arrived at court not as a defendant, but as a grandmother standing her ground. The media swarmed her. Reporters asked if she was afraid of prison. They asked if she would stop baking if they let her go.
Mary looked into the cameras, her voice gravelly and firm.
"If the narcs think I'm gonna stop baking brownies for my kids with AIDS," she said, "they can go fuck themselves in Macy's window."
The quote ran in newspapers across the country.
The court didn't stand a chance.
Testimony poured in. Doctors from San Francisco General Hospital wrote letters explaining that Mary’s brownies were medically necessary. Patients testified that she was an angel of mercy.
The charges were dropped.
Mary walked out of the courthouse a free woman. But she didn't go home to rest. She realized that her personal victory wasn't enough. As long as the law was broken, her "kids" were still in danger.
She needed to change the law.
August 25 was declared "Brownie Mary Day" by the San Francisco Board of Supervisors. It was a nice gesture, but Mary wanted policy, not plaques.
She teamed up with fellow activist Dennis Peron. Together, they opened the San Francisco Cannabis Buyers Club—the first public dispensary in the United States. It was a safe haven where patients could get their medicine without fear of arrest.
But Mary wanted more. She wanted the state of California to acknowledge the truth.
She campaigned for Proposition 215. She traveled the state, despite her failing health. She spoke in her simple, direct way. She didn't talk about liberties or economics. She talked about compassion. She talked about pain.
She forced voters to look at the issue through the eyes of a grandmother.
In 1996, Proposition 215 passed. California became the first state to legalize medical marijuana.
It was a domino effect. Because one woman refused to let her "kids" suffer, the public perception of cannabis shifted. The Economist later noted that Mary was single-handedly responsible for changing the national conversation.
She never got rich.
She had always joked that if legalization ever happened, she would sell her recipe to Betty Crocker and buy a Victorian house for her patients to live in.
She never sold the recipe. She never bought the house.
Mary Jane Rathbun died in 1999, at the age of 77. She passed away in a nursing home, poor in money but rich in legacy.
Today, over 30 states have legalized medical marijuana. Millions of people use it to manage pain, seizures, and nausea.
Most of them have never heard of Mary.
They don't know that their legal prescription exists because a waitress in San Francisco decided that the law was wrong and her heart was right.
They don't know about the 600 brownies a day.
They don't know about the thousands of hospital visits.
Mary didn't set out to be a hero. She told the Chicago Tribune years before she died, "I didn't go into this thinking I would be a hero."
She was just a mother who had lost her daughter, trying to help boys who had lost their way.
She proved that authority doesn't always equal morality.
She proved that sometimes, the most patriotic thing a citizen can do is break a bad law.
Every August, a few people in San Francisco still celebrate Brownie Mary Day. But her true memorial isn't a date on a calendar.
It is found in every oncology ward where a patient finds relief. It is found in every dispensary door that opens without fear.
It is found in the simple, quiet courage of anyone who sees suffering and refuses to look away.
Mary taught us that you don't need a law degree to change the world. You don't need millions of dollars. You don't need political office.
Sometimes, all you need is a mixing bowl, an oven, and enough love to tell the world to get out of your way.
Sources: New York Times Obituary (1999), "Brownie Mary" Rathbun. San Francisco Chronicle Archives (1992, 1996). History.com, "The History of Medical Marijuana."
What I mean by that is I doubt there is even one single truly “neurotypical” person on this Earth whose brain is actually average in every important dimension of brain variation.
In the 1950s, the Air Force realized that planes were crashing because cockpits didn’t actually fit the pilots’ bodies. Wrong size = danger!! They commissioned a researcher to develop a new, more correct set of standard dimensions for the seat, yoke, etc.
That researcher, Gilbert S. Daniels, came up with 10 body measurements that matter to cockpit size. He gathered measurements of several thousand pilots. And the number of people who were at the average for all ten measurements? Zero. Not a single one.
“Average” proved to be a statistical construct, not a thing that actually exists as a person.
In many ways, the built world was not designed for you. It was designed for the average person. Standardized tests, building codes, insurance rates, clothing sizes, The Dow Jones – all these measurements are based around the concept of an “average.
High-dimensional data has this property: it is extremely unlikely that there will be a data point situated at the exact center.
It’s the high dimensionality that’s important here. One person might be at some sort of average on •one• dimension, but for them to be at the average on •all• dimensions grows exponentially less likely as the number of dimensions increases.
It’s like trying to roll all threes with a set of dice. Odds of that with one die? 1 in 6. Odds with two dice? 1 in 36. Odds with 10 dice? 1 in ~60 million.
Daniels was looking at just 10 easily quantifiable body measurements. How many important dimensions of variations are there in a human mind? How hard are they to measure? How likely is it that even one single “average” mind exists on Earth?? The odds are vanishingly small.
[Napkin sketch: assume there are a paltry 20 dimensions of brain variation. (Surely that’s low.) Assume there’s a 1 in 5 change of being completely “normal” in each. (Surely that’s high.) Even that absurd hypothetical gives a 1 in 11,490 chance that a •single• completely average mind exists in a population of 8.3 billion.]
None of the above is even proper neuroscience or psychology. It’s just a framing of the question, a way to avoid ridiculous assumptions and broken approaches, a way to avoid hurting people.
Variation is normal. Let’s expect it, design for it, work •with• it — in others, and in ourselves.
“Make the seats adjustable” is a thought I bring to teaching, for example: Does the context I’m creating for learning accommodate people with all different kinds of minds? What variations am I not accommodating? Can I make some things more individually adjustable to better embrace those variations? Can multiple instructors / learning environments / schools offer the flexibility that I can’t offer myself?
Total adjustability is impossible; infinite flexibility is impossible. But as an ongoing effort, as a •direction•, this work is both feasible and useful.
Several replies think thoughts along the lines of this one from @dalias, and I strongly agree. The •most• neurodivergent who simply cannot conform to narrow, normative expectations are doing the hard work of creating flexibility for •everyone• (see “curb cut effect”).
@lispi314@udongein.xyz @karlhigley@recsys.social I have a pet theory that there's really no such intrinsic thing as neurotypicality, just susceptibilities to being forced into various kinds of conformity and eases of performing it.
We’ve often use the word “neurotypical” to mean “neither autistic nor ADHD.” That might be useful as a shorthand, I guess, but it’s that mode of thought I’m specifically arguing against here: creating a single catch-all category defined as a negative, calling it “normal,” and assuming that it fits most people.
That doesn’t stand up to empirical scrutiny, and I don’t think it’s particularly healthy or helpful.
A lot of conversations about neurodivergence take the form of the first image below. I’m arguing to adopt the framing of the second image instead (except 100- or 1000-dimensional instead of 2-dimensional).
We’ve identified a few clusterings in a space of extraordinary and beautiful variation, and given those clusterings names. How useful those names are! How little they capture, even so! How much variation remains unnamed! How much variation must exist within every human being!
There is variation in everyone, but society / context / environment makes that variation more burdensome for some than for others. “Neurotypical” is not a thing that anyone •is•, but rather an archetype that human systems are designed for / evolved around.
When we recognize that “neurotypical” is an archetype and not an actual person, we can reach the same insight that the Air Force reached: you don’t build things to some single optimal set of “normal” dimensions; you make things more adjustable, flexible, accommodating of variation.
Wow, this is an awesome thread! I didn't know this stuff even crossed your mind, and then you suddenly just lay it all out and obliterate the silly assumptions people make. As someone who does fit (a bit awkwardly at times) in the clusters we have defined as ADHD and autism, I really appreciate this perspective!
@hosford42 I’ve never found either autism or ADHD diagnostic lsits to describe me all that well, but I’ve certainly spent enough of my life feeling •divergent• for this all to be on my mind! (I often use the phrase “my as-yet-unnamed neurodivergence.”)
Like many autistic people, I have encyclopedic recall for things that interest me -- neurodivergence being one of those topics. If you ever want to explore possible matches for your particular neurodivergence, I'd be happy to share that collected information.
@hosford42 That’s a kind offer! I don’t really feel like I’m in need of a diagnostic label: I have a pretty good sense of how my brain works, and I’m quite comfortable with it and generally able to navigate the world both effectively and happily. Now 13-year-old Paul probably could have used that…!
(I’ve heard mention of some people arguing for “giftedness” as a form of neurodivergence in itself, and although that term really gives me the ick, I suspect the category may be a good fit. Among other things: “encyclopedic recall for areas of interest” describes me too! I remember doing one autism self-diagnosis where the entire first section on “giftedness” was all, “yup, yup, oh wow, yup! Maybe I am…” and then the •all• the subsequent sections were just “nope, nope, nope, not that either, nope….”)
this is very much my reasoning for why I don't fully buy that cis het people exist. I mean practically speaking they do and they're the majority, but the ideals might not actually describe anyone.
I think even if you were to try clustering things, it would be multiple clusters, not one big cluster around the means of all the parameters.
Adjustable seats is a great metaphor for freedom in general. You can't paternalistically design society for everyone even if you wanted to.
@hosford42 @thomasjwebb That just made me realize that for as chilling Elysium was for me, it's likely to have made others equally aroused thinking about having so much power.
@thomasjwebb Strong agree. I’m cis het on paper — yet have enough gender noncomformance to have caused me personal pain and distress in my peer interactions, and to have forced me to fight hard for my own gender identity against the current. I don’t feel entirely comfortable claiming the banner of “queer” because that word importantly captures hardships I’ve never known and struggles that are not mine…but I’ve always felt more at ease in queer-friendly spaces, because I know that I too am more likely to be safe and accepted there too.
"Autigender" was an interesting word I learned in this space. I'm cis het male, sure, but I very much do it *my* way, not the standard one. I'm not performatively male. I'm just me, and that happens to line up mostly with those descriptors most of the time.
While I very much like the goal of "make the seats adjustable", I would really like to see a little more acceptance of the (neuro)diversity of teachers --- there used to be an attitude that a student should figure out how they can learn from a teacher --- now it is all on the teacher to figure out how to accommodate the student.*
Maybe there's some way to talk about meeting in the middle?
* You wouldn't believe the amount of work they are now demanding I do to accommodate hypothetical students who have never actually asked for changes. I'm getting very close to saying "F*** it. I have other, more important things to do." I do like teaching. I enjoy it. A lot of students love my class. Many have said that it was their favorite class they took in college. But teaching is supposed to be one small part of my job (if at all). The extra work being demanded is interfering with the rest of the important things that I do. At some point, these burdens are going to preclude me being willing to do the work.
“Make the seats adjustable” is a thought I bring to teaching, for example: Does the context I’m creating for learning accommodate people with all different kinds of minds? What variations am I not accommodating? Can I make some things more individually adjustable to better embrace those variations? Can multiple instructors / learning environments / schools offer the flexibility that I can’t offer myself?
Total adjustability is impossible; infinite flexibility is impossible. But as an ongoing effort, as a •direction•, this work is both feasible and useful.
Accommodation should always be a two-way street. A dialog, not a declaration. The abilities and resources of *both* parties have to be taken into account.
I saw this firsthand when an artist friend of mine didn't provide alt text. Someone chewed her out for it, and then she didn't feel comfortable posting her art anymore. What the other person failed to account for was that the artist herself is *also* disabled, and struggles to find the words to describe images. The end solution that made her feel comfortable posting online again was when people made it clear to her that it was okay to post and then ask for help with the alt text.
Whether someone is officially classified as disabled or not, there are always things they can't do. The whole point in accommodation is to treat *all* people as having intrinsic value, and to make room for each other as best we can.
I wonder what this means for something like neurology, where we don't even know how many dimensions there are, or if that number is even fixed and not its own dimension!!!
@alter_kaker Yeah, just vast terra icognita; most of our understanding of the mind is still “here be dragons.”
The things is, my “average is highly unlikely to exist” argument doesn’t depend on knowing any of that. If we assume the space of possibly variations is highly multidimensional, then it’s basically QED without any further knowledge needed!
Well, but we typically talk about human things being typical more with standard deviations than with exact numbers anyway. If "neurotypical" is anything in the first standard deviation, you're WAY more likely to get multiple dimensions in the same target. Given a normal distribution, about 2/3 of the population will be "typical" on any given trait.
@michaelc OK, if one sigma is “typical” (probably too broad to actually accommodate people well, but let’s roll with it) and we still assume a measly 20 dimensions of brain variation, then we’re up to a whopping 0.03% of the population being “neurotypical” — an extraordinarily rare condition, even under a ridiculously broad definition.
I've heard it phrased as "everyone has some neurodivergent traits, but not everyone is Autistic/ADHD/AuDHD".
When someone says "everyone is a little bit autistic", wherever they are coming from, it comes off as minimizing, dismissive, invalidating, or all three.
I don't think that's what Paul was saying, though. We have identified these categories because those of us who fall in them (like me) often struggle more than others to conform to societal expectations. But that doesn't mean that anybody is "normal". You can say nobody is normal, or nobody is typical, without denying that outliers exist and face greater challenges in society.
One slightly more blunt stat I tend to mention is that the average person has about one testicle. Which is more about bimodal distributions than your point about high dimensionality. But it still tends to rattle the cage whenever I hear people trying to optimize for the "average case" without much introspection about it.
this is in so ways why we as a species are 'failing'. we have built a society on an average (determined by the most privileged among us).
racial average. gendered average. abled average. prosperity average. medical average.
none of this averaging works because on an individual basis it is never accurate. we are all a bunch of overlapping blobs and spending all of this time trying to precicely categorize each and every one of us into specific boxes is both a gigantic waste of time and an oppressive tool that guarantees our needs won't be met.
this is of course extremely relevant in the conversation (and widespread adoption by those in power) around AI, as one big gigantic averaging tool.
Yes. And note the “average” here isn’t always even truly the midpoint of the population, but rather the locus of power: much of society is built around men, for example, when being male is not even typical.
the UNIX v4 tape reminded me of this story by Ali Akurgal about Turkish bureaucracy:
Do you know what the unit of software is? A meter! Do you know why? In 1992, we did our first software export at Netaş. We wrote the software, pressed a button, and via the satellite dish on the roof, at the incredible speed of 128 kb/s, we sent it to England. We sent the invoice by postal mail. $2M arrived at the bank. 3-4 months passed, and tax inspectors came. They said, “You sent an invoice for $2M?” “Yes,” we said. “This money has been paid?” they asked. “Yes,” we said. “But there is no goods export; this is fictitious export,” they said! So we took the tax inspectors to R&D and sat them in front of a computer. “Would you press this ‘Enter’ key?” we asked. One of them pressed it, then asked, “What happened?” “You just made a $300k export, and we’ll send its invoice too, and that will be paid as well,” we said. The man felt terrible because he had become an accomplice! Then we explained how software is written, what a satellite connection is, and how much this is worth. They said, “We understand, but there has to be a physical goods export; that’s what the regulations require.” So we said: “Let’s record this software onto tape (there were no CDs back then—nor cassettes; we used ½-inch tapes) and send that.” Happy to have found a solution, they said, “Okay, record it and send it.” The software filled two reels, which were handed to a customs broker, who took them to customs and started the export procedure. The customs officer processed things and at one point asked, “Where are the trucks?” The broker said, “There are no trucks—this is all there is,” and pointed to the tape reels on the desk. The customs officer said, “These two envelopes can’t be worth $2M; I can’t process this.” We went to court, an expert committee examined whether the two reels were worth $2M. Fortunately, they ruled that they were, and we were saved from the charge of fictitious export. The same broker took the same two reels to the same customs officer, with the court ruling, and restarted the procedure. However, during the process, the unit price, quantity, and total price of the exported goods had to be entered—as per the regulations. To avoid dragging things out further, they looked at the envelope, saw that it contained tape, estimated how many meters of tape there are on one reel, and concluded that we had exported 1k to 2k meters of software. So the unit of software became the meter.
I once read a story about the people writing the software for the NASA Apollo missions. There was a functionary in charge of weight accounting, who came to them and asked how much the software would weigh.
They told him it weighted nothing, but the functionary had heard *that* one before and insisted—everything had to be accounted down to the last ounce. He demanded to see it.
They showed him a stack of punched cards, and he was triumphant. “You see,” he said smugly, “it doesn't weigh only ‘nothing’!”
“No, you misunderstand,” they replied. “The cards aren't going on the spacecraft. Only the holes.”
When you start an STM32 MCU with its default configuration, its CPU will tick along at a leisurely number of cycles on the order of 8 to 16 MHz, using the high-speed internal (HSI) clock source as …
Se a qualcuno può interessare, al seguente link un presentazione su come programmare una FPGA sotto linux, con tanto di introduzione su come funzionano.
Put him in the role of a surgeon in front of a camera, and he will do and say things the average non-surgeon viewer will agree are surgeonish. After an hour of that, we are, as average non-surgeon viewers, satisfied and entertained.
Put him in an operating theatre, and the patient will fucking die because he's not a surgeon and knows nothing about really doing surgery.
Questo sito utilizza cookie per riconosce gli utenti loggati e quelli che tornano a visitare. Proseguendo la navigazione su questo sito, accetti l'utilizzo di questi cookie.
![Two curves showing the proportional error, |(asin(x)-f(x))/asin(x)|, for two approximations to the arcsine function.
The first curve, x + (π/2-x)(1-√[1-x^2])^2, rises from 0 to a max around 0.0085 at x=0.4, dips to around 0.004 at x=0.8, rises again to almost 0.0085 again close to x=1, then rapidly drops back to 0.
The second curve, ½(π-4)x^2 + x + 1 - √[1-x^2], rises faster than the first to a slightly higher maximum, 0.009 at x=0.3, drops all the way to 0 at x=0.7, then rises to around 0.003 at x=0.9 before falling to 0.](https://social.gl-como.it/photo/preview/640/4910344 "Two curves showing the proportional error, |(asin(x)-f(x))/asin(x)|, for two approximations to the arcsine function.
The first curve, x + (π/2-x)(1-√[1-x^2])^2, rises from 0 to a max around 0.0085 at x=0.4, dips to around 0.004 at x=0.8, rises again to almost 0.0085 again close to x=1, then rapidly drops back to 0.
The second curve, ½(π-4)x^2 + x + 1 - √[1-x^2], rises faster than the first to a slightly higher maximum, 0.009 at x=0.3, drops all the way to 0 at x=0.7, then rises to around 0.003 at x=0.9 before falling to 0.")
Comunque, qualcuno(TM) dovrebbe scrivere un'UI basata su wayland ispirata a Microsoft Bob, con l'aspetto di un'avventura punta e clicca.
senza i crash e la roba backend che aveva senso solo all'epoca, ovviamente :D