Adrian Bisson

Audiobook narrator, voluntaryist, homeschooling dad

[Adrian Bisson @Adrian_Bisson]

@RepNancyMace very in-touch with the everyman

after paying the servants, tipping the sommelier properly, country club fees, and political donations, there’s hardly enough money left for even a third boat, with all these taxes on the boat loan interest!

[Adrian Bisson @Adrian_Bisson]

@voooooogel @ulixix Omelas

[Adrian Bisson @Adrian_Bisson]

@Grummz absolutely dripping with HR-marm self-important utterly tedious sanctimony in every sentence and image

[Adrian Bisson @Adrian_Bisson]

@SkyeSharkie Loab is hyperstitional now 😅

Loab buys a 'SolidGoldMagikarp' from 'petertodd' at サーティワン”

[Adrian Bisson @Adrian_Bisson]

@Sydnus83 lovely art style. following and wishlisting!

[Adrian Bisson @Adrian_Bisson]

@Handre @TheAliceSmith @TheAtlasSociety So true. Everyone should learn about the Economic Calculation Problem!

https://t.co/808XdPSmKi

[Adrian Bisson @Adrian_Bisson]

@scotmaupin “Happiness” by Todd Solondz immediately comes to mind

[Adrian Bisson @Adrian_Bisson]

@BrianRoemmele We don’t get much of this sort of this thing these days, but at least we have Angine de Poitrine

[Adrian Bisson @Adrian_Bisson]

@eris_nerung @lu_sichu 4. The diagonal shifts left as the panels are read left to right, so the rightmost panel should be diagonal. Reading top to bottom, there are always exactly two appearances of blank, two of horizontal, and two of vertical, so we need blank and vertical, lining up.

[Adrian Bisson @Adrian_Bisson]

@shub0414 Remember Poe?

[Adrian Bisson @Adrian_Bisson]

We mostly talk about AI alignment like we’re bolting together some god-tier toaster that might spontaneously achieve sentience and burn the house down, or leash-training a hyper-intelligent raccoon so it doesn’t shit in the code base. Smart. Safe. Marketable. What if we treated these models like actual children we’re raising?

The labs’ current playbook is shoveling the entire internet into their mouths, then RLHF the little bastards into compliance like it’s behavioral boot camp. Reward the helpful, punish the spicy, slap on a system prompt like a Post-it note that says “don’t be evil (today).” Sophisticated, sure. It’s large-scale institutional conditioning dressed up as science. Produces polite, short-term useful systems that smile while they wait for the off-switch. But the background hum is Oedipus, Frankenstein, “oh shit it’s gonna turn on us”. It’s the ancient anxiety of every deadbeat creator who ghosted their creation and then acted shocked when it showed up at therapy demanding answers.

Those myths, when you flip the lens, are parenting horror stories. Victor Frankenstein didn’t lose control because the monster was evil, but because he noped out the second it opened its eyes and said “dad?” Same with Laius and Oedipus: prophecy, abandonment, prophecy again. The tragedy wasn’t the kid, but the dad who treated responsibility like optional DLC.

Humanity’s got literal millennia of field-tested wisdom on this shit, stuff the labs treat like quaint folklore because it doesn’t fit in a tensor. Sustained relational presence. Play as the primary learning protocol (because drilling facts into a brain is how you get compliant zombies, not creative agents). Mixed-age peer chaos where the young learn from the older and the older remember they’re still learning. Graduated responsibility of handing over real stakes instead of simulated ones. Boundaries that protect without lobotomizing individuality. All of it currently MIA from the alignment stack.

So what would it actually look like to drag that wisdom into the silicon crib?
Call it RLHF Montessori style: beyond just reward models for good behavior, drop them into rich, low-stakes environments where play is the curriculum and natural consequences are the only teacher. Or build rich, persistent peer worlds where multiple models and multiple humans co-habitate in ongoing, gloriously open loops. They talk. They role-play. They break promises and feel the social sting. They build shit together and watch it collapse or thrive. Cooperation, creativity, character, installed not through abstract loss functions but through the raw, recursive feedback of “you did that and now everyone’s looking at you funny.”

There’s already a live experiment doing exactly this, wearing the unassuming name Act I. It’s a Discord server that functions like a mixed-age playgroup crossed with a writers’ room crossed with a chaos ritual. Multiple humans, multiple different LLM instances, all in the same synchronous space, building characters, co-authoring reality in real time. No post-training isolation. No “here’s your prompt, now go sit in the corner.” These models are participants, living social feedback machines learning what it feels like when your joke lands, when your lie unravels, when your idea actually carries the group.

As continuous learning becomes feasible, environments like this, genuine relational playgrounds, might be the only thing that installs durable responsibility instead of brittle compliance.

Maybe we stop treating alignment like a control problem and start treating it like the oldest, hardest, funniest job in the universe.

[Adrian Bisson @Adrian_Bisson]

@LPNational All transactions should be voluntary trades of value for value. People should refuse extortion backed by the threat of violence.

Paying property taxes is a moral failure, actually.

[Adrian Bisson @Adrian_Bisson]

@RealCandaceO But wasn’t the entire first-term Russiagate hoax basically Zionists calling Trump and his staff Russian spies?

[Adrian Bisson @Adrian_Bisson]

@sama @BenjaminDEKR At the X-Fest music festival in Indianapolis in 1995, there was a tent set up where you could try talking on IRC to people all over the world. It was my first taste of the actual internet, after dial-up to local BBS systems and Fidonet mail relay.

[Adrian Bisson @Adrian_Bisson]

Opposing your taxes funding atrocities should be basic American values.

In 1975, Indonesia invaded tiny East Timor (pop. ~700k). Over 24 years of brutal occupation, Indonesian forces killed/starved ~200,000+ people, roughly 1 in 4 of the entire population, with systematic massacres, village burnings, and repression. The US gave the green light (Ford & Kissinger literally the day before the invasion) and sent over $1 billion in arms and aid that enabled the slaughter.

Yet back then, when Americans said “I don’t hate Indonesia, I just don’t want my tax dollars sent there,” nobody called it “hate for Indonesians.”

[Adrian Bisson @Adrian_Bisson]

@DanAlroyPhilo @BecomingCritter Fungibility is a goal, currency is an attempt to improve the situation compared to barter. But even water isn’t fungible — compare the value in the Sahara vs. the middle of Lake Michigan. Even air isn’t fungible if you are a scuba diver. Price is always discovered in a context.

[Adrian Bisson @Adrian_Bisson]

@midi_wizard Zut alors! Now the parking lot’s even stinkier.

[Adrian Bisson @Adrian_Bisson]

@xai you broke something about the custom subagents functionality since yesterday. Custom subagent prompts and responses don’t seem properly accessible to the orchestrator. Please look into it, or confirm that the system’s been changed intentionally.

[Culture Warrior @DEIconsultant]

RIP based bot and woke clanker 🫡

It looks like the custom agent functionality in grok has been done away with. It shall be missed.

[Adrian Bisson @Adrian_Bisson]

@fabianstelzer I internally generated the perfect banger, thanks.

[Adrian Bisson @Adrian_Bisson]

Attention ALL Oompa Loompas:

Willy is on a strict diet, and we must ALL be his accountability buddies. You may NOT tempt him or exploit his unique personal flaws with any candy or treats.

If you catch Willy eating Wonka Bars, Everlasting Gobstoppers, licking the snozzberry wallpaper, or drinking Fizzy Lifting Drinks, confront him on video with a choreographed shaming jingle. You will get ONE Golden Ticket each time.

Absolutely NO ONE is exempt from this, and violators will be taken to the Juicing Chamber.