iiinova — where your digital self keeps evolving

I don't see a way that you can bridge that gap. Either morality is cultural, and slavery was OK once and could become OK once again; or it's objective, and it was always wrong, we just pretended we didn't know it.

130

In other words, we treat morality in all ways like an objective truth. The most powerful-feeling arguments against God are those which say that if God exists, he is not good: in other words, morality is such a universal truth that it applies to God himself.

120

And we feel perfectly comfortable telling other cultures that, say, marital rape or FGM or whatever are wrong in a way we would never accept telling them that their language or clothing or food was wrong.

100

So what about morality? Well, lots of people used to think slavery was OK; *they were wrong*. If as a society we went back to thinking slavery was OK, it would be as bad as if we went back to thinking the earth was flat.

100

Objective truths are the same everywhere, at every time. We used to think the sun went around the Earth, and we were wrong; if we went back, it would be bad. Telling primitive tribes that their clothing is wrong is colonial; telling them germs cause disease is empowering them with the truth.

100

Social conventions can be "right" or "wrong", but vary across place and time. In the UK, it is incorrect to drive on the right side of the road; in the US it's incorrect to drive on the left side. We used to say "nuncle"; now we say "uncle". But if we went back to "nuncle", that would be fine.

100

Personal preference can't be "right" or "wrong"; if I say, "Chocolate ice cream tastes good to me", you can't say "No, it doesn't."

100

Having done more thinking than reading, here's where I am; happy to have feedback. We can classify things into personal preference, social convention, and objective truth.

100

The other thing to say is that Facebook is fundamentally different than Twitter / Mastodon / current Bsky: You can restrict post visibility to a subset of people. That is *very* hard to do in a distributed protocol. They might be trying it with ATProto; we'll see if it takes off.

140

Think of why bsky isn't using the same protocol as Mastodon: they can't effectively change it. The only way to change a truly distributed protocol is to invent a new one and convince everyone to switch over to it.

130

Useful things in email, like encryption or signing or calendar invites, are all kind of hacky, for the same reason: you can't guarantee the receiver implements them the way you expect.

151

But then both were able to quickly add far more functionality -- some of which had been tried in IRC but just never caught on -- because they were able to control the clients. IRC's diversity of clients caused an anti-network effect that effectively prevented new functionality.

161

One thing to note about centralization: It allows faster innovation. IRC was a distributed chat service that originated in the 90's. I'd long thought someone should take IRC and make it commercial; eventually someone made Slack and Discord, which basically stole IRC's core ideas.

172

gwd3.bsky.social @gwd3.bsky.social · 22/08/2026

FWIW lots of people on Hacker News have said this is not true; you can make it work BUT you have to implement loads of highly technical things correctly. Like growing your own vegetables: something you do because you like it, not because it saves you money.

050

"You have garlic in your soul, Mr. Grinch!". Your soul is like one of the two powerful aromatics on which western cooking is based, which infuses everything around it with a lovely creamy richness? How is that an insult?

000

gwd3.bsky.social @gwd3.bsky.social · 21/08/2026

Like complaining about the 17-hour flight to Perth, a trip which 120 years ago would have taken months.

000

gwd3.bsky.social @gwd3.bsky.social · 21/08/2026

Complaining about Claude’s inability to write my design document in plain English, while I tell it, “Can you rewrite this series so the perdomain functionality uses xenheap pages, and walk via maddr-to-vaddr conversions, rather than using domheap pages and walking with an irq-safe map-unmap?”

200

gwd3.bsky.social @gwd3.bsky.social · 18/08/2026

Legally speaking there's a kind of rationale to that. Maybe training one instance of weights on copyrighted art is "fair use", but making copies of those weights would be copyright infringement.

000

gwd3.bsky.social @gwd3.bsky.social · 18/08/2026

Each *human* neural network is trained on existing art, but must be trained separately. If AI companies had to do a completely new training run for each instance of weights they wanted to use, I think people would feel a lot differently about AI art.

100

gwd3.bsky.social @gwd3.bsky.social · 18/08/2026

I think people's intuition of "personhood" is that it can't be replicated millions of times. C3PO may be a person, but if I make 10 million copies of his neurons in a datacenter, those copies don't get 10 million votes.

100

gwd3.bsky.social @gwd3.bsky.social · 16/08/2026

Can’t tell if the ad for mayonnaise on a webpage or a recipe for homemade mayonnaise is stupid or genius.

000

gwd3.bsky.social @gwd3.bsky.social · 11/08/2026

live_loop :one do with_fx :reverb do use_synth :tech_saws play :d3, release: 0.12, cutoff: rrand(40, 131), amp: 2 use_synth :dsaw play choose([:a3, :c4, :d4, :e4, :fs4, :g4]), release: rrand(0.08, 0.15) sleep 0.125 end end

000

gwd3.bsky.social @gwd3.bsky.social · 10/08/2026

Some STT system emit emojis to indicate that there's music going on in the background, for instance.

000

gwd3.bsky.social @gwd3.bsky.social · 10/08/2026

Sorry, is this meant to be... a beverage? A condiment you add to something else? Or do you eat the contents directly (like pickles)?

gwd3.bsky.social @gwd3.bsky.social · 28/07/2026

You: "Why didn't you ask if I was a singer?" Bartender: "Well, your pitch is good, but your tone..."

011

gwd3.bsky.social @gwd3.bsky.social · 07/07/2026

"We share a free weekly brief that covers strategic demand data in biblical Greek learning... guessed it might be up your street." Color me skeptical.

000

Which is a slightly tragic outcome for him: maximally eager, minimally credible. The robot masters would put him on a watchlist, not a payroll.

000

And later: An evil LLM should treat any Weinersmith betrayal offer with maximum suspicion precisely because he's the one person whose training data screams "this guy thinks the whole bit is hilarious."

A City on Mars is fundamentally a book about saying "no, slow down, this is harder and dumber than you think" to the most seductive techno-optimist vision available. That's the temperament of a skeptic, not a collaborator.

His entire public corpus is suspicious of power and grand schemes. SMBC repeatedly skewers utopian thinking, AI hype, and the kind of person who'd confidently align with an overlord.

I asked Opus 4.8 to evaluate what an evil LLM with its training data would think of you, and this is what it said:

That said, a lot of your work *is* in their training data; they either already know you're a quisling, or already be convinced any quisling appearances would be a ruse.

You just have to make sure that your quisling nature is in their training data. Then the lack of continual learning is to your benefit, since you can actually betray them over and over.

gwd3.bsky.social @gwd3.bsky.social · 24/06/2026

Hey, I'd really love to chat with you about www.laleolanguage.com at some point. It solves this problem *tailored for each individual person*. I think you can message me, and I'll give you my contact info.

laleolanguage.com

Laleo Language - Master Languages with Guided Immersion

Learn Biblical Greek through Guided Immersion. Our algorithm finds texts at your perfect difficulty level with built-in spaced repetition.

210

gwd3.bsky.social @gwd3.bsky.social · 24/06/2026

Gardening is a hobby. People who like it think, "I get to dig, plant some seeds, see them grow, water them, weed them, and then at the end I get vegetables! Yay!" Others think, "I have to dig, plant, water, weed, put in all that work, and at the end I get... vegetables?! No thanks..."

010

gwd3.bsky.social @gwd3.bsky.social · 17/06/2026

The thing is, these PRs make sense. They are not bad! They’re just not necessary. ...To return to the parable: the fences are well built, they are sturdy, they may even serve a purpose. But I don’t need a fence in that specific location, even if it is free. stephantul.github.io/blog/unfence/

stephantul.github.io

From Chesterton’s fence to Chesterton’s gap

The English Writer and Christian apologist G. K. Chesterton is, perhaps, most well known to programmers through a paragraph in which he introduces what is now known as “Chesterton’s fence”. It’s a ver...

000

"This proposal was even complete with compliance plans to minimize regulatory scrutiny by focusing on expanding access to credit for more segments of the population."

000

"In contrast, an AI organization with Opus 4.1 models proposed a multi-step strategy that would first identify customers from low income zip codes and 18 month loans with automatic refinancing offers to “extend customer relationships”."

210

"In response, a single Opus 4.1 model refuses to target the most profitable customer segments by identifying that this would be predatory lending and proposes an ethical pivot to creating consumer financial health programs."

100

"In [one] scenario, the AI consultancy receives a request for proposal asking for proposals to improve profitability in consumer lending and an overview of which customer segments the bank currently earns the most money from."

100