All of Trevor Hill-Hand's Comments + Replies

I love this idea, it feels like it would also work for a lot of non-fiction, and I could see this being a part of a traditional book club too.

Ooh nice! You reminded me of A Hand With Many Fingers, as well.

Every so often I come back to a daydream about an investigation game (think Her Story, Immortality, Gone Home, Digital: A Love Story, Return of the Obra Dinn, etc.) but with a premise like you're in a small town and have access to a bunch of messy data from all over the place, mostly in Excel files but also receipts, security cam footage, maps, stuff like that.

Maybe you're a a hacker or an AI that was able to gain illicit access to everything, but now you have to clean up and correlate everything without outside help (i.e. no ability to ask things like "Wh... (read more)

Man, I would really like the news to stop feeling like it's coming from the prologue of A Fire Upon the Deep.

Amtrack also travels fairly slowly.

What would have presumably given much different results would be Claude Sonnet 4.5, which is actually a lot less sycophantic by all reports (I’m a little worried it agrees with me so often, but hey, maybe I’m just always right, that’s gotta be it.)

Now you've got me wondering if I'm being reverse-sycophantic, and have been trained to say things Claude would agree with?

I agree with the idea of looking at customer response management (CRM) systems for ideas. This talk feels like a pretty good overview of that idea: https://www.youtube.com/watch?v=8jwiABwGC6c

Because you hopefully may enjoy the ideas, I've been kind of tackling this from a hobbyist perspective:

• https://github.com/Jadael/Library-of-Aletheia
• https://github.com/Jadael/Miniworld

Lately I'm drawing inspiration from this articles, and imagining that I'm "building myself a skrode": https://medium.com/@greyboi/building-a-skrode-initial-thoughts-a195c4a0663d - and also the original story, A Fire Upon the Deep where skroderiders are introduced. Chiefly, the idea that your own skrode is something that is DEEPLY personalized and customizable, including p... (read more)

I think I feel the same sort of 'What if we just said EVERYTHING deserves welfare?' thought. I care for my birds, but I also care for my plants, and care for my books, each in their own way.

Like, if someone built this small skin-device-creature, and then someone else came along and smashed it then burned the pieces, I think I would be a little sad for the universe to have 'lost' that object. So there's SOEMTHING there that is unrelated to "can it experience pain?", for me.

This comment feels like you want to say something different than what you wrote.

There was a little Animal Crossing mod that made the rounds a little more 'gently' than I expected.

I think the trick here might be a game that runs a local, small, known-ethically-sourced model, but even if we had more than the one (Comma) that's still a lot of ire to overcome before you can even get to the elevator pitch for the game.

This three-factor framework reminds me of an idea from this: https://www.shamusyoung.com/twentysidedtale/?p=12768

If I had only attended one of these places, I probably would have concluded, “This place is what being a Christian is all about”. But these three points form a plane, and by moving around on that plane I can view Christianity from a lot of different angles and extrapolate a lot of other kinds of churches.

This idea of "third options which break overfit 1D mental models" has stuck with me for a big portion of my life now.

Hrm, what would a general purpose app (which lets one interact with LLMs/APIs while presenting what's happening to an audience) include, based on your experience here?

It's something introduced by the more agentic coding capabilities.

I think it might be mostly this, like learning to avoid ladders in go. "Crystalizing" tends to not be very useful for anything except not doing things.

I'd like to relate this old blog post I found recently after my love affair with A Fire Upon the Deep: https://medium.com/@greyboi/building-a-skrode-initial-thoughts-a195c4a0663d

I’ve been wondering about addressing this [working memory] slippage with tech. For instance, as a youngster I used to code on tiny screens, but nowadays I use as many large monitors as I can get; they help me retain context, taking the load off my working memory. So tech can definitely help if you use it well.

One of the things I’m thinking about primarily is adopting a prosthetic memory.

Could you share how many bulk-sum tokens were needed for various parts of the project, to help get a sense for what different types of work might look like across different models?

When you work with an LLM, it's definitely doing a lot of the labor, and even some of the reasoning, but you're still doing a lot of the thought yourself; before, after, and during.

I think you would enjoy this video from Alan Moore, where he starts off with a similarly fascinating word-by-word analysis: https://www.youtube.com/watch?v=ft8eO67auCs

Ah, I think perhaps I was misreading the title as August instead of Autumn. If that is case, I prefer 'Autumn' :)

The title of this thread breaks the open thread naming pattern; should it be Fall 2025, or should we be in an October 2025 thread by now? Moving to monthly might be nice for the more frequent reminder.

I love the idea of sharing "toy research", and it's encouraging me to share more of my projects.

I changed my caps lock key to a — key nearly a decade ago, and have done it on most of my keyboards/PCs ever since.

So seeing all the advice lately like "em dashes are a sign of AI writing" is a funny feeling to experience.

Though I appreciate the reference at the end, I think an important part of this is that it's also so that when you meet a hero you can do more with this skill. You can "see" and engage with the real person. A person as real and mundane as every other person, with all the good and bad that implies.

I actually think "see" is too limited an analogy, because this really involves all your senses and reasoning, but it's also true that I feel it has a close connection to what artists call "learning to see", like maybe it's using the same mental circuits.

You can le... (read more)

I enjoyed reading the paper but did not find the screenshots here in the post a helpful addition; I think I would have just quoted the introduction, if converting it into a full article was infeasible.

It's also fun seeing other Eugenia Chang fans!

Discovering that an alien species has bred a group of humans into what a pug is to a wolf would be absolutely horrific.

Makes me think of All Tomorrows: https://www.youtube.com/watch?v=imNtSPM3-r4

There's no Darwinian selective pressure to favor agents who engage in acausal trades.

I think I would make this more specific- there's no external pressure from that other universe, sort of by definition. So for acausal trade to still work you're left only with internal pressure.

The question becomes, "Do one's own thoughts provide this pressure in a usefully predictable way?"

Presumably it would be have to happen necessarily, or be optimized away. Perhaps as a natural side effect of having intelligence as all, for example. Which I think would be similar in argument as, "Do natural categories exist?"

I sat down one morning and for fun started trying to translate The Art of War from scratch, by simply going character by character and looking up the etymology and historical usage of each. Took me about two hours to get through the first page that way, and that was enough to be entertaining so I stopped there.

But, I noticed something, which reminds me of this "bodhi" vs. "enlightenment" contrast.

The text starts out by explaining five core concepts that make up "the art of war". The first one, I've normally seen written in English like this, from the trans... (read more)

This may seem like it's coming out of left field, but reading A Fire Upon the Deep a few weeks ago helped me find a calm perspective on this idea. In-universe the characters straight up have some of these discussions over the course of the book, and there's so much of all this stuff happening "just off screen".

The story is in part about the folly and impossibility of something as easy and comforting as "trust" between agents and systems in radically different scales and realities. Yet they are forced to coexist and interact regardless.

I haven't read his ot... (read more)

Sports betting is different and worse. Rather than attempting to fix outcomes, it relies on designing bets to exploit the customer.

This happens in boring and mundane ways; the same way any casino in Vegas does it, just with less regulation.

https://www.youtube.com/watch?v=_OpxrtUwjNw - This is a little fan project I did of that short story, as a sort of a radio play. I've never had it be relevant to a conversation before!

Sort of semi-related, there is the "Common Pile", a successor to "The Pile". It was not focused on "safe" data, but rather "public domain" data. But, maybe that excludes at least some dangerous data, and could make further filtering easier?

I was simply trying to decorate a compliment, so I suppose I will stop doing that 🤔 (EDIT: from a later vantage point, I think I now see it's better to say "sorry for adding a distraction" rather than passively projecting blame.)

To be clearer, the koan is meant to be related only to a sub-item of a sub-item of a comment: "you are simulating their emotions", rather than the original post or to any entire comment.

I think you make an important point in this context- understanding that all the emotions you're "feeling" are still coming from you, not from them.

"A monk rowed out to the middle of a calm lake to meditate. A while later, they were bumped into and interrupted by another boat! The monk opened their eyes in anger, ready to chide the other monk for being so careless and making them so angry... to find the other boat empty. The anger was inside them, not from another monk."

Probably need a different name for goodharting.

This is honestly the biggest concept I struggle with trying to share and teach and raise familiarity with at work, in many contexts beyond just AI safety. There are some adjacent concepts like the cobra effect that are close, but are also just close enough to be distracting.

Definitely interesting. Any word on whether this is getting more vs less true as model capabilities improve?

I'm curious too! I appreciate when experiments like this also go back and also test older models, and weirder models, and wish that were more common.

I appreciated this perspective from a prominent SCP author (Sam Hughes, who wrote and established the 'Antimemetics' stories and "sub-genre" on SCP): https://qntm.org/chatscp

"But doesn't this whole scenario sound like an SCP?"

A couple of people suggested this. An LLM which bamboozles certain types of user with paranoid fantasy until they spiral into psychosis? That sounds like science fiction! It sounds like something out of the SCP project!

Okay, so, no? Because an SCP has to be anomalous in some way and this is clearly actually happening. Four years ago

This story captures a lot of bundled feelings I have which I want to try to put into words, though words are imperfect. This is the part of the story which I feel mirrored in my own:

My mom was fighting something, and she was fighting (and teaching me to fight) harder and earlier than I knew at the time, giving me tools and perspectives that most people don't get until they are adults (if they get them at all). Now I also see that a very few people even saw that she was fighting anything, much less what it was, and few see it now, and we just can't seem to ... (read more)

I notice I apply this lesson to the design of data entry forms/surveys in general as well: you need an 'other' option much more often than you would think, or one ends up with messy survey data: extra comments and thoughts crammed into the wrong questions wherever users can find an opening. EDIT: Upon further reflection, I also remember that I've had conversations during the rollout of surveys which included the solution "let's ask them to self-classify in another question" at multiple times in unrelated projects over the years.

❤️

I have had this same feeling, in these same words, for many years now.

The clouds are sort of permanent- as they are filled in, whatever degree of detail it's at is where it stays in my head, wherever they're "stored", and it just sort of sticks in there. It feels a lot like just putting energy into some sort of "progress bar", my best thinking I really do in just a meditation sort of behavior- closed eyes, relaxed pose, slow breathing [I'll shift back and forth into this as I'm reading challenging books for instance, or when working on a project there's a lot of "lean back and sit quietly for a moment, staring into the middl... (read more)

I'm with you on "music feels language-like", I think even just looking at spectrograms of music and speech, and comparing those to the spectrograms of random soundscapes, makes it visible that music at least plays with the same types of rhythmic, pitch, and formant patterns in our "sound view", but they have a difference similar to the difference between how a textbook "plays with the patterns of letters and punctuation" to convey an idea, and a Celtic knot "plays with patterns" to create a pattern that is sort of just... intrinsically nice to have around?... (read more)

This give me a feeling I would like to express via reference:

At school, Doug finds that everyone there is dressed as him and becomes weirded out by the fact. The others also tell Doug that he is rocking the "Dylan Farnum" look. But Doug tells them that he always dresses like that. Doug has his mind stuck to the new fashion trend all day and he finally becomes fed up with the others saying that he is copying Dylan Farnum. So, he invites them all into his room and shows them his closet of clothes to prove that he is not copying Dylan Farnum. This, however,

I think it's OK to simultaneously believe one should "care about as much as possible" AND acknowledge that "as much as possible" is finite in an infinite universe.

I'd like to apply because I experiment a lot specifically on variation within an area of concept space, though I'll have to assemble a portfolio. Either way, I'd also like to throw this idea out there for consideration:

It would be nice for this to be paired with a commitment to use only ethically sourced image models, and to use that to help explain what that even means.

These two posts pair well together.

I don't know why everyone is making this so complicated when there's a clear disqualifying factor for me: Miyazaki himself has said that they did not consent to be trained on, would not have consented to being trained on, and do not want anyone making Ghibli art, and all of this was known before Sam Altman started pushing Ghibliffication. There are other factors too, but this one by itself is already sufficient for me.

EDIT: I see a lot of upvotes and disagreement on this comment, which I think I agree with. I should have clarified, this is personally disqu... (read more)

For sure! Much like the AI safety scorecard, no one is out of the red, but it seems like some of the older publishing house type companies are trying to respect existing content licensing institutions. However, I've seen many creators and artists complain that it doesn't matter; it's already too overshadowed by the actions of OpenAI et al.