Misha

Misha

@the-goblin.bsky.social

Prof of Game Design at CCNY. Games: Sumer (Switch, Steam), Goblin.bet, Odd Goblin (itch). Cohost of Dead Letters podcast (@dead-letters).

193 followers295 following232 posts

Misha reposted

Misha reposted

Replying to @lorenschmidt.bsky.social

this death actually provides a really good checklist for RPG designers to go over when adding new features: (1) is it an interesting challenge with redundant soft workarounds, or is it a hard counter? be aware of the shape of the interactions this will have with other systems.

Cross-Country Road Trip Day 1, San Diego->Morro Bay: Got the full LA experience (ate excellent tacos outside an Amazon warehouse then got stuck in gridlock traffic). Here's Bandit posing in front of "Malibu Shaman". Fun fact: her name at the rescue we got her from was "Malibu"

D1 Critter of the Day: Heermann's Gull, from Carpinteria Bluffs. I love their little goth beak and leggings. I only saw my first one a few days ago and they're quickly becoming one of my favorite gulls.

Misha reposted

I love Edward Gorey’s theory of art: “Art…is presumably about some certain thing, but is really always about something else, and it’s no good having one without the other, because if you just have the something it is boring and if you just have the something else it’s irritating.”

Solid post, I'd even go so far as to say that one of fantasy's defining traits is being the setting most built around play-enabling shorthand.

Misha reposted

I've been listening to the Dead Letters back catalogue, and I think both Sam and Misha make excellent points about Wolves. Sam is absolutely correct that the way it is written makes it exciting to read - and is thus crucial to getting the prospective GM to actually read it, 1/7

new DEAD LETTERS episode! this time, it's the vast, legendary, landmark WOLVES UPON THE COAST, written by that one bald guy! we talk about the Adventure Writing Trust Horizon, reading, horses with unhinged jaws, poetic monster manuals, legwork, and the GM as medium! open.spotify.com/episode/2Z1e...

This is a fun episode, not least because I get to let loose my hot take that the sublime is bad for ttrpgs

lo and behold, another episode of DEAD LETTERS! this episode is about the legendary VEINS OF THE EARTH by Scrap and PatStu! we discuss blogposts as drafts, 3D environments, the Shaver Mystery, tangible concepts, alien psychology, and more! come listen! open.spotify.com/episode/3W3t...

Other than the more obvious stuff, this is one of the things that spooks me the most about LLMs, although I've been struggling to articulate it. When you ask them a question they take all of human writing and select the one 'best' answer, implicitly positioning that answer as the sum of the truth.

Crystal Fleming is on sabbatical@alwaystheself.bsky.social

He kept coming back, repeatedly (in the way AI often loops) to this idea of AI (specifically Claude) representing "a position without an address", and it became increasingly clear to me that there was some odd mystification happening. Of course LLM products have an address. It's a corporate address.

View full thread

At least something like web search positions its answers as "here's a collection of what others have said about this topic", but even within that model the fight for SEO is a blight on society. I think this would be a huge fundamental problem for LLMs even if they were run altruistically.

My conspiracy theory is that this gets at something deeper in the goals of LLM/AGI development. Some pathological attempt to turn the whole world into data, capture all of truth and reality in a computer, and by extension render it all enframed and controllable from the comfort of the CEO's chair.

Other than the more obvious stuff, this is one of the things that spooks me the most about LLMs, although I've been struggling to articulate it. When you ask them a question they take all of human writing and select the one 'best' answer, implicitly positioning that answer as the sum of the truth.

Crystal Fleming is on sabbatical@alwaystheself.bsky.social

He kept coming back, repeatedly (in the way AI often loops) to this idea of AI (specifically Claude) representing "a position without an address", and it became increasingly clear to me that there was some odd mystification happening. Of course LLM products have an address. It's a corporate address.

The world is sprawling and complicated and any question you could ask about it has vast numbers of valid and insightful answers. There's something about a small number of huge corporations positioning themselves to be the selectors of the one true one that's deeply and fundamentally unsettling to me

Misha reposted

very good post! extremely insightful and actionable. if you like my "tangible tips" posts about editing, you will find lots of useful ideas here

Misha@the-goblin.bsky.social

Closing out my series on key design by making a list of rules for myself to follow when writing modules, then testing it by rewriting a one page dungeon I made recently. The goal: retain all the strengths of bullet points/formatting with thoughtful use of plain text structuring. It went really well!

Closing out my series on key design by making a list of rules for myself to follow when writing modules, then testing it by rewriting a one page dungeon I made recently. The goal: retain all the strengths of bullet points/formatting with thoughtful use of plain text structuring. It went really well!

baatag.blogspot.comElements of a Key 4: Rules for Key-WritingAlright, last time I read a bunch of keys from various modules and looked at what types of info they included and how they presented it. I...

Misha reposted

indie game marketing is untenable right now, bro allegedly took a shot at the president and it got him 22 concurrents