Drew Harry
@drewwww.bsky.social
research & data @work, phd @medialab, engineer @olincollege. he/him.
socio-technical systems, computer-mediated communication, online communities, data visualization
playing: card games, roguelikes, mobas, wargames, 18xx
making: https://drewww.itch.io
socio-technical systems, computer-mediated communication, online communities, data visualization
playing: card games, roguelikes, mobas, wargames, 18xx
making: https://drewww.itch.io
Pinned
Drew Harry
@drewwww.bsky.social
· Sep 15
RECLAIMER
a tactical roguelike
drewww.itch.io/reclaimer
Use your dash to pause time and establish the perfect attack, pushing enemies into walls to conserve ammo.
Escape with as much loot as you can!
@roguetemple.bsky.social #14drl #gamedev
a tactical roguelike
drewww.itch.io/reclaimer
Use your dash to pause time and establish the perfect attack, pushing enemies into walls to conserve ammo.
Escape with as much loot as you can!
@roguetemple.bsky.social #14drl #gamedev
I regret to predict that in five years "chat" is going to morph to imply "ChatGPT."
February 7, 2026 at 10:11 PM
I regret to predict that in five years "chat" is going to morph to imply "ChatGPT."
Reposted by Drew Harry
I had "mines" scribbled in a notebook for a while and finally got around to building them. I thought of them as another area-denial tool for enemies, but one that has some interaction potential.
Here I'm placing mines and blowing them up from a distance.
Here I'm placing mines and blowing them up from a distance.
February 5, 2026 at 5:52 AM
I had "mines" scribbled in a notebook for a while and finally got around to building them. I thought of them as another area-denial tool for enemies, but one that has some interaction potential.
Here I'm placing mines and blowing them up from a distance.
Here I'm placing mines and blowing them up from a distance.
I had "mines" scribbled in a notebook for a while and finally got around to building them. I thought of them as another area-denial tool for enemies, but one that has some interaction potential.
Here I'm placing mines and blowing them up from a distance.
Here I'm placing mines and blowing them up from a distance.
February 5, 2026 at 5:52 AM
I had "mines" scribbled in a notebook for a while and finally got around to building them. I thought of them as another area-denial tool for enemies, but one that has some interaction potential.
Here I'm placing mines and blowing them up from a distance.
Here I'm placing mines and blowing them up from a distance.
This is descriptively valuable for bucketing certain sorts of beliefs.
I think a single "performance" axis gives away the game, though. Lots of software is better-than-human at some things. LLMs certainly are too. And they're better in surprising domains, as opposed to, say, a calculator.
I think a single "performance" axis gives away the game, though. Lots of software is better-than-human at some things. LLMs certainly are too. And they're better in surprising domains, as opposed to, say, a calculator.
I was going to write a whole effortpost about model evaluations but instead here, just take this chart
February 4, 2026 at 10:27 PM
This is descriptively valuable for bucketing certain sorts of beliefs.
I think a single "performance" axis gives away the game, though. Lots of software is better-than-human at some things. LLMs certainly are too. And they're better in surprising domains, as opposed to, say, a calculator.
I think a single "performance" axis gives away the game, though. Lots of software is better-than-human at some things. LLMs certainly are too. And they're better in surprising domains, as opposed to, say, a calculator.
Thinking about Arrival.
February 4, 2026 at 6:35 PM
Thinking about Arrival.
Weapons can now have an odds to miss, and a range of potential miss angles.
Broadly, output randomness increases tension and push-your-luck feelings. I think that fits my gameplay pillars, so I'm playing around with it more to see what feels right.
Broadly, output randomness increases tension and push-your-luck feelings. I think that fits my gameplay pillars, so I'm playing around with it more to see what feels right.
February 3, 2026 at 1:58 AM
Weapons can now have an odds to miss, and a range of potential miss angles.
Broadly, output randomness increases tension and push-your-luck feelings. I think that fits my gameplay pillars, so I'm playing around with it more to see what feels right.
Broadly, output randomness increases tension and push-your-luck feelings. I think that fits my gameplay pillars, so I'm playing around with it more to see what feels right.
Reposted by Drew Harry
Without strong public transit, mobility doesn't work for anyone. Join me this morning at 10a at the Ferry Bldg to kick off signature gathering for Connect Bay Area. We need your help to win
luma.com/le1g7slf?tk=...
luma.com/le1g7slf?tk=...
Signature Gathering Kickoff for Connect Bay Area Transit · Luma
To avoid devastating Bay Area transit service cuts, we need to gather over 200,000 signatures to get Connect Bay Area Transit, the five-county regional transit…
luma.com
January 23, 2026 at 3:45 PM
Without strong public transit, mobility doesn't work for anyone. Join me this morning at 10a at the Ferry Bldg to kick off signature gathering for Connect Bay Area. We need your help to win
luma.com/le1g7slf?tk=...
luma.com/le1g7slf?tk=...
Added reload UI sugar. When you run out of ammo in your weapon, there's a visual marker that trails your cursor. When you reload, there's a notice next to your character.
January 5, 2026 at 4:00 PM
Added reload UI sugar. When you run out of ammo in your weapon, there's a visual marker that trails your cursor. When you reload, there's a notice next to your character.
Found this old GIF of a sneaking-focused roguelike prototype. Interesting sense of scale to it. I really like how vision looks walking through doors or peering through a door at long-distance.
January 4, 2026 at 8:03 PM
Found this old GIF of a sneaking-focused roguelike prototype. Interesting sense of scale to it. I really like how vision looks walking through doors or peering through a door at long-distance.
Reposted by Drew Harry
Playing @drewwww.bsky.social's RECLAIMER. I would appear to be in danger.
January 4, 2026 at 5:49 PM
Playing @drewwww.bsky.social's RECLAIMER. I would appear to be in danger.
I've started reading these to my kid, and it's a delight to return to Tolkien as an adult. When I read them as a teen, I was trying to suck the plot out of them as a pure adventure story. But there's so much here to chew on that I missed.
With a knock on your door, a kindly wizard invites you to join us on Shelved By Genre's LORD OF THE YEAR, in which we will consider the popular but contested works of JRR Tolkien, kicking things off with a long rumination upon the first four chapters of THE HOBBIT: rangedtouch.com/2026/01/02/t...
January 2, 2026 at 8:06 PM
I've started reading these to my kid, and it's a delight to return to Tolkien as an adult. When I read them as a teen, I was trying to suck the plot out of them as a pure adventure story. But there's so much here to chew on that I missed.
First steps on a little tutorial system. I always put this off and regret it. I'm going to try to make something that onboards players and gives them some contrived scenarios to see if it's fun.
January 2, 2026 at 12:29 AM
First steps on a little tutorial system. I always put this off and regret it. I'm going to try to make something that onboards players and gives them some contrived scenarios to see if it's fun.
Reposted by Drew Harry
🍄 happy new year ✨
may the ritual nourish you
may the ritual nourish you
January 1, 2026 at 12:19 AM
🍄 happy new year ✨
may the ritual nourish you
may the ritual nourish you
@lazerwalker.com ran into this and I recall you being curious about next gen go clients. Haven't spent much time with it but it seems on the path towards that destination: github.com/kaya-go/kaya
GitHub - kaya-go/kaya: A modern, elegant Go (Baduk/Weiqi) game application
A modern, elegant Go (Baduk/Weiqi) game application - kaya-go/kaya
github.com
December 28, 2025 at 12:57 AM
@lazerwalker.com ran into this and I recall you being curious about next gen go clients. Haven't spent much time with it but it seems on the path towards that destination: github.com/kaya-go/kaya
Made a little guy that turns if its path gets blocked.
December 22, 2025 at 2:19 AM
Made a little guy that turns if its path gets blocked.
Thinking about how to communicate as much as possible in a compact space. I want players to know (1) how much damage an attack will do, (2) how an enemy will be moved by the attack, (3) how much health an enemy has, and (4) if an attack will be lethal or not.
This animated bar is my current idea.
This animated bar is my current idea.
December 20, 2025 at 7:30 AM
Thinking about how to communicate as much as possible in a compact space. I want players to know (1) how much damage an attack will do, (2) how an enemy will be moved by the attack, (3) how much health an enemy has, and (4) if an attack will be lethal or not.
This animated bar is my current idea.
This animated bar is my current idea.
Worked up a new "intent" system, where NPCs can target a spot a turn in advance and then fire the next turn. I'm inspired here by tactics games that "show the barrel of the gun" like "Slay the Spire" or "Into the Breach."
The NPC has a laser weapon with two shots in it, then it needs to reload.
The NPC has a laser weapon with two shots in it, then it needs to reload.
December 19, 2025 at 8:11 PM
Worked up a new "intent" system, where NPCs can target a spot a turn in advance and then fire the next turn. I'm inspired here by tactics games that "show the barrel of the gun" like "Slay the Spire" or "Into the Breach."
The NPC has a laser weapon with two shots in it, then it needs to reload.
The NPC has a laser weapon with two shots in it, then it needs to reload.
Beyond research, craft ITSELF drives huge amounts of human effort. It is a mistake to think that people make things only to have them and would welcome a way to skip the making itself. Doing the work is not a "cost."
This is such a misunderstanding of why writers, artists, and designers do research to begin with, which is to find something unexpected, not capture the average or find reference for the default.
And, as always, the biggest misapprehension: We don’t want to NOT do this part. It’s fun to learn shit
And, as always, the biggest misapprehension: We don’t want to NOT do this part. It’s fun to learn shit
Anyone who supports this is an embarrassment. "Instead of typing it out they'll generate their idea"? Are you insane?
Call yourself whatever you want but you aren't creatives, that's for sure.
Call yourself whatever you want but you aren't creatives, that's for sure.
December 19, 2025 at 6:53 PM
Beyond research, craft ITSELF drives huge amounts of human effort. It is a mistake to think that people make things only to have them and would welcome a way to skip the making itself. Doing the work is not a "cost."
The npc bot (orange) can lock on to a player it sees nearby and declare its intent to shoot next turn. Anything in the red line will take damage, even if it's not the intended target.
December 17, 2025 at 2:22 AM
The npc bot (orange) can lock on to a player it sees nearby and declare its intent to shoot next turn. Anything in the red line will take damage, even if it's not the intended target.
This is another way of saying what @jamellebouie.net was saying earlier this week. The text-based platforms are marginal among teens, and video-based platforms are dominant.
Teens turn to a variety of online platforms, but YouTube continues to stand out. Roughly nine-in-ten teens report ever using it. TikTok, Instagram and Snapchat are the next most widely used platforms.
December 16, 2025 at 9:19 PM
This is another way of saying what @jamellebouie.net was saying earlier this week. The text-based platforms are marginal among teens, and video-based platforms are dominant.
I'm curious about this too. I see a few contributing dynamics:
- Attribution in software engineering is rare
- Open source culture
- High average pay, low sense of precarity
- Disconnection between "how" (messy LLM-internals) and "what" (the product); resulting UI has less of an LLM-smell.
- Attribution in software engineering is rare
- Open source culture
- High average pay, low sense of precarity
- Disconnection between "how" (messy LLM-internals) and "what" (the product); resulting UI has less of an LLM-smell.
i am not a coder so i’m an extreme outsider here but i’m a little surprised that’s the group that has most embraced ai. isn’t the whole point of hiring programmers, the reason they get big money, not so [product] works but knowing how and why it works so it’s efficient, reliable and scalable
December 16, 2025 at 8:59 PM
I'm curious about this too. I see a few contributing dynamics:
- Attribution in software engineering is rare
- Open source culture
- High average pay, low sense of precarity
- Disconnection between "how" (messy LLM-internals) and "what" (the product); resulting UI has less of an LLM-smell.
- Attribution in software engineering is rare
- Open source culture
- High average pay, low sense of precarity
- Disconnection between "how" (messy LLM-internals) and "what" (the product); resulting UI has less of an LLM-smell.
Playing with a bouncing grenade-type effect. It's a bit awkward at this grid size. You can get a lot of situations where it's hard to produce an intuitive result. The preview graphics help, but aren't a total solution.
Not sure if this is stable enough to use in a real game, but I'm glad I tried.
Not sure if this is stable enough to use in a real game, but I'm glad I tried.
December 13, 2025 at 4:29 PM
Playing with a bouncing grenade-type effect. It's a bit awkward at this grid size. You can get a lot of situations where it's hard to produce an intuitive result. The preview graphics help, but aren't a total solution.
Not sure if this is stable enough to use in a real game, but I'm glad I tried.
Not sure if this is stable enough to use in a real game, but I'm glad I tried.
Step out of the way and push the rocket towards another target.
December 10, 2025 at 4:02 PM
Step out of the way and push the rocket towards another target.
Rockets that fly towards the player and then explode. Not smart; once they spawn they cannot change direction.
December 8, 2025 at 5:01 PM
Rockets that fly towards the player and then explode. Not smart; once they spawn they cannot change direction.
My 2012 dissertation is, basically, 171 pages about why not to do this and what to do instead.
I love the giant slab of Twitch Chat on stage during the Streamer Awards. Looks like the monolith in 2001: A Space Odyssey.
December 7, 2025 at 9:27 PM
My 2012 dissertation is, basically, 171 pages about why not to do this and what to do instead.