Show more

So for the past couple of weeks, I've been playing with using GPT to transform HTML to JSON, basically enabling one-shot web scrapers.

Like a lot of things with these models, it went from "oh that's interesting" to "wait, it can do that?" really quickly.

I've now turned my experiments into a full-fledged library: jamesturk.github.io/scrapeghos

If you're interested, definitely read through the tutorial, I tried to recapture some of the "wow" discoveries: jamesturk.github.io/scrapeghos

@textfiles Hey what do you think of this graphic Jason? I just whipped it up, and I made it to be go with the link to your essay 🙂

Fuck all the way off with this “during #COVID” past tense bullshit.

P.S. It's a great game and it's absolutely impossible to NOT hallucinate that the main character of this game speaking in 's voice.

RIP, queen.

Show thread

I want to draw inspiration from this to add to the tactical espionage rules for ...

Dome Keeper is a ridiculously well designed game.

On the surface it seems to be about mining and shooting at monsters but really it's all about non-stop decision making, risk taking, planning and disaster management. And shooting at monsters, yes.

I really enjoy being bad at it.

store.steampowered.com/app/163

If you're scared that revealing the rules to the players will ruin immersion, work on your level design.

Q: How do you collectively call china, russia and iran?

A: The z-axis.

The more I engage with game design, the more I notice some disastrous ideas that keep turning games with great potential into "great idea, poor gameplay" for decades.

Today's highlights include:

1. Potions and buffs, which are ways to reset and/or replace game state by consuming other resources, sometimes repeatedly. The and suffer significantly from this issue. The primary reason I stopped participating in otherwise amazing story-focused modern community is this mechanic. Also having a life and a job. But mostly potions and buffs.

2. Dramatic progression, where end-game characters have quantifiably greater capabilities than early game ones. Scaling up both enemies and characters creates artificial barriers for creative gameplay. I appreciate 's progression, which features only a 15% difference in crucial stats between the end-game and the first day of your first mission.

Unfortunately, both issues have deeply infiltrated game design, including the most popular TTRPG, which shall remain unnamed.

Another abstract game design mistake is forcing players to make decisions without clearly explaining the rules.

This problem worsens in games with dramatic progression. Players must create "a build" to avoid irrelevance due to dramatic progression, relying on vague descriptions of decision outcomes. suffers immensely from this issue.

This problem can also apply to narrative games. The reason I didn't enjoy as much as anticipated and am close to dropping is that these games' premises break during play.

Broken Sword starts as noir detective with bombs, evolving into a wishy-washy modern fantasy with spells and rituals. "Steel Sky" starts as dystopian drama but devolves into cringe comedy. Sherlock Holmes games feature mysticism, and don't get me started on the TTRPG offenders: the likes of and other -adjacent games. Imagine having omnipotent entities as villains, while mortals are trying to solve crimes... Spoiler: the killer is a Cthulhu.

First time I'm playing JA (as you can guess by the fact that I have a lot of unspent cash and just one merc with me to compensate for Medic 0... I think I'll restart the campaign as soon as I learn the hotkeys during this run).

Absolutely randomly I ended up with an all-female team of mercenaries fighting for the freedom of the oppressed people. Ironically given that Miguel says "Enrique has sent *men*" .

This is so infuriating. We have let the law break things in the digital world that just worked in the physical world. @internetarchive matters so much in this world: theverge.com/2023/3/24/2365580

Stanford has created an AI model called Stanford Alpaca.
youtube.com/watch?v=xslW5sQOkC

It's capable of doing what ChatGPT does and more, except at a fraction of the cost: $600. They were allowed to do this because Alpaca didn't have to be trained same as GPT was, it only had to learn all of GPT's tricks watching GPT play.

Microsoft placed a $2 billion bet on OpenAI, so that makes them look $2 billion less smart than last week.
the-decoder.com/stanfords-alpa

#OpenAI #AGI #gpt #chatgpt #StanfordAlpaca

I'm so fucking tired of the logic of

"It's the responsibility of neurodivergent people to act neurotypically and to annoy no one"

and I want it to stop

Reduce, reuse, recycle... repair!

This week, we proposed the ‘right to repair’ initiative, which will make it easier and cheaper for consumers to #repair defective goods.

By repairing more goods and using them longer, #EU consumers can not only enjoy sustainable consumption but will also achieve considerable savings!

Learn more about the proposal → europa.eu/!KKwHFh

#EUGreenDeal #RightToRepair

I sometimes get asked for good resources on Open Systems Theory and Sociotechnical Systems Design and though it useful to collect them. View this as a personal curated list of materials to check out.
#SocioTechnical #OpenSystems
linkedin.com/feed/update/urn:l

Show more
Doma Social

Mastodon server of https://doma.dev.