apt for this week, next week, and the next several weeks ;)
Replying to @haxBTC
Rambling thoughts on software, bugs, and the new "AI kills your code" era: primal.net/e/naddr1qvzqqqr4g…
34
Replying to @glazworks @gurenyov
yes, anthropic has earned our business. but our business isn’t reselling theirs.
i think marking up tokens is a bad business: the labs squeeze you from below, competitors squeeze you from above, and the value your layer adds over the tokens diminishes with every generation of models.
we want our best creators to keep creating. make something great, bring their friends, build an audience, make a living. our bet is that the economy around those games becomes worth far more than the compute it takes to make them.
and we’re already seeing the beginnings of it. creators are selling ad space inside their games. we don’t run ads. we never asked them to sell ads. they built places people wanted to spend time, and figured out that someone would pay to be there.
that’s a much more interesting business to me than charging someone another $20 because they ran out of prompts.
free creation is deliberate and not designed as a promotional one time thing. charging people to upload videos would have made early YouTube’s hosting bill look better but imagine what it would have cost them.
we still have plenty to prove. but the thing we’re trying to prove is that people want to build, play, and spend time together on Spawn. creators already do, and player retention keeps improving.
the plan is to take a share of the economy that grows around their work and use it to keep creation free forever. i want creators making game not worrying about token meters
this is why i'm bullish on things like @spawn
the agents can make anything now, but most people don't want to make random software (and most people won't care about random software)
but i think basically everyone wants to make a game, and those games have to exist somewhere
Like everyone else I put Fable 5.2 through the ringer on game dev, making an atmospheric kikis delivery service inspired broom riding game.
I'm really impressed with how well it did with the visuals. It's so pretty!!
1
1
86
"ok next on the board is Big Walk, they did a great pre-release ritual of withholding sales from Microsoft due to their funding of jews, so that should score them some BIG points!"
"Ohh sorry, you asked AI for help, therefore your group now want you dead. Big Walk -1,000 points"
Big Walk used AI in 'the final stretch' of development, kicking off intense awards debate:
aftermath.site/big-walk-ai-i…
Readers added context they thought people might want to know
House House used LLMs sparingly near the end of Big Walk's development, mainly to fix bugs in third-party code. They estimate well under 0.1% of the game's code was affected, and they didn't use generative AI elsewhere.
bigwalk.game/faq/#ai
pcgamer.com/games/puzzle/b…
22
hax.btc retweeted
You live a normal life without a printer in your home but one day every 2 and a half years you need a printer
hax.btc retweeted
Apparently, White college girls making monkey noises was a bigger story than a White college girl being shot to death by a pack of blacks.
Illinois college cheerleader, 19, was riddled with bullets in Halloween mass shooting near campus trib.al/i3HK943
hax.btc retweeted
Democrats win elections because the median voter is no longer capable of understanding the future as a concept
Every Sunday I'll post some links that I found interesting for AI, starting with...
OpenCode/Anomaly shipping lots of cool shit:
teapawt.net/thdxr/status/211920994…
Antseed verifying your models are who they claim to be:
teapawt.net/geniusyinka/status/211…
lil local models:
teapawt.net/0xCodez/status/2118424…
Andrej Karpathy:
“A 1B model can be much more effective than most current LLMs.
90% of current models waste a ton of capacity and money on remembering stuff that doesn’t matter.”
here is his 5 core thesis about Small LLM’s:
1. the intelligence was never the knowledge. Strip the memorized internet, keep the reasoning. What’s left is the cognitive core, and it’s small.
2. the core might fit in ~1B params. Maybe less. “Plenty of room at the bottom.” We’ve been paying for a library when we needed a brain.
3. size stopped being the bottleneck. What makes an agent work is the tools it can call and how tight the feedback loop is, not parameter count.
4. intelligence lives in the system, not the weights. A raw LLM is a CPU with no OS. Small core + good harness > giant model alone.
5. you can own it. A model on your hardware is one no one can rate-limit, price-hike, or deprecate on you.
in 12 months the gap between teams shipping on-device and teams paying per-token frontier bills won’t be capability.
TwIL an 8.8B model at 90%+ of frontier isn’t a benchmark flex.
It’s the thesis Karpathy’s been repeating all year, made real.
study more on this below ↓
1
27
I want to connect with more interesting, intelligent, insightful, and innovative AI people.
I want to see more workflow setups, pipeline feedback, and knowledge sharing.
I want to be in a space that explores the new AI ERA together, so I made this: teapawt.net/i/communities/16470391…
2
27

















