AI enthusiast • Streamer • YouTuber • Christian • Vibe coder
Building weird internet things, documenting the process, and writing the thoughts my brain refuses
Dungeon party chat
Warlock: haven’t been here since 04
Priest: I was born in 04
Warlock: I have kids older than you
Priest: you should have them play
Warrior: funny you say that he’s my dad
This is what wow forever about!
Searching for open-weight models can be so much better.
The right model for you shouldn’t be buried in a directory. The models must flow.
Introducing model trees. Check the revamped models page ↓
Featuring @elder_plinius@support_huihui@QwenDevs
That's b/c you're not smart enough to understand the beauty of price-based allocation. Uber rapidly raising prices (to the market-clearing price, i.e. the price at which everyone who wants to take an Uber at that price can, and every Uber has a customer):
1) Allows people who need the scarce service the most to get it.
2) Incentivizes people to look for substitutes: call a friend to drive you, walk if you can, etc., etc.
3) Incentivizes others to provide that service: people come with private cars, even from distant places, to offer rides for profit, etc. (and, in the process, lower the price).
In contrast, forcing them to sell below the market-clearing price does NOTHING to improve the availability of transportation. It just imposes allocation according to some other function: who has a personal connection with an Uber driver or management (corruption), who's lucky enough to get a ride, etc., etc.
So you end up with a situation where a single mother with two babies has to stay at the airport, while a young man who can walk, call a friend, or has another option that's cheaper than the market-clearing price but more expensive than the "capped price" simply takes the (cheap) Uber that the mother would have gotten at the market price.
https://t.co/fX7rCBrXel
> playing WoW
> take a small break
> check twitter
> bitcoin under 69k
> planes stopped working
> another jensen affair
> go back to playing WoW
I burned a shitload of compute so you don't have to. Releasing the Harness Leaderboard.
Which harness should you run your open model in? I ran 28,849 trials to find out.
The leaderboard compares nine coding-agent harnesses on open models. In each comparison only the harness changes: the model, provider and reasoning effort stay fixed, and every task gets three attempts.
The trials ran GLM-5.3-Flash, DeepSeek-V4.1-Flash and Flint on Terminal-Bench 2.1 and DeepSWE, for $4,746 of inference.
Averaged over the three models, mini-swe-agent (@KLieret), a bash-only agent, solves 51.4% of tasks and @openclaw 27.3%. Four others are statistically tied with mini-swe-agent.
One of them is Pi (@pidotdev), at 49.4% for $0.29 per solved task, about 45% less than mini-swe-agent's $0.53.
The best harness depends on the model: mini-swe-agent on GLM-5.3-Flash, Hermes Agent (@NousResearch) on DeepSeek-V4.1-Flash and Kimi Code (@Kimi_Moonshot) on Flint.
@opencode costs $3.35 per solved task, 11.7 times as much as Pi.
See Harness Hub's Harness Lens for what the nine harnesses do differently. It documents the prompts, tools and compaction of 34 coding-agent harnesses and links every claim to its source.
Leaderboard: https://t.co/AiWiQtBOz5
Harness Lens: https://t.co/I8jbcWa9JH
Congrats to M Hardcorpse on achieving the World First Hardcore Worlds death! 🪦
He died 9 seconds after leaving Tutorial Island 🏝️
Officially verified by the Jagex Anti-Cheat and Data teams! 🏆