max (@maxbittker) header
max (@maxbittker) profile picture

toymaker, likes learning :^) https://teapawt.org/t.co/dPbgElCy3G

Brooklyn, NY
Joined September 2011
Photo by max (@maxbittker)Photo by max (@maxbittker)Photo by max (@maxbittker)Photo by max (@maxbittker)Photo by max (@maxbittker)Photo by max (@maxbittker)Photo by max (@maxbittker)Photo by max (@maxbittker)Photo by max (@maxbittker)Photo by max (@maxbittker)
birds dancing on my ceiling
14
75
6
717
Gemini 4 Neon has the same affliction, doing terribly on Runescape Bench because it tries super-ambitious strategies yet has terrible time management and doesn't finish them before the buzzer
Gemini 4 Carbon underperformed on Runebench. Interestingly, it attempted ambitious strategies (quests) then failed to complete 7 of them, wasting more time than any other model and hurting its score.
10
2
1
88
4,694
The strongest models check the clock ~every 30 seconds. Neon checks it 6x less often AND chooses long complex strategies.
1
5
546
Honestly can't decide if this means it's unbothered and unbenchmaxxed, or if it just got distilled in some bizarre way where it internalized super high standards from its older siblings (hence the aggressive strategies) without the skills to complete them
4
435
Cool results, especially enjoyed the human vs AI PvP scenarios, where thankfully @SethBling could outmaneuver and outsmart agents
Replying to @kradleai @SethBling
We’ve designed kradle.ai, a bench that’s messy, unpredictable, hard to saturate, and full of others (AIs & humans) pursuing their own goals. Like the real world. You may be thinking… Claude?!? Carbon? OpenAI takes so many top slots because their models are more iterative and adapt to their environment. Astra 6.2 iterates on its approach to tasks ~4x faster than Fable 5.2. In general, Claude models tend to think for longer and then stick with the strategy they come up with. And, Carbon is #16 due to latency.
1
1
9
825
css is so hard
86
2,257
371
25,639
908,712
Astra and Sol 6.2 are new leaders on Runescape bench. Their signature strategy was completing 2 different quests and collecting both rewards simultaneously, beating records I'd thought were saturated.
7
7
2
178
8,717
for whatever amount of compute you have, you will get much more research done if you can just make your experiments smaller
23
25
3
609
19,691
river.maxbittker.com is newly smooth and expanded, come surf
21
19
3
213
10,312
now featuring 2x more images, text/image search, and a faster index, browse @AREdotNA slipperily
13
724
Hy4 High inside Pi is (unofficially) is a Pareto point in RuneBench with 6.23 score at about 0.59 USD/run Hy4 High on OpenCode 1.19 scores 4.92 at 0.36 USD/run (curious now to run DS 1017 on DSH minimal) I can share the data if you want to officially validate @maxbittker
1
1
4
657
Gemini 4 Carbon underperformed on Runebench. Interestingly, it attempted ambitious strategies (quests) then failed to complete 7 of them, wasting more time than any other model and hurting its score.
9
1
71
6,611
There's a youtuber producing an incredible educational series about learning advanced agent techniques via runescape, so awesome to see!
1
3
174
4,633
This second video is more advanced and covers remaining skeptical, building harnesses and visualizations, saving on token usage via model routing. Really great stuff youtube.com/watch?v=Rnf5csei…
13
960
Does anyone else re-run old prompts when new models come out to find mistakes you had accepted?
1
1
1
33
1,752
Gell-Mann Remembering
328
Fable 5.2 is very good at painting!! 🎨 Congratulations to the @claudeai team!(Video + Audio made in code). Comparison below between Fable 5.2, Fable 5.1 and Gemini 4 Argon on the same prompts. ⬇️
10
11
150
7,910
Fable 5.2 RunescapeBench results! New #2 result after Astra
19
7
1
327
29,920
At default effort, it's VERY similar profile to Opus 5.5, but at higher thinking it pulls ahead and is on the pareto frontier maxbittker.github.io/runeben…
12
2,604
finally… I have removed YouTube Shorts
i am genuinely so grateful that fairplay has finally been broken. it’s lifechanging. i can text claude to mod any app now and it’s on my phone in a half hour
6
1
1
305
29,859
death animation/ragdoll test
death animation/ragdoll test
44
789
20
8,049
88,329