DW (@dken_w) header
DW (@dken_w) profile picture

building some stuff in gaming

Joined November 2020
Photo by DW (@dken_w)Photo by DW (@dken_w)Photo by DW (@dken_w)Photo by DW (@dken_w)Photo by DW (@dken_w)Photo by DW (@dken_w)Photo by DW (@dken_w)Photo by DW (@dken_w)Photo by DW (@dken_w)Photo by DW (@dken_w)
Pinned Tweet
Been vibecoding a little spaceship arcade arena shooter with Astra and Opus 5.5. Game is still sloppy, but the combat looks pretty cool so far. Here’s a bit of late-game gameplay
8
983
Welp… Even Fable 5.2 itself hates its own work the most.
LLM drawings, same prompt: 1. Fable 5.2 xhigh (57m, $21.77) 2. GPT 6.2 Astra xhigh (18m, $2.88) 3. GPT 6.2 Astra max, slightly guided Surprisingly, when asked to pick which one is better, Opus, Sol, Neon all consistently prefer 2 over 1. Personally i like 1 much more. Is there an obvious gap between LLM taste and human taste?
1
2
179
It’s hard for me to make a credible pixel art bench if the LLM judges’ tastes don’t align enough to human’s
1
12
LLM drawings, same prompt: 1. Fable 5.2 xhigh (57m, $21.77) 2. GPT 6.2 Astra xhigh (18m, $2.88) 3. GPT 6.2 Astra max, slightly guided Surprisingly, when asked to pick which one is better, Opus, Sol, Neon all consistently prefer 2 over 1. Personally i like 1 much more. Is there an obvious gap between LLM taste and human taste?
16
9
3
248
13,972
people shit so hard on 6.2 Astra naming "Bell", but honestly i got "Bell" from Fable 5.2 as well
I was saving this for my video but I'll spoil it now. Ben and I both made new levels in games using 6.2 Astra during the early testing. He did Majora's Mask, I did Outer Wilds. We both prompted ourselves, but it was along the lines of "make a creative new level for [game]" We both got water levels with music themes, specifically around bells. Both were named "Bellwether". This model regresses to the mean like nothing I've ever seen. Go ask it to make a new level for a game you love, lmk if it chooses "bellwether" as the name.
1
143
GPT-6.2 Astra’s drawing: a photographer darkroom. In ~2500 lines of pure C++. It’s also the first time i see a model picks c++ instead of js or python to draw arts.
4
3
95
3,246
Asked Fable 5.2 to take over and revamp it. It seems pushing towards more of a 3D game rendering...
by gpt-6 astra. 32 colors. #pixelart
2
8
421
by gpt-6 astra. 32 colors. #pixelart
1
1
7
1,242
Made with python, numpy, and c++ for ray tracing
116
Made this with Opus 5.5. Not exactly a game, but more of a "visual and sound experience" of a peaceful night ride. 🔗 passing-light.modmod.tv/ *Best with headphone for spatial audio 🎧* It's only 3 min. Hope you find it relaxing.
Pixel art by Opus 5.5 "Passing Light"
1
4
401
codex tip: when building internal tools, ask for an "application interface" and it'll drop unnecessary padding and make things neat and compact
3
1
5
528
Pixel art by Opus 5.5 "Passing Light"
2
1
1
13
1,036
Made with @kitao's Pyxel, which is actually a retro game engine. In 1200 lines of python. gist.github.com/dkenw/b0f4aa…
1
3
234
fyi, opus 5.5 medium ranked higher than max/xhigh in the previous version of FrontierCode (1.1), but it was more of a grading issue that 2.0 has solved.
such a big gap between opus 5.5 xhigh and high
197
such a big gap between opus 5.5 xhigh and high
Introducing FrontierCode 2.0: our new coding benchmark that measures a model’s ability to go beyond the spec. Most coding evals measure a model’s ability to write code that meets a spec. With FrontierCode 1.0, we asked a new question: would you actually merge the PR? FrontierCode 2.0 is comprised of more diverse tasks, featuring larger scopes and higher mergeability standards. We are now asking: not only would you merge this PR, but should you? One key feature of FrontierCode 2.0 is underspecified prompts. Rather than including every engineering decision, prompts are deliberately ambiguous, giving the models room to demonstrate their judgment. The result is a benchmark that is far from saturated, with the top model scoring just 45%.
43
25
9
1,270
263,743
the graph shows claude models are way more sensitive and reliant to thinking effort than GPT models. (green: gpt models; orange: claude models)
3
1
32
14,220
Opus just has a better game designer sense. It knows all the low-hanging fruits, like swing smear, sigil, shockwave rings
Pushing GPT 6 Astra for the best sorcerer attack animation in pixel art This one is its ~40th iteration
1
173
hey @X please fix your GIF color rendering when u upload a GIF (sRGB), X converts it to MP4 using Rec.601, but doesn't store these color metadata in the file. then, when viewing it on chrome or X app, they fall back to Rec.709 (a different color formula), causing a shift in hue.
121
Pushing GPT 6 Astra for the best sorcerer attack animation in pixel art This one is its ~40th iteration
opus 5.5, animated pixel art wizard, pure code prompt shared below 👇
2
1
6
563