What did they do to Astra so cracked at vision (and computer use). That ARC 3 score turned out to be no joke/fluke. That huge gap between Astra and Fable (in this case) is basically every hard vison/spatial benchmark i've seen including non-benchmarks like playing games (Portal, Factorio, RimWorld).
I think Opus 5.5 is at same level now. I have seen too many videos made by Opus 5.5 today on twitter.
<a href="https://x.com/victormustar/status/2102707412704919910" rel="nofollow">https://x.com/victormustar/status/2102707412704919910 horse galloping pixel art
<a href="https://x.com/LexnLin/status/2102133072585965759" rel="nofollow">https://x.com/LexnLin/status/2102133072585965759 moving train pixel art animation
<a href="https://x.com/jkeatn/status/2102441348075057539" rel="nofollow">https://x.com/jkeatn/status/2102441348075057539 painting with code
<a href="https://x.com/LCSlates/status/2102503027340988559" rel="nofollow">https://x.com/LCSlates/status/2102503027340988559 video, very detailed prompt though
<a href="https://x.com/aj_dev_smith/status/2102504509637587339" rel="nofollow">https://x.com/aj_dev_smith/status/2102504509637587339 generated song/music with code
<a href="https://x.com/aj_dev_smith/status/2102575577563570450" rel="nofollow">https://x.com/aj_dev_smith/status/2102575577563570450 another song
Aren't these the same thing, to some extent? I remember my music teacher told me, that as long as you can hear your false tones, she can teach you how to sing, no matter how bad you are at it. But if you can't hear it, she cannot help you.
AI is the same - as long as it can see well, it can tell the difference between what it outputs and what its supposed to. If you subtract the two, you have an error, and you can hill-climb on that.
famouswaffles · · focus · HN ↗
SpatialBench - <a href="https://x.com/spicey_lemonade/status/2096365630190698516" rel="nofollow">https://x.com/spicey_lemonade/status/2096365630190698516
ZeroBench - <a href="https://zerobench.github.io/" rel="nofollow">https://zerobench.github.io/
Robot Arms - <a href="https://openai.robocurve.org/gpt-6-astra/">https://openai.robocurve.org/gpt-6-astra/
smusamashah · · focus · HN ↗
<a href="https://x.com/victormustar/status/2102707412704919910" rel="nofollow">https://x.com/victormustar/status/2102707412704919910 horse galloping pixel art
<a href="https://x.com/LexnLin/status/2102133072585965759" rel="nofollow">https://x.com/LexnLin/status/2102133072585965759 moving train pixel art animation
<a href="https://x.com/jkeatn/status/2102441348075057539" rel="nofollow">https://x.com/jkeatn/status/2102441348075057539 painting with code
<a href="https://x.com/LCSlates/status/2102503027340988559" rel="nofollow">https://x.com/LCSlates/status/2102503027340988559 video, very detailed prompt though
<a href="https://x.com/aj_dev_smith/status/2102504509637587339" rel="nofollow">https://x.com/aj_dev_smith/status/2102504509637587339 generated song/music with code
<a href="https://x.com/aj_dev_smith/status/2102575577563570450" rel="nofollow">https://x.com/aj_dev_smith/status/2102575577563570450 another song
prideout · · focus · HN ↗
torginus · · focus · HN ↗
AI is the same - as long as it can see well, it can tell the difference between what it outputs and what its supposed to. If you subtract the two, you have an error, and you can hill-climb on that.