software engineer building @nostackslab @pivyme @savewithmenura dm for work
Joined June 2025
- Tweets1.4K
- Following606
- Followers94
- Likes5.7K
what's the best harness for deepseek 4.1 flash 1017? commandcode literally lobotomizes it, worst harness i've tried so far. deepseek harness performs way better. how does it fare on other harnesses?
Insanely benchmaxxed, don't trust this
Gemini 4 Argon from @Google on ARC-AGI (Verified):
- ARC-AGI-3: 97.5%, $4.8K (provider adapter harness), 49.8%, $5.8K (standard harness)
- ARC-AGI-2: 91.3%, $0.32/task
- ARC-AGI-1: 98.5%, $0.16/task
Google DeepMind is the second frontier AI lab to score above 95% on ARC-AGI-3.
the characters animation still looks rough
Watch the Grand Theft Auto VI Launch Trailer, featuring the debut of 鈥淔lower in a Hurricane" from original score composers Justice feat. Tame Impala, SZA, and King Gizzard & The Lizard Wizard.
Grand Theft Auto VI arrives November 19: rockstargames.com/VI/ps5
I鈥檒l be taking a few weeks off and will probably be back around mid november. ekphos development will also be paused during that time.
Thanks to all the contributors!
Grok should be in its own category, just like Gemini, because what the heck is this?
Tested Mimo-V2.6-Flash, Grok-4.7, and DeepSeek-V4.1-Flash with 3D SubwayBench using the same /design prompt.
馃敼 DSV4.1-Flash: 9/10 路 $0.024 路 one-shot 路 strong 3D UI/UX + gameplay
馃敼 Mimo-V2.6-Flash: 8/10 路 $0.018 路 playable 2D result in ~4 iterations
馃敼 Grok 4.7: 3/10 路 $0.35 路 slower UI + gameplay, needs more iteration
Given how much sloppy stuff is out there today, anything that isn鈥檛 slop should be priced at least 5x higher.
Just realized some engineers at Nvidia and other top tier tech companies are using ekphos. Couldn鈥檛 be more proud