FRAMEWIREIndonesiaUpdated Aug 14Live wire
0:00 / 0:00

Claude opus 5 averaged 1.70km of travel a day.

Gpt-5 got 1.80, gpt-4o 2.57, and qwen2.5 3b running on my laptop got 4.15. clustering the same places by coordinates got it down to 0.80km. I compared against whichever model won each city, and still didn't lose one. 8 wins, 2

JC BuildsAug 141
0:00 / 0:00

Claude is basically dead to me

Grok 4.6 is ridiculously good I gave it the same black hole simulation test I used with Opus 5 and somehow it came out just as good, arguably better For the kind of work I do, the capability gap between them is starting to feel nonexistent

Luckey FaradayAug 1411
0:00 / 0:00

I'm sorry but... this editor is fully ai assisted.

Opus 5 and mostly gpt 5.6. Has been a month already since I started working on that.

Giò | Jaydem WorldAug 14
0:00 / 0:00

Using Opus 5 - High to write a 3 line JSON Object

brian.Aug 14