@cnakazawa Pretty sure its intentional for cost/satisfaction users, most users are probably more latency-sensitive than quality-sensitive (and have low token limits)
The challenge we face now as engineers is to automate the arguing process lol, like Gary's "boil the ocean" line
.@openclaw is NOT ready for normies, will be years
cant even send me a video file (not that large) and keeps compacting itself?
(Codex harness btw, not a crappy model)
Current AI coding setup:
$200 Codex max
$20 Claude (minimum to get Claude Code)
$20 Cursor (for tab-edit suggestions only lol)
GPT does everything, but when it shits out a bad UI Opus cleans it up. I manually audit in Cursor.
I wonder how long this will last
Codex tried to get me in trouble last night. After a quick investigation...
cum = cumulative
releases = release items
smooth = smoothstep()
AGI is cuming