first time using oai models for coding. it's so refreshing…

Samuel

@samuel.fm

first time using oai models for coding. it's so refreshing to ask it simple questions and for it to reply immediately rather than using 300 tool calls and taking 30 minutes

8 5 60

Sam Clemente 16 Jul

Honestly I've been on the Codex train for a few months now

Redowan Delowar 16 Jul

I didn't think they had in them anymore. But sol 5.6 with ultra in fast mode consistently beat fable with max settings.

0x_meow 16 Jul

yeah i've come around to oai these last two months - the latest models are great and their usage limits are generous

1 2

0x_meow 16 Jul

Also (and this was the main reason tbh) i can use the subscription with any harness / coding product i want

1 7

jake 16 Jul

This is super important.

Daniel Stpete 16 Jul

Web URL please

1 1

jack 16 Jul

codex is so much slower than Claude tho what 😭

wnz.bsky.social 16 Jul

That's one of the reasons I switched several months ago.

Rowan-Paul 16 Jul

I wonder if this an infra thing or their models are just quicker

1 1

Samuel 16 Jul

it's definitely faster per token but the main thing is it doesn't try and deep research literally everything. I can ask fable what the current git branch is and it will take like 10 minutes

2 16

Hendrik Mans 16 Jul

Sol medium and up has some of that, too. But it feels to me like it's sort of bored by most "normal" coding work. (But yeah, both the agent harness as well as the models are much more fun to work with than Anthropic's right now!)

Kuba Suder 🇵🇱🇺🇦16 Jul

Wow…

Ariel M. (she/her)17 Jul

i used a local tiny qwen model (~2gb?; i've got 4gb vram lol) today for a few simple questions and yknow what? quick, straightforward. no frills. going to keep trying it for the small questions, see where it struggles vs is fine