secondstate

@LexnLin I do one off targeted critique reviews via fable sometimes thru Cursor and it generally costs 7-11$ per prompt lol I keep maxing out my referrals though, so least I get ~25-30 fable prompts per month I guess

omg I literally just made a 3d game and used 2 prompts: first ate 25% weekly (about 1,5h runtime) second ate 75% weekly (about 2h runtime) and it didn't finish 😭 both using fable 5 high Now I paid $10 for on-demand usage and told it to create a handoff.md really quickly. it ate $9 😭😭

@thdxr @dillon_mulroy i was hoping you can because I’ve been point my agent at opencode for a reason

@dillon_mulroy yeah agree it helps that i can't tell the difference between good effect code and slop effect code

it's nowhere near perfect, but using effect really does help cut down on the amount of slop these models produce

5.6 sol and its goofy summaries

@sec0ndstate yes in the end this kind of agent is just a websocket client :P

@gakonst certainly reminds me of how alloy looks :)

the API here is particularly fun because it resembles an Ethereum transaction send (submit a turn) & wait for receipt (wait for turn result) while a background event loop is receiving all the intermediate traces from the API and can then decide whether to stream to stdout or to write them to a very efficient tui etc.

follow along, i've copied 1:1 the core codex event loop, context management, codemode & tool calling for 5.6-sol, so that it can be embedded anywhere via an ergonomic library-first api also have bindings for wasm/node/web + python via pyo3 github.com/gakonst/nanoco…

@0x0000G @ethereumfndn agreed, was super sad to see that porto got discontinued.. slopforking doesnt work on its own because then it loses its ability to be easily usable across the ecosystem, but them taking it over would help that

It is outrageous that the @Ethereumfndn is not building something like porto.sh (or taking over it).

@vanilagy lmao wat

Mediabunny had a bug where nested encode workers would randomly stop working in Firefox. The solution: (💀)

@z0r0zzz @Alibaba_Qwen EZ bro, just quantize it to 0.0133 bit precision, it'll run on a raspberry pi

Qwen3.8 is launching and going open-weight soon!🌐 With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out. Can't wait to hear what you build. Stay tuned! 🚀  Token Plan international：qwencloud.com/pricing/token-… China：platform.qianwenai.com/pricing/token-…

@thsottiaux @Ananth7e ICYMI: there are scheduled tasks in codex!

@Ananth7e What happens if I go on a holiday for 3 weeks

omg.... tibo has reset codex limits 8 times in the past 9 days. anthropic could nev-

@dillon_mulroy @EffectTS_ beatings will continue until effect morale improves

WELL WELL WELL LOOK HOW THE TURN TABLES

@sec0ndstate with a sprinkle of orange

RED canonical belts and suspenders smoke test load-bearing papering over

@z0r0zzz check out my link x.com/z0r0zzz/status…

least favorite class of poster: the linkooor

@ThePrimeagen @opencode but then dax will have to work double shifts

@opencode Can you reset the limits every 19 hours?

2x the kimi k3 usage on opencode go for 1 week

@gakonst react-scan by the rerender fixer king himself: @aidenybai but react-doctor is super cool as well

there's some amazing tool that i've seen on the timeline which is about debugging react re-renders, can somebody find that for me? been looking all day

@thsottiaux pls keep it

We've had no 5h limit in Codex plus and pro for a few days. Do you think it is better or are you finding it difficult to manage the usage included in the weekly limit effectively? If we were to make this different, what should it look like in an ideal world?

@maria_rcks i raise my bid x.com/sec0ndstate/st…

Ok ok... here, I made it easier to understand

yes, terra is a meme

yes, but is it over 9000?

@heyandras If any skill mentions subagents/orchestration, it will do that, so you explicitly have to tell it not to, or cut it from the skill. I usually let it do 5.6 luna subagents to run builds/tests/ verbose crap, but tell it to either never spawn sol agents, or under specific terms.

GPT 5.6 Sol with High reasoning is so weird, it uses subagents all the time for not particular reason. The model is good, just this auto-subagents-everywhere-everytime is a bit annoying tbh.