haruka

Built one of the top voted Discord bots on (80k+ servers). Now building GuildMate & @ReRide_AI.

[haruka @SilentHacks0]

SimCord is out!
https://t.co/hkJtofinSx

You can mock/test your https://t.co/EtIvI8kvVZ bot against a simulated Discord backend

This lets agents verify their commands, buttons, modals etc *actually* work, without needing to use Discord!

[haruka @SilentHacks0]

@thsottiaux I couldn’t seem to get a skill to invoke a subagent properly

[haruka @SilentHacks0]

@thdxr Man that’s a relic of ye old days I’d forgotten about

[haruka @SilentHacks0]

@0xSero My benchmark for harnesses is if I can scroll in tmux with it and I can’t figure it out with omp

[haruka @SilentHacks0]

@Angaisb_ Hope it’s a clear winner over Opus 4.8

[haruka @SilentHacks0]

@BenjDicken Big 19 already, they grow up so fast

[haruka @SilentHacks0]

@roanoke_gal Nice rec on the timeline, taking a look at this now too

[haruka @SilentHacks0]

@CommandCodeAI How about a usage reset hehe 🫣

[haruka @SilentHacks0]

I like how Claude code can get notified from a background task. So many other harnesses get stuck polling

[haruka @SilentHacks0]

@mcwangcn So it says you’re covering costs out of your own pocket, I wonder how long you will be able to sustain that 🫠

[haruka @SilentHacks0]

@ChrissGPT Wait what the full thing and not a lobotomised spark variant

[haruka @SilentHacks0]

@morganlinton What about the price per task? I think GPT 5.5 can be more token efficient and overall be cheaper than 5.4 @ a given task

[haruka @SilentHacks0]

@Angaisb_ I don’t get how it’s different to their original strategy. Like they already had a staggered release before, letting certain corporations have early access

[haruka @SilentHacks0]

@argofowl I wish there was some consistency between effort wording across all models, annoying sometimes to remember “max” vs “xhigh”

[haruka @SilentHacks0]

@AndrewCurran_ So this is the “weird reason” it was delayed?

[haruka @SilentHacks0]

@umi33563 @teortaxesTex @_xjdr Yeah what that’s not something you can just pass off in one sentence

[haruka @SilentHacks0]

@CanopyWave_AI From what I’ve read, you get a quota of “fast” inference and after that you’re stuck to these rate limits. No thing as a free lunch 😔

[haruka @SilentHacks0]

@AiBattle_ Hope it comes with vision

[haruka @SilentHacks0]

@argofowl Polymarket prediction ended up being more accurate 🫩

[haruka @SilentHacks0]

@synthwavedd Booo

[haruka @SilentHacks0]

@matrix_build Looks interesting, can I try it out?