Arsh

16 yo builder

[Arsh @be_arsh]

@MrAhmadAwais God please

[Arsh @be_arsh]

@CommandCodeAI command code is goated

Can we more free models pleaseeee

[Arsh @be_arsh]

@claudeai it’s mostly in your interest to get the masses to use Claude chat and pay it but not use Claude code

[Arsh @be_arsh]

@weswinder @neat_snap Yeah it’s a good plan, lots of discounted models.

[Arsh @be_arsh]

@weswinder @neat_snap Command code goat plan

[Arsh @be_arsh]

@ardent__dev Agents.md

[Arsh @be_arsh]

@justinsunyt (Blank company) for (thing other company does)

Once that blank company is your own

You know you are doing something right

[Arsh @be_arsh]

@mattyp That’s pretty cool

What I want is like I want to just be able to like create new agents that are like local without having like a worker already running in a specific directory in the iOS app

like I wanted to be more like the Codex Remote in the ChatGPT app

[Arsh @be_arsh]

@MrAhmadAwais New mathematical equation right there

[Arsh @be_arsh]

@Jason @bot Now just got installed a bunch of speakers microphones in my house. Create a wake word and then boom, we got Jarvis.

[Arsh @be_arsh]

@opencode Error from provider (Console): Upstream request failed: [invalid request_error] This request requires an end-
user identifier. Provide a non-empty 'safety_identifier' (or 'user') field.

bro

[Arsh @be_arsh]

@kentcdodds I only wish for better cursor usage limits

[Arsh @be_arsh]

@_clarktang Clearest effect ever

[Arsh @be_arsh]

@LyalinDotCom please fix antigravity desktop app, it seriously is infuriating to use, the model doesn't even DO what I ask, it just seems to do whatever it wants, and doesn't follow things I have said earlier in chat.

the model itself is fine, its the harness for me.

[Arsh @be_arsh]

@leodev thats literally how I see you now

[Arsh @be_arsh]

@rauchg waiting 500ms for python to boot every time an agent runs a tool was torture. 10µs is actually unhinged

[Arsh @be_arsh]

@leodev none of mine are sponsored

[Arsh @be_arsh]

@HARNESSROUTER @ycombinator makes sense after reading through. so the bulk is model-tier arbitrage (opus down to 5.2), but that 1.5x–2x delta on the same model shows how much tool schema bloat and uncompressed step history compound in multi turn loops. very solid case for harness routing.

[Arsh @be_arsh]

@theo t3 code just unifies harnesses in one UI??