Chris

Building games & animations with Grok • SuperGrokPro daily • Vibecoding multi-agent systems • Shipping real projects, not just prompts

[Chris @manxlab]

@matthewmillerai Don’t need Astra for computer use. Sol just as good in my experience. Vastly better than Anthropic with any model for computer use

[Chris @manxlab]

@vineerpasam All the time. And I also remind it when it was wrong, and tried to authoritatively push for a certain direction

[Chris @manxlab]

@bridgemindai Hopefully Opus 5.1 includes an attitude adjustment. Its personality is exhausting.

[Chris @manxlab]

@buildwithhassan I’ve started using 3.8 flash regularly for audits. It finds bugs all the time, is fast and cheap. Inter-model work is very useful.

[Chris @manxlab]

@sattyyouneed Taste + the ability to actually ask the agents for something of value, that operates efficiently and isn’t total slop. Getting models to actually achieve anything resembling excellence is entirely in the hands of the developer

[Chris @manxlab]

@hiarun02 No difference. Unless only Astra can do the work, I’ve switched back to Sol. ~1/5th the consumption

[Chris @manxlab]

@ziwenxu_ Ran 19 min task on Astra low = -5%; ran 17 min task on Sol med = -1%. They were the same task.

[Chris @manxlab]

@buildwithrajath Same for me. 5% in a few minutes of Astra light on 5x plan. Still drains crazy fast

[Chris @manxlab]

@Steve_Yegge My own custom one. Up to 32 panes across 2+ workspaces, endless AI models, orchestration, cross model delegation….

[Chris @manxlab]

@thsottiaux Got it

[Chris @manxlab]

@bridgemindai @OpenAI usage has been poor and my experience with Astra in general has been bad(one shot visual creations are not what I build) other subscriptions have really shined this week. Muse as a builder is a powerhouse, grok as a builder is a powerhouse, DS 4.1 powerhouse, Claude orchestrates

[Chris @manxlab]

@CommandCodeAI 🛬 time to …

[Chris @manxlab]

@bridgemindai @thsottiaux @OpenAI Interestingly Claude is also glitching

[Chris @manxlab]

@bridgemindai @thsottiaux @OpenAI My weekly usage % just dropped from 32% to 2% while I wasn’t using codex at all

[Chris @manxlab]

@bridgemindai @OpenAI Even delegating to other models and only running on low, Astra just guzzles usage credits

[Chris @manxlab]

@tokyo_roberto @rezoundous Even with orchestration and running on low the remaining % just DROPS

[Chris @manxlab]

@lightsilver323 https://t.co/Zp4xTua3wI

[Chris @manxlab]

@adhamuxi SuperGrok Heavy + Muse Code Power Usage gets close to no stress

[Chris @manxlab]

@thsottiaux Squeezed it down to 2% before the reset. Near perfect timing!

[Chris @manxlab]

@MarcosHernanz It is super fast. I wish it would get stuck less and it does make quite a few errors. For the price it is very good.