Jorge Cardenas Mayor

[Jorge Cardenas Mayor @JCM_Chile]

@iuditg I move to anthropic

[Jorge Cardenas Mayor @JCM_Chile]

@adamc0dez Mmm try to ask to Claude itslef , probably he will find the issue. Could be that it has a system variable overridden the model picker from cli or a bug in the autodetection. Tip, ask him to search the web the last model if he Challenge You that the last model is opus 4.8

[Jorge Cardenas Mayor @JCM_Chile]

@maxedapps Try ultra code and explicitly tell it to use subagents , that's what worked for me. Also, I read in other threads that the system prompt says not to use subagents unless the user asks. No idea if that's true, but it would explain why so many people are reporting the same thing.

[Jorge Cardenas Mayor @JCM_Chile]

@kimmonismus Been going through the open GitHub issues and my guess is this is cache-related, plus context getting truncated. A lot of people are filing tickets about Sol looping on simple tasks after a compact, and there are other threads about missing cache hits. Looks like the same root.

[Jorge Cardenas Mayor @JCM_Chile]

@evi77ain 1.5 yesterday with sol high using security codex.

[Jorge Cardenas Mayor @JCM_Chile]

@0x45o repeat the same thing that probably said to 100k other users.

[Jorge Cardenas Mayor @JCM_Chile]

@GoogleAIStudio you should open ChatGPT-style app connections for Gemini across all tiers. You're already inside so many orgs through Workspace, but their tier blocks third-party apps. Let them see what AI plus third party tools can do. they'll upgrade on their own once they feel it.

[Jorge Cardenas Mayor @JCM_Chile]

@markproduct Probably, and it's way better than Twitter used to be.

[Jorge Cardenas Mayor @JCM_Chile]

@cheaf25master Depend on the tasks, i would prefer hybrid

[Jorge Cardenas Mayor @JCM_Chile]

@elonmusk @elonmusk everyone is racing for coding benchmarks, even the cheap Chinese models. build a llm w/ Grok 4.5 intelligence for real-world work at 1/4 the cost and you'd be god in education, logistics and services. That price point is the one thing standing between us and shipping.

[Jorge Cardenas Mayor @JCM_Chile]

@vansh22b Same here. I love Codex and pay for both (Pro x20 for Codex, Max x5 for Claude), but this is forcing me to switch and even to think about cancelling Codex altogether. Looking at Kimi maybe. Sad about it.

[Jorge Cardenas Mayor @JCM_Chile]

@sama @sama you reset yesterday and my whole weekly limit was gone in a day. Same audit that cost 10% last week. Sol high orchestrating, Sol low executing, nothing fancy. Love Codex, but at this rate it's unusable for us. Appreciate the resets, but we can't depend on them.

[Jorge Cardenas Mayor @JCM_Chile]

@footofwrath @0ttik1 @thsottiaux agree

[Jorge Cardenas Mayor @JCM_Chile]

The reality @0ttik1 is that the weekly usage feels like the 5 hours Windows. That is something to talk about. Now You put something more on the table. the problem is more token or less usage? I don't know , i didnt measure My consumption before 5.6 and don't have token right know 😢

[Jorge Cardenas Mayor @JCM_Chile]

@ouzdor Agree

[Jorge Cardenas Mayor @JCM_Chile]

@Meta @finkd i propose to you a challenge, try to create a instagram account with api access in less than a several hours and frustration, or a business account or anything, then try to recover your password. Really you should do the test.

[Jorge Cardenas Mayor @JCM_Chile]

@fridai05 @thsottiaux The better fix is not having to depend on resets at all. Before 5.6 I was banking 4 resets unused; now I depend on random ones because a full week of usage drains in a day with normal use. My setup is Sol high as orchestrator, Sol low to execute. Nothing fancy.

[Jorge Cardenas Mayor @JCM_Chile]

@kritikakodes reach