The Merstradamous 🇦🇱🇺🇲

80k in Codex credits ($10k) gone in 12 hours with absolutely nothing completed. CODEX IS A GARBAGE HARNESS OPENAI IS A GARBAGE AI COMPANY WITH GARBAGE MODELS I AM ABSOLUTELY DONE WITH OAI

So it looks like Codex has brought back the 5hr limit to the pro 20x account but they renamed it the weekly limit 🤦

6hrs of Codex usage used up 88% of my weekly $200 pro 20X limits. I am done with Codex and Openai It was fun while it lasted

Used up 20% of my weekly usage since the latest Codex reset which was 90 minutes ago. @thsottiaux I thought you were doubling usage with the new codex upgrades/updates not cutting it in half further. Also just to give a bit more context, I'm still on the pro 20x that still has about another week before your new $200 plan takes effect. Once the new $200 plan takes effect that would mean I will use 40% of my weekly usage limits in 90 minutes. There is no way I keep my 4 $200 accounts if they get 200 minutes of usage each. There's literally no reason for anyone to pay $1/min when there's other providers out there that are just as good or better for much much cheaper.

Was watching the Tesla roadster reveal live on YouTube last night and Elon was saying live and that Livestream "to celebrate the roadster reveal I've launched the roadster website and you can send your BTC eth Sol there and we'll send you back double" That's when I came to X to watch the real reveal 😂 It's actually crazy because I thought that YouTube live stream was legit real. It didn't look AI whatsoever. It's crazy and scary how good AI is becoming

Ok I burned 70% of my Codex pro 20x weekly limit in 4 hrs 🤦 time to try @opencode and their free deepseek 4.1 flash deal

Wtf is going on with Codex usage limits? How do I go from burning 8% of my weekly limit per hr to burning 13% per hr after the latest global "reset"? Did my pro 20x plan already get cut in half?

Codex has got to be the absolute worst harness in existence. It's actually comical how bad it is

This laptops limits are 46 agents running in parallel. Now I need to get my other laptop, my Mac mini and my VM running agents simultaneously as well and have all my agents and devices coordinating the shared mission together

Why is it that when I use a banked reset my usage drains more than 2x faster 🧐 Idk what's going on with OAI but this product is becoming sht. Everyday I'm noticing something worse with them I hate it that Anthropic is really the only other option right now because I can't stand them Hopefully Gemini 4 Argon drops soon and hopefully it's at least equivalent to 6.1 Sol 🙏

Is it just me or did Oai lose their Aura at an exponential rate? Whatever happened to pacing?

So crazy how a week ago, the most beloved @OpenAI has become hated like @AnthropicAI It goes to show that the consumer ain't taking no B.S from anyone and can very easily switch to a new competitor very easily, especially since all of these AI "labs" are progressive at insane speeds No company has a moat anymore and messing with the user will only leave your company and your investors holding a trillion dollar bag of garbage

I'm realizing more and more every day how incredibly horrible the Codex harness is

The argument that @OpenAl can offer significantly fewer tokens and lower TPS because its models are "more token efficient" misses the point. Token efficiency and subscription limits are two different things. If a model can accomplish the same work with fewer tokens, great. That efficiency should translate into more value for the user-not become an excuse to give the user fewer tokens. If every efficiency gain is followed by tighter limits, the user never actually captures the benefit. The provider does. Real efficiency should mean: same subscription → more work → more throughput → more value. Not: better efficiency → lower allowance → same or less usable capacity. If your models become 2x more efficient, users should effectively be able to accomplish ~2x more-not have the limits cut until the improvement disappears. Otherwise "token efficiency" is just a margin improvement being marketed as a user benefit.

Would be great if subagents could wake the root/main chat when they need attention instead of just dropping queued messages that sit idle until the user manually wakes it. @ChatGPT @sama @thsottiaux Right now the alternatives are pretty wasteful: keep the root constantly polling or set 15 minute heartbeats just to check whether a subagent finished or needs something That's unnecessary overhead. Subagents should be able to trigger the root only when there's actually something worth waking it for. Event driven > polling

Damn this is fire 🔥

Free dots today while you provide the data that makes them valuable. Then tomorrow, they charge you to access the intelligence your activity helped create, tighten the limits, and raise the price. The user supplies the data, the feedback, the activity, and ultimately much of the value then gets sold access back to it. This playbook is getting old. People deserve technology that works for them, belongs to them, and becomes more valuable for them the more they use it. That's a big part of why I'm building what I'm building. This model needs to change