Tibo, focus on real-world value, not model assumptions, overconfidence, or blind trust in benchmarks and tests.
I'm building a real project, and even with clear instructions, detailed explanations, and examples, the model repeatedly does the opposite of what I ask. It's wasting time instead of saving it.
Back when I used Codex 5.3, AI genuinely helped me get real work done. That was when AI felt useful and worth paying for.
I also wonder whether OpenAI runs experiments on selected users and whether my account could be part of one. I have no evidence of that, but the inconsistent performance makes me question it. Users deserve transparency if experiments are affecting their experience.
We need reliable AI that delivers actual results, not just impressive benchmarks.
@thsottiaux Codex context handling has gotten significantly worse.
I ran "/compact" at 77% context usage, and it only dropped to 73%.
When Astra first launched, it was amazing. Long-running tasks are where Codex finally understands what I want. Cache hits mean little if context management still hurts.
@Manisha27493225@OpenAI Agreed. This isn’t about judging reasoning styles. In real projects, agents lose architectural decisions over long sessions and handoffs. Users find fixes, like developer/reviewer cross-checks. Reward reproducible failures and workflows proven to improve outcomes..
@OpenAI, I have a proposal.
AI doesn't improve through research alone. Real users also discover weaknesses, solve reasoning problems, and develop better ways to work with AI.
Shouldn't users who make meaningful contributions receive recognition and rewards?
@OpenAI, if meaningful user contributions help improve future AI systems, consider rewarding verified contributors with extended Plus/Pro subscriptions, credits, or recognition.
Users aren't just consumers of AI. Some actively discover how to make it better.
Let's recognize that value.
@OpenAI Privacy must come first.
Users shouldn't have to expose confidential source code, proprietary projects, or intellectual property to receive recognition.
Participation should be voluntary, with clear consent and fair evaluation.
If you’ve been using ChatGPT and Codex since the early days, you already know this isn’t just a Claude vs OpenAI thing. There have always been ups and downs. Limits change, models change, and companies make decisions users don’t like. I don’t care much about the temporary noise, buddy. I care about one thing: does the tool still help me think, build, solve problems, and get real work done? If yes, I’ll keep using it. Rough phases pass. Be positive, be productive, and use your brain for something useful instead of feeding it garbage and outrage all day lol. Judge the value over time, not one bad week..
Thanks for being upfront before launch. I think the core point is getting buried under API economics, so here is the user side question.
For a $200 Pro user, the new plan gives roughly half the API-equivalent usage of the old one. Today, that is less for the same price.
The rest of the post is about how it improves later. Cheaper models and better efficiency can mean more work per dollar over time, but they don't replace capacity that is reduced now. Unannounced features can't be valued until we see them. And no 5h limit is welcome, but it improves flexibility, not the total allowance.
It looks like Pro is moving closer to API economics, whereas before it was a more heavily subsidized bundle. If that is the direction, saying so plainly would help a lot.
The clearest comparison would be:-
Old $200 Pro: how much real work?
New $200 Pro: how much real work?
Same models, same representative workloads..
The API dollar figure matters less than one practical question: does the same $200 get me more, the same, or less work today? Real usage numbers would build more trust than asking us to translate API pricing ourselves.
@BlueDragonJuno@ATMFL80@thsottiaux Nothing hypothetical about 45% remaining + a scheduled reset in 3 days. The courtesy reset gave me 55% more now, but delayed my next 100% by 4 days. That’s the entire point..
@cheesealloverme@thsottiaux Using quota just because it might suddenly disappear isn’t a (strategy). That’s exactly the problem. A weekly allowance should be usable when needed during the week, not rushed through because an unexpected reset might wipe the remainder..