Ron Paul

Part-time blockchain developer, new Web3 dApp tester, and self-motivated researcher. HODL: Azuki, MAYC, CloneX, mfers, BEANZ, WGG,

[Ron Paul @RonPualS]

@yacineMTB DeepSeek harness and official API

[Ron Paul @RonPualS]

@pvncher What’s missing is a proper file handoff protocol and a way to pass approvals between tasks (e.g., via auto-review).

Instead, it keeps interrupting me and forcing me to manually confirm things in a separate Codex task.

[Ron Paul @RonPualS]

@petergostev Their “permanent caching” is basically just waiving the uncached-token charges.

[Prof. Feynman @ProfFeynman]

Study hard what interests you the most in the most undisciplined, irreverent, and original manner possible.

[Ron Paul @RonPualS]

@scaling01 ...multi words in one token...

[Ron Paul @RonPualS]

@thsottiaux Wow.. keep cache forever

[Ron Paul @RonPualS]

@Angaisb_ Yes. A reset later I guess.

[Ron Paul @RonPualS]

@leodev I think you’re overthinking this, dude. Removing resets wouldn’t magically prevent us from losing access. So why get rid of them? At least with resets, we can get more work done when everything is running normally.

[Ron Paul @RonPualS]

@thsottiaux @AbdoKerdawy Not shipped yet?

[Ron Paul @RonPualS]

@thsottiaux Not shipped yet. When?

[Ron Paul @RonPualS]

@thsottiaux Maybe this is why some users are skeptical of these on-paper usage increases. Even if the overall quota goes up, the 5-hour limits could still prevent people from actually using it all.

Think about that.

[Ron Paul @RonPualS]

@thsottiaux What if the 5-hour cap is so tight that hitting your weekly limit becomes practically impossible?

At that point, is the weekly allowance a real limit, or just a theoretical number?

[Ron Paul @RonPualS]

@adonis_singh Not really.

[Ron Paul @RonPualS]

@testingcatalog We cannot measure the real TPS in codex app now.

[Ron Paul @RonPualS]

@daniel_mac8 We need an official TPS counter in the Codex app.

[Ron Paul @RonPualS]

@TokenGremlin Have you tested it yet?

[Ron Paul @RonPualS]

@rockatanescu A ticket-based workflow seems like a reasonable way to help agents break down large development goals into smaller, verifiable steps. If decomposition and intermediate validation reduce costly rework, why shouldn’t agents work this way?

[Ron Paul @RonPualS]

@ConnorTalksAI There’s a bit of a problem here. The new version got noticeably slower after release, and now that the speed has been restored, we’re being told it’s an “improvement”? Does that really count?

[Ron Paul @RonPualS]

If TPS is what you’re trying to improve, shouldn’t Codex show the actual TPS for each task? Without that metric, how are we supposed to verify the improvement?

Also, the numbers don’t add up: if the previous result was ~20 TPS, a 50% improvement gets you to ~30 TPS, not 50.

Show us the TPS counter, then we can talk about the improvement.

[Ron Paul @RonPualS]

@lkuczborski In the bottom of the open menu.