tehKluma

Streamer • Developer • Community Builder | @ggteammidnite Outreach | @NeonChat Owner | @VeloraTV Community Manager • Velora Creator | @BisectHosting Partner

[tehKluma @tehKluma]

I’ve had enough of people tagging me, replying to me, or showing up in my comments trying to convince me to stop using Velora.

I’ve seen the posts. I’ve seen the screenshots. I’m aware of the controversy. You do not need to keep sending it to me.

I’m continuing to use Velora. That is my decision.

People can criticize the platform, leave it, or refuse to support it. That’s their choice. But I’m not joining a pile-on over old drama every time something from the past gets dragged back up.

You make your choice. I’ll make mine. Please stop tagging me about it.

[tehKluma @tehKluma]

$8 paid with XMoney xD lmao

[Velora.tv @VeloraTV]

We have changed the way people think about streaming and creation - and this is just our first year.

Here's to year two.
We're all in this together. 💛

[Velora.tv @VeloraTV]

A message from Founder/CEO @corylegendre for Velora's 1st birthday! 💛

[tehKluma @tehKluma]

Anyone got $6 bucks I can throw in to keep my Premium here? so i can go on the tangents you all know me so well for?

Thanks!
XMoney is same
CashApp $kluma
Venmo KlumaTV

Thanks!

[Team Midnite | TeamMidnite.tv @ggteammidnite]

Happy First Orbit @VeloraTV from all of us at @ggteammidnite

Crew:
- @computerwhiz15
- @MotherMedic818
- @tehKluma
- @MsChromatic
- @TheAussiePanda

[tehKluma @tehKluma]

@ChatGPT @OpenAI @sama 8 back and forths CEO in the trenches time??? Someone gotta look at my concern here
12 mins is not enough to do anything with in Codex

Big shout-out to @antigravity for an almost 4 hour 20 min session of coding and no lockouts because I used up my usage here it also gave me Claude for 20 mins or so after

[tehKluma @tehKluma]

LMAO no @OpenAI at half the price we'd see more compute, less lollygagging of your usage

[tehKluma @tehKluma]

Hey @AnthropicAI — genuine question for the people building Claude Code: how does your coding usage hold up during sustained repository-level development?

I'm asking because I just had a Codex session burn through my available 5-hour usage allowance after roughly 12 minutes of actual development work, followed by hours of waiting for the reset.

I don't expect unlimited compute, but can Claude Code reasonably handle a multi-hour development session involving repository investigation, multiple files, testing, and iteration without hitting the equivalent of “welp, see you in five hours” after 12 minutes? 😅

Because at this point I'm genuinely comparing which $20 subscription gives me enough usable coding time to actually finish the damn thing I'm working on.

[tehKluma @tehKluma]

Hello, @openai @chatgpt

Thank you for reaching out to OpenAI Support about the unexpectedly rapid Codex usage consumption and the impact it has had on your development workflow.

We understand how disruptive it can be to reach the Codex usage limit after a short period of repository-level development, especially when it prevents you from completing an active workflow. We also understand your concern about not having enough visibility into how quickly your available usage is being consumed.

Codex usage is not determined by elapsed time alone and can vary based on factors such as the model being used, reasoning level, task complexity, context length, Fast mode, image generation, and whether multiple sessions are active. As a result, a complex repository-level task involving code investigation, multiple files, testing, and iteration may consume the available allowance at a different rate than a shorter or less complex task.

We also appreciate your feedback about clearer real-time usage information, advance warnings before reaching the limit, more predictable capacity for development tasks, and clearer communication when model changes occur. We will pass this feedback on internally. While we can’t provide implementation timelines or ongoing status updates, your input greatly helps us continuously improve our offerings.

If you have any other questions or concerns, please let us know. We're here to help.

Best,
Shiela
OpenAI Support

Prior message

Codex usage limits are making the product impractical for sustained development. I'm a ChatGPT Plus subscriber paying $20/month. I use Codex on Windows 11 and authenticate directly through my ChatGPT account. I'm not using the API, an API key, or separate access-token billing. In my most recent session, I started using Codex at approximately 3:48 AM and hit the usage limit at approximately 4:00 AM—about 12 minutes of actual use. I was then told I couldn't use Codex again until approximately 8:48 AM. I understand that the five-hour window is an accounting window rather than five guaranteed hours of runtime. However, from a customer's perspective, being able to exhaust the included usage in roughly 12 minutes and then losing access for the remainder of that window makes Codex extremely difficult to use for sustained software development. The task itself wasn't unusual. I was working on a real repository and asked Codex to investigate and implement changes involving multiple files. That's exactly the type of work I subscribe to an agentic coding tool to perform. Repository-level development naturally requires the agent to inspect existing code, understand dependencies, make changes, test them, and sometimes iterate when the first implementation exposes another issue. The current usage model makes that iteration difficult. I now have to think about whether allowing Codex to investigate a problem thoroughly might consume enough of my allowance that I won't be able to test the result, correct it, or continue working afterward. A coding agent being thorough shouldn't effectively penalize the user by ending the rest of their development session. This is also noticeably different from my previous experience with ChatGPT Plus and Codex. In the past, I could work through meaningful development sessions: implement a feature, test it, make corrections, and continue onto other tasks without repeatedly encountering limits that stopped my work. Recently, I've found myself spending significantly more time dealing with usage restrictions and waiting for resets. The lack of predictability makes the problem worse. Before starting a repository-level task, I have no meaningful indication of whether it will consume a small portion of my remaining allowance or nearly all of it. By the time I find out, Codex may already be unavailable for several hours. For comparison, I also use Google Antigravity at a similar subscription price. I can often work with it continuously for several hours—sometimes 4–5 hours—which makes it much easier to complete an actual development session. This has reached the point where I'm seriously considering cancelling ChatGPT Plus and continuing my development work with Antigravity instead. I don't expect unlimited AI compute for $20/month. I do expect a paid coding product to provide enough predictable capacity to complete meaningful work, particularly when repository-level agentic development is one of its primary use cases. ChatGPT Plus and Codex previously provided enough usable capacity for me to do that; my recent experience has not. I would strongly encourage OpenAI to reconsider how repository-level Codex work consumes the included allowance. Increasing the included capacity would help, but there are other possible approaches as well: reducing the cost of repository investigation and iteration, allowing a lower-capacity model after the primary allowance is exhausted instead of imposing a hard multi-hour lockout, or providing much clearer real-time usage information before and during an agent task. Most importantly, users should be able to see when a task is consuming an unusually large amount of their allowance before it results in losing access for several hours. I'd also appreciate confirmation that the usage in this particular session was counted correctly. Exhausting my available Codex usage after approximately 12 minutes is extreme enough that I would like to know whether this was expected behavior or whether something caused my allowance to be depleted incorrectly. I want to continue using Codex. The issue isn't that limits exist; it's that the current limits and lockout behavior are making it increasingly difficult for me to actually finish my work.

[tehKluma @tehKluma]

One more thing — credit where it's due: @Google, thank you for somehow managing to give me access to your latest AI development tools without making me afraid that 12 minutes of actual repository work is going to consume my entire usable allowance. 😂

That's really the comparison I'm making here. I don't expect unlimited compute from anyone. I expect a paid developer tool to give me enough predictable capacity to actually use the “latest and greatest” model for development.

A Ferrari isn't particularly useful if it has enough fuel to get out of the driveway and then tells me to come back in five hours. 😭

[tehKluma @tehKluma]

One additional note for the Codex team: when the latest and greatest AI model is capable of doing more complex development work, but that same capability burns through the available usage so much faster that I can use it for less actual development, that starts to feel backwards. A Ferrari isn't very useful if the fuel tank only holds enough gas to get out of the driveway. 😅

I don't expect unlimited compute. I just expect a paid coding product to provide enough usable capacity to finish a normal development session without having to wonder whether asking it to investigate something thoroughly is going to lock me out for the next several hours.

[tehKluma @tehKluma]

@OpenAI @ChatGPT @sama @btaylor @SueDHellmann @zicokolter
12 mins is insane! someone better look into this!

[tehKluma @tehKluma]

@MotherMedic818 @ggteammidnite Holyfire lol mainly

[Team Midnite | TeamMidnite.tv @ggteammidnite]

🌙 Calling all the insomniac gamers! 👀🎮

Since sleep apparently isn’t on tonight’s quest list… what are y’all playing at this ungodly hour? 😂

Might as well get some XP while we’re awake. 💜

[tehKluma @tehKluma]

@MotherMedic818 @ggteammidnite That's understandable still enjoy it just not as much 🤣

[tehKluma @tehKluma]

I have a want to get this out there more
@NeonChat is my platform I'm trying to get on Google PlayStore for all of anyone to join
Please if your interested please let me know

Love to have you tryout the Android app

[tehKluma @tehKluma]

https://t.co/dxhgQ8RWEa

Hey anyone wanna new bank???
@VaroBank is looking for new victims jk jk awesome bank

[tehKluma @tehKluma]

@Fallout New Random Encounter: The Hash-Singing Slasher

Arrives at your C.A.M.P. at 3 AM
Hostile until supplied with potatoes
Uses your cooking station without permission
Sings continuously
Produces +25% quality hash browns
Refuses to elaborate
Leaves

Bethesda, ship it. 😂

[tehKluma @tehKluma]

THE HASH-SINGING SLASHER 🎶🔪🥔

He doesn't murder you. He breaks into your CAMP at 3 AM, makes hash browns, sings the entire time, and leaves.

Honestly, that's just a Fallout 76 random encounter waiting to happen. 💀
@Fallout

[Sonny on 76 @Sonny_on_76]

Waiting on my new friend, The Slasher @Fallout #Fallout #Fallout76

[tehKluma @tehKluma]

@ItsLethalTTV X really said your posting privileges have been revoked. Please return to bed and try again tomorrow. 😭