same shit happens to me each second day it says that then it allows me to send 3-10 more messages and repeats the same shit again...

EDIT: i have pro+ for 1 year and i didnt signed for weekly limits... it wasnt of terms of service when i was subbing

Unfortunatly it concerns all models. Like the global rate limit issue from earlier

Yeah, this is a known issue on Copilot, unfortunately. Best way is to pivot to claude or codex extension for dev work.

I've given up on using Copilot after so many problems. It's astonishing how something new happens every month that makes it hard to work, even though I'm paying for it. I decided to switch to Claude Code Max for about $200, and literally all my problems are solved. It sounds like a generic sales pitch, but it's true: goodbye rate limits, goodbye problems with a specific AI model being in high demand, and on top of that, better code quality.

We're all getting duped. Folks are still.cashing in not knowing.

When I opened support.github.com to submit a new ticket, a virtual assistant popped up asking if I wanted a refund. I guess GitHub does have some self-awareness about this. But I have zero intention of forgiving them. I feel like my money was stolen, and I will NEVER use Copilot again 👎.

They only just changed the date again, before it stated the 21st. Mind you they want payment on the 19th, lol

We'll see if they remove the limit though

So yeah, BS.

I should have come back sooner to say, but after about 10mins it started working again. I did 'upgrade' like it requested so I'm guessing that does seem to alleviate it for a bit? Hard to know, feels like one of those things that could be manipulated to farm upgrades 🫥😵‍💫

Thank you for informing us, @smart-trust . I had considered purchasing another subscription with a different account, but that doesn’t feel like a reasonable or professional solution.

I’m genuinely concerned about the lack of response from Microsoft and GitHub on this issue. It reflects poorly on their communication and overall user experience.

Despite the frustration, we are still dependent on GitHub and its AI tools for our workflows. That said, situations like this highlight the growing need for more reliable and competitive alternatives in the market.

This looks like a standard response. They’re sending the exact same message to everyone.

Yep, it came within 5 mins of ticket submission so it absolutely is.

I noticed. I have to be very precise and mention every little detail. It doesn't think out side of the box and neither does it come up with improvements. Furthermore, it doesnt rethink about its own implementation and therefore ignores mistakes it makes. This is due to that selecting auto also lowers reasoning. Its very noticeable and annoying

Thing is, prior to all this i was building an agentic harness in theory (not knowing exactly but having the idea) while auto is inconsistent, if you already have proper semantics, memory, protocols in place for any of your framework needs, you can leverage that against the lack of intelligence. Doesn't excuse anything but it will give you all a chance to keep working. Codex just started working for me today so maybe things are back up. I havent tried copilot specifically outside of auto for the last 5 days and watching these forums and using Claude simultaneously till it gets hit with limits.

I digress, my harness has helped tremendously with this crappy github situation

I noticed. I have to be very precise and mention every little detail. It doesn't think out side of the box and neither does it come up with improvements. Furthermore, it doesnt rethink about its own implementation and therefore ignores mistakes it makes. This is due to that selecting auto also lowers reasoning. Its very noticeable and annoying

The reasoning levels differ, the inconsistency is the switch itself since some models properly recall and some dont. Some follow instructions verbatim and some skip steps.

This is why I mentioned making your own harness around copilot or whatever you are using.

It also helps to tack on a set of instructions, like a conversational starter kit to ensure they always do their due diligence from the start which is another method I use as well and its proven super efficient especially with this issue. That and modifying Claude superpower and creating starter kits has saved Hella time.

Holy shit guys I think I know why they're doing this!

They must be collecting funding from investors just now!!
Because the ammount of subscriptions sold will 3x!
😂😂😂

Fair enough, but unless you're gonna sue them there's no real impact being made. They did this deliberately and vaguely explained and what not, so I agree find an alternative. I was just helping those who had issues bearing with what is until we see wtf is actually gonna happen. I actually want to see how heavy the blowback is. I wonder if they'll try to cancel non-payment during this time because I definitely plan on riding this out with my card locked, lol. 🤣 I have claude as backup so im not really worried.

However I also want to say, think about your current issue, dependency. Once you remove that it won't matter what "sub" you pick, you'll have better control of the overall outcome. Even with all this going on, my work suffered maybe 30% in productivity and thats a vague guess from the simple fact codex is glitchy and I have to do more monitoring because auto uses some models that still hallucinate and I havent fully mitigated that on longer sessions so I dont use autopilot.

which is ??