Just got a $1k bill from an orchestration platform we use for running ~20k jobs on OUR OWN infrastructure. Guess how long we are gonna keep that subscription going...
@cjzafir I loaded up $5 to test some vibing on a toy project. Worked for several hours completely refactoring the thing and debugging some weird rendering and perf issues. Finally went to look at my balance, thinking I needed to top it off… $4.88 left.
@mitsuhiko it’s good to know that companies the size of Microsoft are still dealing with optimistic updates combined with eventually consistent elasticsearch. I always felt exceptionally dumb when I caused bugs like that 😅
@itsharmanjot This needs a healthy dose of reality. Their license is much more restrictive than you depict. Self-hosting Nango is quite limited and specifically for hobby projects. And the license prevents circumventing any of this. Their right, of course.
@AWSSupport I can send a private message, but I have been chatting via the web chat with your support. We are on business+, and now your support agent has completely ghosted us for an hour. This is UNACCEPTABLE.
Hey @AWSSupport, I have filed a business critical issue and your support system is consistently failing. Every time the call comes through, your system says it experienced a technical issue connecting us and you are now outside your SLA for responding.
@0xblacklight@FireworksAI_HQ I was running into it quite a bit with just one agent running. Made me decide not to convert to the new firepass v2 plan, unfortunately.
@catalinmpit You gotta use oppositional models. For example, if gpt wrote the code… it should not be reviewing it. Beyond that, I tend to pull the branch into my working directory, reference the PR, and ask the agent to review it in tandem with me.
@TheScholastics I tried to buy a copy of Maya 20ish years ago from a third party seller. Got a burned cd with sharpie on it. Called the police with the address of the seller in NYC and they laughed at me. Said no cops were going into that hood 🫠
Oh dang, the memory cards I use in my camera have tripled in price, if you can even find them in stock anywhere. Haven’t really payed too much attention to the memory shortage lately, guess I should have. Yikes.
Feel like models prefer that you use "main" instead of "master" in git. Constantly having to wait a few turns before coding agents figure it out. Easy fix in agents.md, but wonder if this gives insight into prioritization of data freshness.
If you have an agent that takes screenshots for UI regression testing and work proof, you better make sure it isn't just hallucinating those screenshots. Just caught one of ours making screenshots up out of thin air. Not even sure how that happened yet.