Outage resolved! ✅ Our systems and support teams are now available. Thank you for your patience and understanding.
Donegal Insurance Group’s Post
More Relevant Posts
-
SRE/DevOps expert, Software development and engineering manager, Entrepreneur with 15+ years of experience || Haifa, Israel / Sunnyvale, CA / Yerevan, Armenia
In recent months, I've been using ChatGPT for at least a couple of hours a day. Based on my experience, I can now predict an upcoming outage with accuracy. Here is the typical sequence of issues I observed last time, a couple of hours before the outage, while I was editing a list of 10 theses for my latest article: 1. Increased response time, which really slows down. 2. This is followed by errors that require regenerating results. 3. Then, it misunderstands what I ask about, even though I usually provide clear prompts and have good prompt engineering experience. 3. The craziest part: it forgets the context. For example, when asked to provide a complete text with 7 points, not 10, or when asked for all 10 points, one of them is something completely new, imagined by ChatGPT. 4. Forced downgrade to GPT-4.0 or, on one occasion, to GPT-3. 5. An official outage occurs a few hours later. One time it happened and later for a couple of days before the release of 4o #openai #llm #observability
Management consulting, technology, and go-to-market. AI, Cloud, Networking, and Blue Economy. Storyteller and explainer of the complex.
Thought this was a funny take on yesterday's outage. Thankfully, I usually have multiple online and local LLMs accessible at any given time. I am not sure I could handle using my own brain again ;-)
To view or add a comment, sign in
-
I just published My First Postmortem: The Great Widget Service Outage https://lnkd.in/d6etpx4K
To view or add a comment, sign in
-
After latest uptime havoc it will be interesting to observe if there would be some serious changes implemented in the update process like roll out speed and user category priority (and testing) and how it will affect the extent of the damage in the future.
To view or add a comment, sign in
-
console. log (018 + 015) ; console. log (NaN) ; console. log(3>2>1) ; console. log (0.1+0.2 ==0.3) ; what will be o/p of this logs?
To view or add a comment, sign in
-
DevRel at EmpathyOps | CNCF Ambassador | Associate Software Engineer at Hughes Systique | Prev: Community Manager, Hashnode | Web Dev | DevOps | 🎙️ Host | Communities and Open source
Hung up on PATs? Looking for some of that STS magic? Trying to avoid credential leaks? 🤓 Join us next week, May 23rd @ 11:30AM EST as we dive into prod practices around credential management & access to our source control systems, & pipelines. 📹 https://lnkd.in/gtYZHTqX
To view or add a comment, sign in
-
Notification of external audits can come at short notice and RTOs must always be ready to respond effectively and quickly in such an event.
To view or add a comment, sign in
-
ADD DEBUG AND TIMEOUTS AROUND COMMUNICATION WITH SSL SESSION ID CACHE DAEMON (SIDD) https://lnkd.in/eBVnbbQh
ADD DEBUG AND TIMEOUTS AROUND COMMUNICATION WITH SSL SESSION ID CACHE DAEMON (SIDD)
To view or add a comment, sign in
-
Is network troubleshooting the bane of your existence? Does MTTR going up get you down? Join me and Lawrence Popa for an Auvik Live Demo on June 12 at 2PM ET where we'll show you some of our top tips for troubleshooting network issues with the help of Auvik Network Management!
Why did the network technician bring a ladder to work? Because he heard the MTTR was too high! 😆 Just a cheesy joke to go with your cheese pizza at our live demo on June 12, 2024, at 2:00 PM ET. 🍕 https://bit.ly/4bMGXl0
To view or add a comment, sign in
10,924 followers