Great question.
I don't use LLMs for writing.
I use agents extensively for brainstorming, research, checking facts, handling markup, finding references, indexing data, and so on.
But I think that asking people to read LLM-generated text breaks a kind of social contract.
@nathankpeck@SeaNmWhite@mauerbac Facts. The quality of slack messages tends to be poor. I wish we used more Email for important threads on my team. But when I started doing it, nobody replied ๐
@borisfyi@awscloud@AWSAmplify That's a good suggestion. Although security headers can negatively impact if they block necessary content. Particularly the CSP policy which restricts the browser to load resources (scripts, images) only from the same origin as the page. Might be overly restrictive as a default.
@AllenHeltonDev Allen, I can't imagine how hard this must be for you and your family right now. Please know that you are not alone. Don't hesitate to reach out if you need anything at all.
@AWSbrett @mikepb78@richdevelops@algolia@meilisearch You can also run Meilisearch on Fargate, but it does rely on disk so youโll need EFS mount to persist the data. I wonder if you can use AppRunner since it scales down to *almost* zero
@jeffragusa@JustJake My bad. I glanced at your profile and thought you worked for Google. But my point is that not documenting limits is frustrating for customers. Not doing it for security reasons is a bit of an excuse because you can figure them out anyways by finding when it throttles.
@jeffragusa@JustJake You are not wrong but your answer confirms the sentiment from everyone on this thread: An attitude of blaming the customer for a service limit that is not documented.
@JustJake They probably have a non-public RPS limit or burst rate on top of the quota. Or said differently, just because you are allowed 18K requests per minute doesnโt mean you can use all of it in the same second. You have to pace it evenly across the 60 seconds.
@yuyokk@AWSAmplify Hi Iurii, Fernando here from the Amplify team. This doesn't sound normal. I sent you a DM to find out more details and see if we can help.
@QuinnyPig That would be setting the observability bar too low. โ If we had to rely on the status page as an indicator that something is not working. Then it means we got our observabilty really wrong.