Toots from 2023-05-28#
“The reality is that this tech is hard. LLMs aren’t magic, you won’t solve all the problems…#
“The reality is that this tech is hard. LLMs aren’t magic, you won’t solve all the problems in the world using them, and the more you delay releasing your product to everyone, the further behind the curve you’ll be.”
https://www.honeycomb.io/blog/hard-stuff-nobody-talks-about-llm
Mastodon Source 🐘#
“The agile question is one I’m most proud of, because agile hasn’t meant anything concrete …#
“The agile question is one I’m most proud of, because agile hasn’t meant anything concrete for a long time beyond “who needs a plan when you can get interrupted and reorged at any time by whim of managerial fiat.”
https://exple.tive.org/blarg/2023/05/27/brown-mms/
Mastodon Source 🐘#
“Our solution combines caching, deduplication, erasure coding, and sparse loading to take advan…#
“Our solution combines caching, deduplication, erasure coding, and sparse loading to take advantage of our needs. Without adding any customer visible complexity…we were able to achieve our scale and cold-start latency goals, while having significant headroom for future scaling.”
Brooker, Marc, Mike Danilov, Chris Greenwood, and Phil Piwonka. 2023. “On-Demand Container Loading in AWS Lambda.” ArXiv [Cs.DC]. arXiv. http://arxiv.org/abs/2305.13162.
Mastodon Source 🐘#
“Approximately 80% of newly uploaded Lambda functions result in zero unique chunks, and are just re-uploads of images that had been uploaded in the past. This appears to be primarily driven by automated testing and deployment (CI/CD) systems. Of the remaining 20% of functions that create at least one unique chunk (and therefore aren’t just trivial reuploads), the mean upload contains 4.3% unique chunks, and the median 2.5% unique chunks”