Source item

outages and notices about rate limits at multiple model providers, if you needed more proof that inference at (growing) scale is hard