AI platform updates and developer tooling #81

Today's Letter

1. AWS, Bedrock Managed Knowledge Base GA
2. Google Search, connected apps support for AI Mode announced
3. Vercel runtime logs add cache miss reasons

AWS, Bedrock Managed Knowledge Base GA

• AWS made Amazon Bedrock Managed Knowledge Base generally available on July 16, 2026.
• The service provides managed agentic retrieval for enterprise data, including scaling and retrieval operations.
• Setup in the AWS Management Console does not require model selection, with default settings for first retrieval.
• Advanced controls remain available for embedding models, rerankers, chunking strategies, and related tuning.
• It includes six native connectors: Amazon S3, SharePoint, Confluence, Google Drive, OneDrive, and Web Crawler.
• Real-time ACL checks run at query time on top of pre-retrieval filtering to keep source permissions current.
• Managed parsing supports visual documents up to 500 MB, audio up to 2 GB, and video up to 10 GB.

Source: aws.amazon.com

Google Search, connected apps support for AI Mode announced

• Google said AI Mode in Search will support securely linked third-party apps and services.
• The feature is positioned as a way to interact with go-to services directly from Search, rather than switching across separate apps.
• The announcement was published on July 16, 2026 on the official Google blog under Search.
• Google highlighted AI Mode as the entry point for these linked-service interactions.
• Entities referenced in the source materials include Instacart and Canva, indicating the rollout is aimed at common productivity and commerce workflows.
• The post frames the change around secure account linking and cross-app interaction inside Search.
• Google did not provide pricing, usage limits, or a detailed rollout schedule in the supplied source text.

Source: blog.google
More: techcrunch.com · pulse2.com

Vercel runtime logs add cache miss reasons

• Vercel added Cache Reason details to runtime logs to show why a request was not a fresh cache hit
• Reasons are shown for cacheable CDN responses, including ISR, Partial Prerendering, and functions that set Cache-Control headers such as stale-while-revalidate
• Dynamically rendered responses on every request do not include a cache reason
• Documented status and reason pairs include MISS for Cold or Request collapsed, BYPASS for Draft Mode or Crawler, STALE for time-based revalidation or tag-based invalidation, and REVALIDATED for tag-based deletion
• The Logs tab now shows the reason alongside cache status for an individual request, with links to the cache status and reasons reference
• The same field is available from the CLI through vercel logs for per-request inspection and vercel metrics grouped by cache_reason for traffic-level analysis
• Vercel also exposed the feature through the Vercel plugin's cdn-caching skill for broader cache debugging workflows

Source: vercel.com
More: unit42.paloaltonetworks.com

Jocoletter curates AI, software, and product trends for developers and builders.

#AWS #Google #Vercel

Subscribe to Jocoletter