The bit about MCP servers beating fixed-pipeline tools for academic research really clicked for me. I've been running a similar setup but with three search providers wired together through a single Claude Code skill. Firecrawl for breadth, Exa for semantic matching, Synthetic.new for full-text. Covered the whole setup here if you're curious: https://reading.sh/how-to-build-a-solid-research-pipeline-in-claude-code-ff7878c5e2b5
Thanks for the shoutout! This is a fascinating emerging area. Working on my next set of experiments...
So incredibly helpful! Thank you for going deep.
The bit about MCP servers beating fixed-pipeline tools for academic research really clicked for me. I've been running a similar setup but with three search providers wired together through a single Claude Code skill. Firecrawl for breadth, Exa for semantic matching, Synthetic.new for full-text. Covered the whole setup here if you're curious: https://reading.sh/how-to-build-a-solid-research-pipeline-in-claude-code-ff7878c5e2b5
The "Alfred Wallace's skill design" link seems to be broken
Fixed. Look at https://www.linkedin.com/posts/alfredhwallace_better-ai-is-not-about-fetching-more-facts-activity-7420923355597561856-QjxR?utm_source=share&utm_medium=member_desktop&rcm=ACoAAADydNkB8cFBZ48tGuFjE4yHqQ6N6mBWEyE