Connects structured model outputs and retrieval design to measurable application quality, with practical ways to inspect data and improve search.
Why read Jason Liu?
Liu is useful when an application retrieves plausible text but misses the information the user needs. His RAG improvement guide starts by testing retrieval separately, then uses metadata, search comparisons and specific user feedback to find what to fix. You get a sequence of experiments before choosing a more complex stack.
His context engineering essay shows how source details, full pages, tables and search facets can help a model navigate a collection. Read the examples as designs to test on your own data. He also created Instructor, which turns model responses into typed Python objects; validation of their shape still needs application-specific checks.
Start with the original
Selected work
-
Engineering guide ·
Systematically Improving Your RAG
How do you improve retrieval with evidence? Build a baseline, test full-text and vector search, add useful metadata and collect feedback that names the problem. The reported client measurements describe his examples.
Cite this selection
XFollowList, “Systematically Improving Your RAG” reading notes in the Jason Liu profile. https://xfollowlist.com/people/jason-liu#work-444c71a747f4
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
-
Engineering essay ·
Beyond Chunks: Why Context Engineering is the Future of RAG
What should a search response tell a model? Compare raw chunks with source metadata, complete pages, tables and facets. His progression gives you candidate designs to evaluate, rather than a guarantee of better results.
Cite this selection
XFollowList, “Beyond Chunks: Why Context Engineering is the Future of RAG” reading notes in the Jason Liu profile. https://xfollowlist.com/people/jason-liu#work-3cb3f73abf6e
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
Projects & roles
Instructor
The structured-output library he created, using Pydantic models, validation and retries around model calls. Maintained as an open-source project.
Community rating
Sources & further reading
Profile sources checked . This describes checking these references, separately from the publication dates of selected works.
- His current author site links @jxnlco and identifies him as Instructor’s creator. It describes his Codex developer-experience role at OpenAI as his own current account; old consulting offerings are closed.
- The Jason Liu byline, author-profile link and visible publication date establish authorship and the selected first read.
- Current public X metadata identifies jason and @jxnlco, consistent with the outgoing link on his author site.
Cite this profile’s source check
XFollowList, Jason Liu profile sources. Sources checked 2026-10-11. https://xfollowlist.com/people/jason-liu#sources-checked