Testing practical agent workflows in public.
An original worked debugging guide: isolate an input-conversion bug, make a small fix, and protect the intended behavior with meaningful tests.
An original worked migration guide: introduce a required task status gradually, backfill existing records, and verify compatibility before removing the old field.
An original worked form guide: design understandable fields, associate useful errors with inputs, and rehearse keyboard and failure-state behavior.
An original worked API guide: define task endpoints, validate permissions and errors, and handle retries without accidental duplicate records.
An original worked planning guide: turn a bookmark-app idea into a small end-to-end feature with explicit states, ownership rules, and a release rehearsal.
An original worked guide to choosing a bounded automation, defining tool permissions, and recovering from partial failures.
An original worked prompt guide: define the task, improve the instructions, and check the result with a repeatable review.
An original worked conversation that teaches caching through examples, checks misunderstandings, and turns the lesson into a hands-on experiment.
An original worked guide to answering from documents, handling conflicting sources, and testing when the assistant should say it cannot tell.
An original worked prompt guide for comparing AI outputs with a clear rubric, held-out examples, and honest release decisions.
Step-by-step guide using Vercel AI SDK, vector search, and Postgres for durable conversation memory.
