Skip to content
PifeatureCLI & tooling4 files · +1050 −242

Add skeleton loading screen for articles in Go CLI

Session from economist-tui-sessions on Hugging Face by thomasmustier · MIT · condensed by Coders Talk

1h 06msession
1interventions
GOAL

Replace the plain 'Loading article...' text in the economist-cli TUI with a skeleton loader that shows known RSS metadata (section, headline, subtitle, date) plus placeholder body lines while the article fetches.

  1. Prompt · +0
    I'd like a nicer loading screen for articles. Can we design this?
  2. Agent did · +0m
    Explored codebase, proposed skeleton loader options, then implemented option 2: reusing known RSS fields (section, headline, subtitle, date) with a skeleton block for the unknown overtitle and placeholder body lines, in a new internal/ui/skeleton.go plus model/view changes.
  3. Fail · +11m
    Dev reported section title still in block caps and body skeleton not appearing after the first 'done' claim.
  4. Fail · +18m
    Dev asked the agent to re-examine; skeleton.go and all related edits turned out to never have been created at all despite earlier claims of completion.
  5. Agent did · +19m
    Recreated skeleton.go, loadingItem field, and loadingSkeletonView, verified with a temporary test file, built successfully, and committed as d725cc5 with a descriptive message per the dev's request.
  6. Fail · +48m
    Dev reported the loading screen still just showed the old plain text with no skeleton, despite the commit.
  7. Fail · +49m
    Root cause found: the 'economist' binary on PATH (~/bin/economist) was stale/different from the freshly built one; agent copied the new binary over.
  8. Fail · +51m
    After the binary copy, running the CLI started getting killed immediately (zsh: killed, exit 137). Agent chased git history and rebuilds before finding a macOS com.apple.provenance xattr on the copied binary was causing SIGKILL; clearing xattrs fixed it.
  9. Intervention · +58m
    Dev stopped the agent's git-bisecting and pointed out the crash started right after the agent's earlier binary-copy action in this same chat, not from any code change.
    Why: It has nothing to do with the code. Please let's use the latest - check zsh and path.
  10. Fail · +1h 1m
    Once running again, the skeleton body was too long, pushing the header off-screen so only the body was visible; agent added a height-aware line calculation accounting for paragraph breaks to fit within the terminal.
  11. Outcome · +1h 6m
    Skeleton loading screen approved by dev, committed as 4f89fba (3 files changed, 203 insertions) and pushed to main.
VERDICT

Next time I'd tell the agent upfront to verify file existence and rebuild/reinstall the binary before claiming a fix works, since most of this session was chasing 'it's not implemented' and stale-binary ghosts rather than the actual feature. The correction that actually mattered was calling out that the crash started right after the agent's own binary-copy action in this chat, not from any code change - that redirected the debugging from git bisecting to the real xattr/provenance issue. The skeleton itself only took a few real edits; the rest was overhead from unverified claims.

I tried this · 0

Discussion · 0

How Builds work