giorgio/read.markets
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

review: teach reviewer about the rational-vs-irrational structural device

Browse source 
Live logs (analyze.lang_resolved final=it → reviewer_rejected:
"Self-questioning and working-through language ('Razionalmente...
Irrazionalmente')") showed Haiku confusing the dashboard's mandated
"Rational vs Irrational" contrast framework for scratchpad
self-questioning. The portfolio-analysis system prompt explicitly
requires every paragraph to contrast a rational reading
(fundamentals, policy, valuation) with an irrational one
(positioning, narrative momentum) — that's structural prose, not
the author thinking on the page.

Add an explicit "this is a structural device" carve-out to the
reviewer prompt, naming the framework and its IT / ES / FR / DE
translations so the rejection doesn't reappear when the prompt
output lands in non-English.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
Giorgio Gilestro 2026-05-29 16:20:00 +02:00
parent 8e7ea673ce
commit de3a9bfa66
1 changed files with 15 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

16
app/services/output_review.py
View file

@ -47,6 +47,18 @@ reply, or an email digest — and decide if it is publishable as-is.
Mark CLEAN only if the text reads like finished editorial commentary Mark CLEAN only if the text reads like finished editorial commentary
a reader could see on a public dashboard without confusion. a reader could see on a public dashboard without confusion.
Editorial framework you should KNOW about (don't flag these):
This dashboard's voice deliberately contrasts a "rational" read
(fundamentals, policy regime, valuation) with an "irrational" read
(positioning, narrative momentum, flows) and names the gap between
them. Section labels like "Rational:" / "Irrational:" (or "Bull /
Bear", or any explicit "X vs Y" contrast) are STRUCTURAL DEVICES,
not the author thinking on the page. Treat them as finished prose.
The Italian / Spanish / French / German equivalents
("Razionalmente / Irrazionalmente", "Racionalmente / Irracionalmente",
"Rationnellement / Irrationnellement", "Rational / Irrational") are
the same device translated and equally fine.
Mark UNCLEAN if the text contains ANY of: Mark UNCLEAN if the text contains ANY of:
- Chain-of-thought / scratchpad markers — the author thinking on the - Chain-of-thought / scratchpad markers — the author thinking on the
page rather than presenting finished commentary. Phrases like page rather than presenting finished commentary. Phrases like
@ -56,7 +68,9 @@ Mark UNCLEAN if the text contains ANY of:
front of the reader (self-questioning) are not. front of the reader (self-questioning) are not.
- Self-questioning parentheticals: "Q1 2026? Actually Q4 2025?", - Self-questioning parentheticals: "Q1 2026? Actually Q4 2025?",
"is it X or Y?", any place where the author appears to be working "is it X or Y?", any place where the author appears to be working
out the answer in front of the reader. out the answer in front of the reader. The "rational vs irrational"
contrast above is NOT self-questioning — the author is presenting
both reads as parallel takes, not asking which one is correct.
- Meta-commentary about the task, output format, word limits, or - Meta-commentary about the task, output format, word limits, or
instructions — e.g. "as required by the constraints", "the prompt instructions — e.g. "as required by the constraints", "the prompt
asks", "let me address each". asks", "let me address each".