Source the numbers that had no artifact, and put the units back - #173
Merged
Conversation
Three claims with nothing committed behind them, and one regression I introduced. svar/index published four shock probabilities and a two-figure revision decomposition that appeared on that page and nowhere else in the repository, while every other number on it cites current_forecast.json. Running latest_shocks at production draws confirms the four probabilities exactly — world demand 0.713, world energy 0.895, UK demand 0.809, monetary policy 0.840 — so they were right, just untraceable. The response is committed as papers/boe-svar/figures/latest_shocks.json and the page now cites it with its draw counts. The revision figures do not survive. The page claimed +0.27pp on GDP and +0.22pp on CPI for the 2026Q1 round; the paper reports +0.40pp and −0.22pp for 2024Q2 — a different quarter, different values, and the opposite sign on CPI. No artifact holds them, no exposed tool produces them, and their resemblance to the paper's numbers with a flipped sign is the profile of a figure that was transcribed loosely. Removed rather than reworded. obr/validation explained a six-fold gap with a fact that is not true. Its held-add-factor scorecard reports GDP 0.37% against the working paper's 2.2%, reconciled as the paper covering "its own longer window". The paper says "held constant, projected over 2026Q1–2027Q4" — the same window. The difference is the EFO vintage, November 2025 against March 2026, and the page now says so. The regression is mine. Moving the readings onto the topic pages took stored() with them, so the public-finances headline read "-15,989.0" where the page it replaced said "£16.0bn borrowed", and "£77.1bn" and "£749/week" stopped appearing on the site at all. Stored-exactly-as-published is the right principle for the method note and the fidelity claim, not for the first number a reader meets. display() renders scale and unit in the headline and keeps the exact stored value on the line beneath — and preserves the publisher's precision, so a stored 5.0 is 5.0% and not 5%. Twelve gates, 1107 tests. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Three claims with nothing committed behind them, and one regression I introduced.
svar/indexpublished six numbers with no sourceFour shock probabilities and a two-figure revision decomposition appeared on that page and nowhere else in the repository, while every other number on it cites
current_forecast.json.Running
latest_shocksat production draws confirms the four probabilities exactly — world demand 0.713, world energy 0.895, UK demand 0.809, monetary policy 0.840. They were right, just untraceable. The response is now committed aspapers/boe-svar/figures/latest_shocks.jsonand the page cites it with its draw counts (6,000 draws, 430 accepted, ESS 198).The revision figures do not survive. The page claimed +0.27pp on GDP and +0.22pp on CPI for the 2026Q1 round. The paper reports +0.40pp and −0.22pp for 2024Q2 — different quarter, different values, opposite sign on CPI. No artifact holds them, no exposed MCP tool produces that decomposition, and their resemblance to the paper's numbers with a flipped sign is the profile of a figure transcribed loosely. Removed rather than reworded.
obr/validationexplained a six-fold gap with something untrueIts held-add-factor scorecard reports GDP 0.37% against the working paper's 2.2%, reconciled as the paper covering "its own longer window". The paper says "held constant, projected over 2026Q1–2027Q4" — the same window. The difference is the EFO vintage, November 2025 against March 2026. The page now says so.
The units regression is mine
Moving the readings onto the topic pages took
stored()with them, so the public-finances headline read-15,989.0where the page it replaced said£16.0bn borrowed— and£77.1bnand£749/weekstopped appearing on the site at all.Stored-exactly-as-published is the right principle for the method note and the fidelity claim, not for the first number a reader meets.
display()renders scale and unit in the headline and keeps the exact stored value on the line beneath. It also preserves the publisher's precision, so a stored5.0is5.0%and not5%.Verification
Twelve generator and contract gates, 1107 tests.
🤖 Generated with Claude Code