Skip to content

Recover the nine lectures behind the English by state-sha, then merge #286, #294 and #296 #298

Open

Description

@mmcky

Nine lectures in this edition are behind their English source by state-sha, and three open sync PRs cannot build until one of them is recovered. Found by translate status at engine v0.29.4 on 2026-10-18 after the day's drain of the sync queue (fourteen PRs merged); the dashboard's date heuristic had shown only some of them. Evidence and causes: QuantEcon/project-translation research/2026-10-18-translate-status/ (private).

Recover by engine run, at the released @v0 (v0.29.4), one hand PR per group

Lecture Route Why it is behind
optimal_growth_uncertainty translate init -f new lecture, parity refusal — QuantEcon/lecture-python.myst#1072
social_security_depletion translate init -f new lecture, parity refusal — QuantEcon/lecture-python.myst#1092; #296 carries only its _toc.yml entry
sir_model translate forward -f translated, then PR creation failed — QuantEcon/lecture-python.myst#1051
var_dmd, svd_intro, measurement_models translate forward -f the sync of QuantEcon/lecture-python.myst#1055 failed on Bad credentials and filed nothing (QuantEcon/action-translation#367)
harrison_kreps translate forward -f the sync of QuantEcon/lecture-python.myst#669 failed on a missing anthropic-api-key input and filed nothing; the la.solve refactor never arrived
mccall_q translate forward -f the myst#760 wave (QuantEcon/action-translation#280 auto-close); #273 fixed the H1 by hand, but #760 also changed rng, heading case and two typos here
ak_aiyagari translate forward -f, or accept the hh → household rename (myst#706) was delivered by hand in #297 for the code; prose and comments may still differ, and the state sha is already current

State-sha advance only (content verified current, state stale): career, jv, odu (#760 changed only their H1, applied in #273) and likelihood_bayes (#248 delivered the default_rng change). One commit updating .translate/state/*.yml, as fa#160 did.

Then

Checks on every recovery PR

Structural parity against the source (directive sequence, heading map); the lecture executes in the preview and the preview log is read for Executing notebook failed, because this edition's preview reports build succeeded on execution errors (#271); the two known forward defects — model preamble and duplicate front matter (QuantEcon/action-translation#288) — and the zh-cn spacing rules (QuantEcon/project-translation#70).

Blocked on: an Anthropic API key available to the session that runs the engine (QuantEcon/infrastructure#47).

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions