Update TODO.md and ROADMAP.md for accuracy post-v2.4.0

[igerber]

Summary

• Fix stale line references in Known Limitations (estimators.py wild bootstrap, predict())
• Rewrite Large Module Files table with accurate line counts; add 4 missing >1000-line files (imputation, two_stage, linalg, estimators), remove bacon.py (now 979 lines)
• Mark NaN Handling section as DONE (all 7 t_stat locations fixed)
• Expand safe_inference() migration table from ~20 sites / 9 files to 25 sites / 11 files with exact line numbers
• Add "Tech Debt from Code Reviews" section (15 items from PRs Add Windows wheel builds using faer pure Rust linear algebra #115, Remove jackknife variance estimation from TROP #124, Fix TWFE within-transformation bug and add methodology review #139, Add Borusyak-Jaravel-Spiess (2024) Imputation DiD estimator #141, SDID methodology review: rewrite to match R synthdid + Rust parallel variance #145, Sun-Abraham methodology review: fix 5 issues, add R benchmarks #153, Add Two-Stage DiD estimator (Gardner 2022) #156, Add tutorial notebook for Two-Stage DiD (Gardner 2022) #159)
• Fix utils.py line reference in RuntimeWarnings section (481 → 515)
• Bump ROADMAP.md version from v2.3.0 to v2.4.0

Methodology references (required if estimator / math changes)

• N/A - documentation-only changes, no estimator or math code modified

Validation

• Tests added/updated: No test changes (documentation only)
• Backtest / simulation / notebook evidence (if applicable): N/A

Security / privacy

• Confirm no secrets/PII in this PR: Yes

---

Generated with Claude Code

[github-actions]

Overall Assessment: ⚠️ Needs changes

Executive Summary

• Doc-only PR; no estimator/method/inference code changes.
• Version bump and line-count updates match repo state.
• Two documentation accuracy issues remain: a methodology-registry mismatch statement in TODO and the safe_inference() migration command/table inconsistency.

Methodology

• P2 — Impact: TODO.md:L96 claims the Methodology Registry says TwoStageDiD/ImputationDiD bootstrap should support Mammen/Webb, but the registry sections for those methods state Rademacher-only multiplier bootstrap. This misstates methodology-source alignment and can misdirect follow-up work. — Fix: Update the TODO entry to reflect the actual registry text or cite the correct document that claims Mammen/Webb. References: docs/methodology/REGISTRY.md:L548-L613.

Code Quality

• No findings (docs-only change).

Performance

• No findings.

Maintainability

• No findings.

Tech Debt

• No findings beyond the documentation mismatch noted above.

Security

• No findings.

Documentation/Tests

• P2 — Impact: The “How to find them” command in TODO.md:L67-L70 doesn’t reproduce the table in TODO.md:L53-L65: it misses dict-based assignments like diff_diff/staggered.py:L777 and diff_diff/sun_abraham.py:L661, and it includes non-target lines such as diff_diff/utils.py:L178 (the safe_inference definition) and diff_diff/utils.py:L641 (inline inference not listed). This makes the migration list hard to verify. — Fix: Update the command to match ['t_stat'] assignments and document exclusions, or adjust the table to include diff_diff/utils.py:L641/clarify scope; consider using rg -n "t_stat\\s*=|\\['t_stat'\\]" diff_diff/*.py | rg -v "safe_inference" or similar.

[igerber]

/ai-review

[github-actions]

🔁 AI review rerun (requested by @igerber)

Head SHA: 00c4616f45f6c588a1b18e9666f288b195efe26e

---

Overall Assessment: ✅ Looks good

Executive Summary

• Docs-only updates in ROADMAP.md and TODO.md; no estimator/methodology code changes or registry edits.
• Verified updated line counts and limitation references align with current file contents (e.g., estimators.py and large module counts).
• One minor doc portability nit in the new grep command for locating inline inference sites.

Methodology

• No findings. This PR only edits documentation and does not touch estimator logic, assumptions, or docs/methodology/REGISTRY.md.

Code Quality

• No findings. Documentation-only changes.

Performance

• No findings.

Maintainability

• No findings.

Tech Debt

• No findings. New section is informational only.

Security

• No findings.

Documentation/Tests

• Severity: P3
  Impact: The helper command uses \s with grep -E, which is GNU-specific and can fail on BSD/macOS, so contributors may miss inline inference sites.
  Concrete fix: Replace \s with [[:space:]] in the regex, or switch to grep -P and note the PCRE requirement.
  TODO.md:L68-L71