PolicyEngine / PolicyEngine/microcosm

META: publication + validation portfolio to a well-validated populace launch

Open
#305 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
0
Forks
4
Avg merge
1d 3h
Merged PRs (30d)
94

Description

Agent routing (labels: tier:fable / tier:standard)

Heuristic: Fable designs and judges; Opus/GPT-class agents build and assemble. Delegate work whose failure mode is loud (tests fail, tables don't reproduce); keep on Fable work whose failure mode is plausible-and-silently-wrong (statistical semantics, claim-bearing prose, sensitive framing). Issues carry the label; acceptance tests are written into the standard-tier issues so the harness judges the output, not the agent.

Two standing Fable roles (instead of Fable-everywhere):

  1. PI/editor — reviews every claim-bearing PR (results sections, abstracts, site explainer copy) and runs an adversarial pass before any submission.
  2. Semantics guardian — changes to the following files require Fable review regardless of green CI: popdgp metrics/views/floors; every repo's holdout/split logic; every data loader's weight handling; promotion-gate thresholds and floor bands; populace-fit's model.py/qrf.py weight semantics. (This portfolio's three caught bugs — the SCF int16 overflow, harness tail-blindness, fragility-bolding semantics — all passed lint and tests while being scientifically wrong.)

Fable-tier items: popdgp#2 (paper design), populace#302 (holdout masking — load-bearing wall for every non-self-referential claim), populace#304 (system paper), calibration-paper results prose, composition interpretation, Belgium note review (#303, EUROMOD framing). Everything else labeled standard. Human-only: authorship, funding, #290 merge, Axiom-paper timing.

🤖 Generated with Claude Code

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

This is a portfolio and routing issue rather than a single implementation task, and it names no single entry point or test. Start by reviewing the listed items, including popdgp metrics/views/floors, holdout and split logic, data-loader weights, promotion gates, and populace-fit model.py/qrf.py. Done means the work is routed to the stated review tier and the publication and validation items are completed with the required reviews.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
data, documentation, machine-learning, testing-qa
Issue type
Documentation
Difficulty
5/5
Estimated time
Over a week
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.