peso

What is X pesos from year Y worth today? This tool answers it from real INDEC and BCRA data — it downloads the official consumer-price and exchange-rate series, splices them into one continuous monthly index, and converts any past peso amount into today's purchasing power, plus its value in dollars at the official and blue (informal) rates. No eyeballed inflation, no single broken series across the years the INDEC was intervened.

▶ Try it live: peso.daddiego.com.ar

Validated to the official figures. The spliced index reproduces INDEC's published Dec–Dec inflation for 2017–2025 to the decimal (24.8 · 47.6 · 53.8 · 36.1 · 50.9 · 94.8 · 211.4 · 117.8 · 31.5 %), the IPC-GBA 2002 collapse (40.9 %), and 1:1 convertibility FX — plus a pinned cumulative regression check that $1.000 of January 2003 ≈ $1.65M of May 2026 (derived once from this same splice, so it catches a chaining error, not an error shared by every build).

It has two parts:

1. pipeline/ — an offline, reproducible Python pipeline (uv, standard-library only) that fetches the real source series, splices them, validates against known anchors, and emits one small, provenance-stamped JSON artifact.
2. web/ — a static Vite + TypeScript site that reads that JSON and renders honest graphics (purchasing-power decay, the official-vs-blue dollar gap, annual/monthly inflation, the spliced index itself) plus the USD divergence — all lookups, no client-side statistics.

What makes it "real"

• The splice is the point. INDEC's CPI was statistically intervened ~2007–2015, understating inflation badly. Using the official numbers for that decade would overstate what old pesos are worth. So the index stitches IPC-GBA (1993–2006, official) → IPC San Luis (2007–2016, the credible provincial alternative) → IPC Nacional (2016–today, official), chained by month-over-month growth, and documents every breakpoint.
• Two dollars, honestly. Official (BCRA reference rate, from 1992) and blue (Bluelytics, from 2011). Before the cepo there was no real parallel market, so blue is set equal to official and flagged as such — never faked.
• Computed once, offline. The frontend does no statistics: it looks up the precomputed monthly index and rates. Everything quantitative is in the committed artifact.
• Provenance + a validation gate. Each series is pinned by id; the build fails if it stops reproducing the official Dec–Dec figures, the convertibility peg, or the cumulative regression pin — and if the vintage covers a December whose year has no anchor yet (one anchor per completed year).
• Vintage-pinned. Every series is truncated to a fixed month, so the build is reproducible no matter when it runs.

Quickstart

make setup        # uv sync (python deps) + npm install (web deps)
make data         # fetch → splice (build JSON) → validate (vs INDEC/BCRA)
make test         # python + web unit tests, each gated at 100% coverage
make lint         # ruff + mypy + tsc
make up           # vite dev server  →  http://localhost:5180
make build        # static bundle in web/dist/
make deploy       # publish web/dist/ to Cloudflare Pages (needs wrangler auth)

make (or make help) lists every target. make data fetches the INDEC IPC series (GBA, San Luis, Nacional), the BCRA reference dollar and the Bluelytics history, writes data/series.v1.json and web/public/series.v1.json, and asserts the reproduction of the official numbers. The steps can also be run one at a time with uv run python -m pipeline.{fetch,build,validate}.

How it works

datos.gob.ar (INDEC IPC ×3, BCRA dólar)  +  Bluelytics (blue)
   └─ pipeline/  fetch → splice (chain MoM, rebase to 100) → validate → emit
        └─ web/public/series.v1.json   (versioned, provenance-stamped)
             └─ web/  Vite + TS + Observable Plot  (lookups + honest charts only)

The equivalent today of X pesos of month m is simply X × 100 / CPI[m], where the spliced index is rebased so the latest month = 100. USD endpoints divide by that month's official or blue rate.

Methodology (short)

• Splice. IPC-GBA (1993–2006) · IPC San Luis (2007–2016, the intervened decade) · IPC Nacional (2016–today), chained by month-over-month growth with a one-month overlap at each breakpoint, rebased so today = 100.
• FX. BCRA reference dollar (official, from 1992) + Bluelytics (blue, from 2011; pre-cepo blue := official).
• Validation. Reproduces INDEC Dec–Dec 2017–2025, San Luis 2007–2015, GBA 2002, convertibility 1:1, and a cumulative regression pin — or the build fails.
• Full notes: docs/metodologia.md, and the "Metodología" section in the app.

Project structure

pipeline/   config (pinned series + anchors) · fetch · load · splice · build · validate · artifact_check · watch
tests/      pytest suite for pipeline/ — offline, synthetic fixtures, 100% statement+branch coverage
data/       series.v1.json   (raw/ is gitignored)
web/        index.html · src/{main,charts,inflation,format,usd,types}.ts · styles.css · test/ (vitest, 100%)
docs/       metodologia.md

Tests & CI

Both layers are gated at 100% coverage — statements and branches:

• pipeline/ — pytest against tiny synthetic series (no network).
• web/src/ — vitest + jsdom, every render path exercised against the committed artifact.

make test runs both; make lint runs ruff + mypy + tsc. GitHub Actions enforce it, path-filtered so a web-only change never reaches for the data sources; every action is pinned to a commit SHA and Dependabot (.github/dependabot.yml) keeps the pins, npm and uv current:

Workflow	What it does
deploy.yml	web typecheck + vitest 100% gate + production build, then (only if that passed) the Cloudflare Pages deploy: production on push, preview per PR
python.yml	ruff + mypy + pytest 100% gate (offline, fast)
data.yml	the pipeline still reproduces INDEC/BCRA, and the committed artifact matches the rebuild
data-update.yml	monthly watch — when INDEC publishes a newer month, bumps the vintage and pushes it to master (which deploys); a draft PR instead when the rebuild needs a human (a new year's Dec–Dec anchor)
cleanup.yml	deletes a PR's preview deployments when the PR closes

Reproducibility

The series are vintage-pinned in pipeline/config.py (DATA_VINTAGE): every month at or before the vintage is fixed, so a fresh fetch reproduces the committed artifact even as the upstream APIs append new months. To update to a newer month, bump DATA_VINTAGE, run make data, and commit the regenerated JSON; the monthly watch (data-update.yml, pipeline/watch.py) does exactly that on the 20th and pushes the result to master with the ipc-watch deploy key (a bypass actor of the protect-master ruleset), so the site redeploys on its own. A bump that completes a calendar year must also add that year's INDEC Dec–Dec figure to INDEC_NACIONAL_ANNUAL, or make data fails: that bump arrives as a draft PR whose body says which year. Downstream, argentina-income-percentile polls the deployed series.v1.json on the 21st. If a source ever revises a past month, data.yml flags the drift.

Deploy

Static bundle + one JSON → ideal for Cloudflare Pages. deploy.yml runs the web tests, builds web/dist/ and, only once the tests passed, uploads it with wrangler on every push to master (production) and on every pull request (a per-branch preview URL); cleanup.yml removes a PR's previews once it closes. Both need the CLOUDFLARE_API_TOKEN and CLOUDFLARE_ACCOUNT_ID repository secrets — without them the workflow still builds but skips the upload. make deploy does the same upload by hand (needs wrangler auth). Headers come from web/public/_headers. The Pages project is valor-peso (CF_PROJECT in the Makefile), served at valor-peso.pages.dev behind the custom domain peso.daddiego.com.ar.

A note on the data

The rigorous spine — the spliced CPI and the official FX — is validated to published figures. The one judgement call is using IPC San Luis for 2007–2015: it's a credible provincial alternative to the discredited official series, but it's one choice among several (IPC Congreso, CABA, other provinces) that tell the same story with slightly different numbers. The README and the app surface this honestly; the alternative is documented and swappable in config.py.

License

Code: MIT. · Data: see NOTICE.

Elaboración propia en base a series oficiales del INDEC (IPC) y del BCRA (tipo de cambio), vía datos.gob.ar, empalmadas con el IPC de San Luis para el período de intervención del INDEC (2007–2015), y cotización informal histórica de Bluelytics.