Alexis Román, founder. We build products, audit the systems they run on, and secure the writes and the commits along the way. The factory's product is measurement: applied computational physics and AI metrology, shipped as instruments anyone can re-run — and as products people pay for.
I work on failure diagnosis at the systems level: CI pipelines, LLM inference serving, and evaluation harnesses. My working rule is that a number that cannot differ is not evidence — every claim below carries the object that produced it, and every gate I ship fails closed when it has nothing to measure.
No adivinamos: medimos — we do not guess, we measure. Platino Labs is an applied computational-physics laboratory that sells software: the discipline that decides whether a numerical scheme converges is the same one that decides what a product page is allowed to claim. The public instruments further down are the pedigree — and every one of them publishes its negative results as results.
Full record: Open Source Diagnostics — a measured record
The instruments further down are the hook. This is the product line they come out of: Studio City, a booking platform for boutique studios, and two verticals that run on the same booking core.
| Product | What it is | Live route | What the site charges |
|---|---|---|---|
| Studio City | booking platform for boutique studios — hourly reservations, owner dashboard, per-hour schedule | studiocity.app | Básico $299/mes · Pro $599/mes · Premium $899/mes |
| Barber Labs | the barbershop vertical: appointment system built on the Studio City booking core | studiocity.app/barber-labs | the same plan ladder — one booking core, one price list |
| Padel Labs | the court-slot vertical: scheduled play built on the same core | studiocity.app/padel | the same plan ladder — one booking core, one price list |
Each route is checked against a negative control on the same host: an unregistered slug renders the
generic template, a real product renders its own title. Measured 2026-10-31 —
/ → Studio City — La plataforma integral para estudios boutique · /barber-labs →
Barber Labs — Sistema de citas para barberías | Studio City · /padel → Padel Labs · Studio City ·
control /zzz-control-negativo-xyz → Estudio · Studio City.
The same rule that governs the instruments governs the products: what is written is what a command can be pointed at. Commercial figures for the line — studios registered, studios on a paid plan, currency collected — are not published here, because this page does not measure them. They live in the evidence room with their command, or they are not claimed.
The entry price, measured — not quoted. On 2026-11-01 the public pricing surface publishes three plans and one headline, and nothing else:
curl -s https://studiocity.app/ | grep -o '\$[0-9]\{3\}/mes' # → $299/mes
curl -s https://studiocity.app/ | grep -o -E 'Básico|Pro|Premium' # → the three tier namesBásico $299/mes · Pro $599/mes · Premium $899/mes, headlined «Prueba gratis · Planes desde
$299/mes · Sin permanencia · Sin comisiones». The entry price is therefore $299/mes, and that is
the only one this page publishes.
A launch condition of $99/mes for the first three months circulates internally and is not
published on the site — so it is not published here either. Two public sources quoting different
numbers for the same product is precisely the defect this page is built to refuse: a prospect who
reads $99 here and $299 there does not doubt the price, they doubt the laboratory. If that
condition ever ships on the site, it ships here in the same commit, and never before.
One factory, three doors. The same apparatus sits behind all three, so the doors are not marketing labels: each one names a different object being measured, and each one can be checked from outside before anyone talks to me.
| Door | What you get | What you can check before the first call |
|---|---|---|
| Build — SaaS | a booking platform and the verticals that run on it: hourly agenda, owner dashboard, price per slot | three live routes whose titles survive a negative control — see What we ship above |
| Audit — diagnosis | CI and supply-chain failure diagnosis, priced per finding, delivered as a report whose every graph is pinned to a commit | live service page · evidence room, 36 assets pinned to a commit |
| Secure — the writes | a pre-flight gate that proves the identity of the destination before writing, and commits that must declare which model wrote them | git-preflight-gate fails closed when it cannot measure · platino-ai-gate rejects Model: none on real code |
What is not in that table, and why. The cloud-spend audit and the credential-history scan run
inside the private corpus: there is no public repository for them and no client case published yet, so
they are listed here as capability, not proof — declared, and NO_MEDIDO from where you are
standing. A page that shows three doors and hides which two have no published case is a brochure; this
one carries the boundary in the same paragraph as the offer.
The arsenal. Everything in this table is public, downloadable, and runs without talking to me — that is the whole point. These are the instruments the audit door above is built from, and they are already pointed at my own repository, in public, with their red days intact.
| # | Repository | What it is |
|---|---|---|
| 1 | auditor-toolkit |
the free pre-flight kit: catches the false green before it reaches the shard — case study and sample deliverable inside |
| 2 | git-preflight-gate |
pre-flight gate: verifies the identity of the destination before writing, and fails closed when it cannot measure |
| 3 | ast-channel-router |
routes by extracted AST structure into declared channels |
| 4 | rf-signal-integrity |
S-parameters and Z₀ against a criterion sealed at 1.0e-4 before the first number |
| 5 | fotonica-monotile |
FDTD over the Hat monotile — contrast declared NEGATIVE / SUSTAINED NULL |
| 6 | acustica-monotile |
acoustic bandgaps over the same tiling — 22 of 29 bins inside its homogeneous null |
| 7 | la-auditora |
the audit service itself: supply-chain and CI failure diagnosis, published as a live page |
| 8 | platino-data-room |
the evidence room: every audit graph pinned to a commit, released as an index |
| 9 | platino-ai-gate |
two git hooks that make every commit declare which model wrote it — and reject Model: none on real code |
Rows 7 and 8 above are not method, they are the product. If you want the audit rather than the essays: the service page is the door, and the evidence room is what the audit hands you — 36 assets, 22 public graphs and 13 ingestion runs, every one pinned to a commit, so a claim can be re-run instead of believed.
| What it is | Where |
|---|---|
| The audit — CI and supply-chain failure diagnosis, priced per finding | la-auditora · live page |
| The evidence — the pinned index behind it | platino-data-room · arch-index-20261029-batch2 |
Before evaluating models I measured systems where a false positive is not a bad metric, it is a published result. Seven instruments are public, and every one of them publishes its negative results as results — the criterion that fails is the one that stays sealed:
| Instrument | What it measures | The object that proves it |
|---|---|---|
fotonica-monotile |
2D TM Yee-FDTD scheme over the Hat monotile; noise floor declared as a bound (Q1..Q4 green) | contrast reported NEGATIVE / SUSTAINED NULL — 28 of 29 bins inside 3σ + riz, failure run 1 vs 3 required · DOI 10.5281/zenodo.23329943 |
acustica-monotile |
acoustic 2D solver; bandgap bounds over the aperiodic Hat monotile |
T(ν) does not separate from its homogeneous null control in 22 of 29 bins · DOI 10.5281/zenodo.23329941 |
rf-signal-integrity |
RF S-parameters and Z₀ against a criterion sealed at 1.0e-4 before the first number |
four nominal meshes FAIL C1.1 and the criterion was not moved — the domain was declared defective (dx ≤ 8.25e-5 m); sweep verdict NO PASA / CONTROL NO DISCRIMINA |
plasma-ade |
1D ADE-FDTD cold plasma; wideband pulses | 9 criteria, 6 met, 3 failed and declared; C4a frozen as a metrological result (solver_plasma_cw_crudo_1649.txt) · red seals 16 → 0 |
gw-rw-zerilli |
1D Regge–Wheeler / Zerilli leapfrog and 2D kick-drift hyperbolics: convergence order and perimetric absorption | Richardson p = 2.0028629063990215 (O(h²), tolerance 0.05) and R_dB = −45.931354463740675 dB under the sealed −40 dB; F3, C3b, C3a and c4 published RED / NOT MEASURABLE / REFUSED |
optica-kerr-2d |
2D TE Kerr propagation, object vs falsifier | object ROJO at 13.11 % against a sealed 0.03, and the χ³ = 0 falsifier passes on both sides (beam diffracts 19.06 % ⇒ the solver cannot confine by itself) |
sfqed-paper |
strong-field QED paper and its measuring apparatus | 35-file verifiable bundle (34 declared + the manifest that cannot declare itself), one-command gate, 7-case must-fail bench that measures the verifier from outside — plus four declared unsealed items (MOTORES.md §6) and its own methodology and boundaries |
The rule did not change when the object changed from a field to a benchmark: a number that cannot differ is not evidence. Nor did it change when the object turned out negative: a criterion that fails is published with its raw capture and its command, and the threshold is never moved to make it pass.
A controlled 50-problem comparison on one corpus, two arms:
| arm | setup | result | cost |
|---|---|---|---|
| A | frontier reasoning model, no scaffolding | 50/50 correct | 0.265545000 |
| B | lightweight model + fail-closed structural gate (validators + bounded retries) | 50/50 correct | 0.023394983 |
Accuracy ties; cost does not. Measured ratio 11.3505× — arm B costs 8.81 % of arm A. The gate itself was a non-factor in repair (0.2000 retries per problem): the structure lived in the answer format, not in error recovery.
Verified artifacts (all three DOIs resolve, checked 2026-10-19):
- Software
platino-corev0.1.1 — 10.5281/zenodo.23459450 - Measured corpus v0.1.1 — 10.5281/zenodo.23288831
- Paper M5, pre-registered prediction (fulfilled at 768 s) — 10.5281/zenodo.23363121
Two shards, [sim_wh_n150] and [sim_bh_p150], were red across many unrelated open PRs. Almost every
author was reading the failure as their own.
Root cause: tests/ttnn/unit_tests/operations/transformers/test_prefetcher_worker_pipes.py calls
device.load_sub_device_manager(...) without the @skip_for_slow_dispatch guard that its sibling
test_prefetcher.py carries, producing
TT_FATAL @ sub_device_manager_tracker.cpp:118: Using sub device managers is unsupported with slow dispatch.
Evidence used before claiming independence:
- the culprit path is absent from the diff of every affected PR (12/12 checked);
- the failure signature is in the job log (
sub_device_manager_tracker.cpp:118→ 6–9 hits per red job), withpipes>0as the positive control for the grep; - the same
TT_FATALreproduces on an unrelated branch and an unrelated PR run.
Delivered as 13 diagnostic comments across 12 open PRs — no competing PRs opened. Where a job was red for a different reason (8 of the candidates), it was excluded with the measured cause instead of being handed a reusable diagnosis.
- Timeout error mapping +
open_sessiontimeout. The client waits 60 s, the server default was 10 s, andcode="timeout"did not exist in the wire protocol's error buckets — so a server timeout was reported to clients asinvalid_request_error(a client fault). The message is a composite f-string that greps to 0 in source; the defect traces to a single emit site. - Client handshake budget honoured when opening a duplex session (the only override in the tree was
in
warmup.py; no environment variable reached the serving path). - Docs:
run_server.shoverride defaults aligned with the scripts; broken relative links to the serving API guides.
Plus a full root-cause write-up on issue #9072 (CI failure, "duplex open timed out"), with the variable that was silently double-defined and the one-line fix for the error taxonomy.
A package set can be coherent and still be incomplete. Measured in a public CI pipeline: the
environment holds 117 packages, uv pip check reports All installed packages are compatible, and
12.06 s later the same interpreter raises ModuleNotFoundError: No module named 'prisma'. Nothing was
installed in between.
Coherence is a relation between declarations; completeness is a relation between declarations and code — and no metadata-based check can see the second one. Here the package that the code reaches through six import hops is declared in a different extra than the one being installed, so every declared set is satisfied while the import is doomed.
Written up, deposited, discussed upstream, and now with a public corpus behind it:
Coherence is not completeness · 10.5281/zenodo.23955654
· thread: astral-sh/uv#13963
· rules: corpus-medido — 56 measurement rules, Apache-2.0 / CC BY 4.0
- Differential diagnosis under noise — separate "this PR broke it" from "the harness broke it", and prove which one before speaking.
- Fail-closed instrumentation — a gate with an empty universe returns red (
RC 31), never a green. A check that searches for one shape does not certify the absence of the others. - Cost accounting for inference — same-accuracy comparisons priced per problem on real runs.
Python · Bash · GitHub Actions (workflow authority topology, SHA-pinned actions, OIDC sinks) · pytest & property-based testing · AST / YAML validation · LLM serving (vLLM) · reproducibility bundles (CFF, Zenodo, ORCID).
- ORCID — 0009-0001-9116-075X
- Software & corpus — github.com/alexisromtz-lab