Goable scores how suitable the weather is for a specific outdoor activity, at a given place and time, and returns it as a 0 to 100 verdict a booking platform can act on and defend. Built by a two-person team, in the open where it matters.
Goable rests on one technical bet: that the useful interface to weather is not a chart, a dashboard or a widget, but a score from 0 to 100 for a specific activity, place and time, grounded in physics and designed to learn from outcomes.
That score is deterministic and audit-logged: given the same inputs and provider chain it replays exactly, so a booking platform can defend the automated decisions it builds on top of it. The engine synthesises peer-reviewed physics; it does not invent any.
The physics layer is built from peer-reviewed literature and cited in a public reference chain. Machine learning calibrates the curves against observed outcomes; it never replaces the physics. A score you can explain is a score a regulator, a captain or a worried customer can interrogate.
Given the same inputs and provider chain, a score replays exactly. The audit log retains the raw weather samples and the response, so a booking platform can reconstruct any automated decision it made, months or years later.
The activity catalogue is CC BY 4.0 on GitHub, alongside the verification metrics, the drift charter and the research dataset. The engine that turns them into a verdict is what you pay for.
Hono, Drizzle, Postgres, AWS Lambda. A short vendor list and conservative versions in the plumbing, so the complexity budget goes into the engine: the physics, the calibration, the drift surveillance, not into chasing the latest framework.
No SOC 2 or ISO 27001 today. Audited certifications are expensive logos that pay off at enterprise contract sizes Goable does not yet have, so the controls are documented on the security page and evidenced on request. The audit becomes a real budget line when a prospect actually needs the logo, not before.
Goable does not resell forecasts. It turns them into a judgment for a specific activity, place and time.
The output is a 0 to 100 score with its reasoning and hard gates, meant to be read inside a booking flow, not studied as a chart.
Goable is not B2C. The score is embedded via API by the platforms that serve end users. There is no public weather app.
| Horizon | Time scale | Direction | State |
|---|---|---|---|
| Near-term | hours to days | Real-time scoring API for operations | In active build |
| Seasonal | weeks to months | Seasonal analytics and parametric primitives | Exploratory |
| Climate | years to decades | Downscaled viability projections | Exploratory |
Only the near-term scoring API is in active build. The seasonal and climate horizons are exploratory directions the same physics and calibration could support, not a committed roadmap and not a shipping product.
A test-driven scoring engine over peer-reviewed formulas, with hard-gate safety rules and an activity taxonomy across water, snow, air and land.
The activity catalogue is published under CC BY 4.0 on GitHub. The Inverse Suitability preprint is on arXiv with a one-command reproducible package that runs on synthetic data.
Probabilistic ensembles, an evidence-gated self-calibration loop, and an audit log that keeps the raw samples alongside every response.
Sealect, a separate founder-built booking platform, is being wired to the public API as a planned reference integration. It is not yet live, and it is not an independent customer.
Signup, plans and the public buyer surfaces. Goable is pre-launch and pre-revenue; none of this has shipped yet.
Historical, portfolio and parametric endpoints for the seasonal horizon. Exploratory, and gated on real outcome data accumulating first.
We publish the frontier work that would otherwise stay behind the engineering. The credibility gap between "we have a model" and "a regulator, an actuary or a peer-reviewer can inspect it" is where a lot of forecasting infrastructure quietly fails. We close it by defaulting to open.
The current headline artefact is the Inverse Suitability preprint: continuous-item response theory adapted to weather, separating intrinsic condition difficulty δ(x) from rider skill θ_r from behavioural outcomes. The paper demonstrates recovery on synthetic data; real-world validation is a future research step. Preprint on arXiv (2607.01961), with a one-command reproducible package at github.com/goable-io/inverse-suitability and zero proprietary data, so the synthetic-recovery proof stands on its own.
Around it, three open data streams round out the research surface: forecast verification (per-cell BSS, CRPS and reliability at cell-day resolution), the sustainability index (an annual, frozen report of session-weighted non-combustion share across k-anonymised zones), and the difficulty atlas (skill-conditioned scoring per cohort). All CC BY 4.0, all computable from the same physics-first engine behind the API.
Goable is a brand operated by Fabio Carucci, an Italian sole proprietor, not a company. Registered in Magliano Alfieri, 12050 (CN), Italy. VAT P.IVA IT04133450041.
No data protection officer is appointed: Goable sits below the GDPR Article 37 threshold. The data-protection contact is contact@fabio-carucci.com.
Compute in AWS eu-west-1 (Ireland). Postgres (Neon) in eu-central-1 (Frankfurt). Vercel fra1, with CloudFront and Cloudflare in front, TLS 1.2 or higher. LLM features use Anthropic under a bring-your-own-key model.
A two-person team: Fabio Carucci (founder) builds the engine, API and platform; Marta Incarnato leads the Sustainability Index. Pre-revenue and pre-launch. No sales team; procurement questionnaires are answered by the founder.
Goable is infrastructure for the businesses that serve the outdoor economy, not for end consumers. If you build booking software, run an outdoor operation, or work on parametric weather cover and want to compare notes, there is a conversation worth having. The contact page routes you to the right channel.