Accountability

We don't ask you to trust us. We publish whether we were right.

A power-feasibility call is only worth the track record behind it. So every screening verdict here is timestamped, and scored against what actually happens in the field. This page is the public scoreboard - and the realized base rates that anchor it.

Why this exists. Anyone can publish a confident number. The only thing that earns trust is being measurably, repeatedly right - and owning it publicly when we're not. Grid and on-site power projects take years to energize, so this ledger fills in over time. That's the point: it's falsifiable.

The scoreboard

logged from launch

Screening verdicts timestamped to date (we record every real run)

Outcomes resolved so far - first verdicts resolve as real project timelines close

±40%

Stated screening-grade band we hold ourselves to, per result

Status - honest version: we launched recently, so no predictions have a confirmed real-world outcome yet. As parcels we screened reach (or miss) energization, we will publish each one here as a hit or miss against what we said and when. No cherry-picking - every logged verdict is eligible.

How a verdict gets scored

When a site is screened we record the coordinate, the inputs (MW, timeline), the verdict (ADVANCE / CAUTION / KILL), the fastest path and its month estimate, and the timestamp. A verdict is later marked:

Hit - the realized outcome fell inside our stated band (e.g., we said on-site ~22 mo and it energized within the screening tolerance; or we said KILL and the grid did arrive far outside the window).

Miss - reality fell outside the band. We publish the gap and the reason, and feed it back into the model.

Outcomes are confirmed from primary records - ISO/utility energization filings, operator disclosures, and verified developer reports - each with a source, the same way every number in the tool is sourced.

Realized base rates - what actually happens

Until our own track record matures, the most honest anchor is the realized record for the whole U.S. grid. These are not our estimates; they're measured outcomes for completed projects:

Realized metric (completed projects)	Value	Source
Median time, interconnection request → energized	~4.5 years (projects reaching operation in 2024) - up from under 2 years in 2000-07	LBNL “Queued Up”, 2025 Ed.
Share of queued projects that withdraw before connecting	~80%	LBNL “Queued Up”, 2025 Ed.
Average months in queue - CAISO / ERCOT / PJM	~43 / ~22 / ~24 months	American Clean Power (2023)

These same realized figures now appear inside every verdict, beside our modeled grid date, as a reality check.

How the realized dataset grows (and becomes a moat)

Market- and utility-specific realized timelines don't exist cleanly anywhere - they live in tribal knowledge and scattered filings. We're assembling them two ways:

1. Completed-project mining. Realized request-to-energized durations computed from ISO/RTO queue records and LBNL data, market by market.

2. Verified developer contributions. Operators submit a real deal's timeline (applied date, energized date, utility) with evidence; we verify the source before it counts. Contributors are credited. Unverified claims never enter the dataset.

Every verified outcome and every scored prediction makes this record harder to replicate - that compounding accuracy is the asset. Screen a site →