TrueLumen reads your drawings, specs, studies, and contracts as one body of work — then shows you every place they disagree, with the evidence cited, side by side.
Exhibit B-1.1 — Technical Parameters
3.1 Capacity. The System shall have an aggregate nameplate alternating-current capacity of not less than one hundred eighty (180) MWac measured at the Point of Interconnection.
3.2 Inverters. The System shall comprise sixteen (16) inverter units of the make and model set forth in Exhibit O, each rated not less than 4.2 MVA continuous at 35 °C ambient, arranged in accordance with the Collection System design.
3.3 Substitution. No substitution of the units specified in Section 3.2 shall be made without the prior written consent of Owner.
The drawing shows 14 inverters. The signed contract requires 16. Nobody noticed for two revisions. surfaced day 1 of a 10-day review
We co-develop TrueLumen with a global energy leader and the venture studio behind some of the most rigorous applied-AI companies.
A global energy company building and operating power at utility scale. TrueLumen is co-developed with AES engineering teams reviewing gigawatt-scale renewable projects — the sharpest version of the problem we solve, where a hidden discrepancy becomes costly rework, a schedule slip, or a safety failure once the plant is energized.
Their reviews set our bar for precision, and every capability is validated against real design sets before it ships.
The venture studio founded by Andrew Ng. TrueLumen is built and backed by AI Fund, which co-builds AI companies from the ground up — bringing the technical depth, operating discipline, and go-to-market rigor behind a portfolio of applied-AI businesses.
Their product bar shaped our core: evidence-grounded reasoning that earns trust in rooms where being wrong is expensive.
A gigawatt-scale project lives in hundreds of drawings, datasheets, and specifications that are all contractually obligated to agree. They often don't — and nobody notices. The mistake is cheap to fix on the page and ruinous to fix in the field: a change order, weeks of rework, at worst a safety failure once the plant is energized.
A single missed decimal nearly triggered a near-million-dollar order for equipment that was never needed.
Today the only thing standing between a buried mismatch and the field is the reviewing engineer — reading hundreds of pages one at a time, inside a ten-day window. It leaves no record, can't be verified, and gets harder every year as the volume of review races ahead of a shrinking pool of senior engineers.
Three things, in order — and it never asserts anything it can't trace to a source.
Prose, tables, diagrams, scans — the meaning, not just the text. A value on a drawing lands in context, not as a floating number.
What agrees, what conflicts, what supersedes. Every check is held to the version of record — the signed contract, the governing code cycle — never just the latest file in the folder.
Each flag carries a one-click citation to the source that governs it. Open it, check it, defend it in the room. Nothing is asserted that can't be traced.
Four commitments hold on every review. Here's what each one looks like in practice.
It points out that two numbers disagree. It never overrules your engineer or declares a rule broken.
Low-volume, high-value comments — never a 275-item laundry list. Below the precision floor, you don't see it.
4 surfaced31 suppressedon one 90% set
Read-only on top of the systems you already run. Nothing moves, nothing migrates, nothing to rip out.
Every review states what was checked, what wasn't, and against which sources — so gaps are visible, not assumed.
TrueLumen reads from your systems of record and hands back the artifact your team already exchanges — a cited comment log, on day one of the review window.
| Drawing | Pg | Comment | Reference | Severity |
|---|---|---|---|---|
| E-201 | 4 | Inverter count reads 14; contract specifies 16. | Exh. B-1.1 § 3.2 | High |
| E-114 | 2 | Conductor size below ampacity study result for this run. | Ampacity calc, tbl 3 | High |
| E-308 | 7 | Ground-grid resistance target conflicts with the grounding study. | Grounding study § 5 | Medium |
| E-102 | 1 | Sheet naming departs from the agreed convention. | Exh. B-5 | Low |
Your contracts and design data are the most sensitive documents on the project. They're treated that way.
TrueLumen stores insight and cited evidence — never a copy of your document repository.
Runs in a dedicated cloud tenancy behind two-factor access.
Your data is used to answer your questions — it never trains any model, ours or anyone's.
Every customer's documents and licensed standards stay walled off from every other.
SOC 2 program underway
We're co-built with teams reviewing gigawatt-scale energy projects — the sharpest version of the problem. From there the path is a single axis: the higher the consequence, the higher the value. Each step is a new customer, not a new product.
We'll run a review against your ground truth and show you what we find — every flag cited, so you can check our work before you believe a word of it. For engineering leaders evaluating the product, and for investors, the same door.
Start free info@truelumen.ai