Project summary

Free, nonpartisian database documenting the AI practices of ~280 companies with 4,800+ cited evidence records, run by a MA 501(c)(3); bill briefings built on the data have been delivered to Massachusetts Senate Offices. AI does discovery through out multi-stage automated research pipeline; humans verify; every record carries field-level provenance.

Raising $10k for four months of verification-pipeline costs (API metering, contract engineering, editorial stipends for the October volunteer cohort).

What are this project's goals? How will you achieve them?

Goal one: a record of corporate AI practice that a legislator or researcher can cite without checking it themselves. Achieved by three things already in production:

1.) automated multi-stage discovery over company disclosures
2.) human verification of every record before it ships, enforced by a corroboration guard in code
3.) field-level provenance that records which step an AI system did and which a person did.

Goal two: a published error rate. A monthly human verification sample starts with the October volunteer cohort; the first quarterly error rate publishes with the methodology page this quarter.

Goal three: citation. Bill briefings built on the record have been delivered to Massachusetts Senate Offices (S.2630, S.301; the north-star metric is a legislative citation, and the government-facing portal (platform.aitrace.org) exists to make that easy. In quarter four we plan to release a versioned dataset with DOIs so researchers can cite a fixed snapshot.

How will this funding be used?

Four months of the verification pipeline's direct costs, nothing else.

• AI API metering for discovery scans and verification passes (measured: about 15 cents per single-company scan, $52-150 per full-corpus run, 6.6 cents per record per verification pass)

• Contract engineering on the verification and provenance tooling

• Editorial stipends for the volunteer verification cohort starting October 2026.

Founder compensation is excluded. The full twelve-month lean budget is $125k, with $79k of it being non-salary.

This funding is scoped to the part that turns drafted records into verified, citable ones, at the $2,500 minimum -> API metering only.

Who is on your team? What's your track record on similar projects?

One full-time founder, Trent Maziarz (UMASS CS graduate), engineering and editorial.

Board expanding to four directors in Novemeber 2026.

Volunteer verfication cohort starting October.

Trent's track record is the thing itself, aitrace.org is live with 280 companies, 1,672 companies, 4,870 cited sources. The pipeline runs in production; the briefings were delivered.

We hold our own tooling to the same standard; the second-generation verifier is going through six pre-registered acceptance evaluations against a human-verdict benchmark, and evaluations 1 and 2 were recorded as failures on recall rather than moving the bar.

That record is published with the methodology, and our practices are equally tracked on the website.

What are the most likely causes and outcomes if this project fails?

Most likely cause? the record goes stale.

A transparency record that is six months behind is worse than none, because it is cited as current.

Staleness happens when the metered scan and verification cadence cannot be paid for, that is exactly what this ask covers.

Second cause, the verifier does not pass its acceptance bar and human verification stays the bottleneck, which limits coverage growth not correctness. failed evaluations are published either way.

Third cause, no legislative citation inside a year, which would mean the briefing format is wrong, not the data. We would change the format.

Outcome of outright failure? The existing record stays public and free as a fixed snapshot.

How much money have you raised in the last 12 months, and from where?

No money raised, except for $5,091.90 cents of Trent's own donations.

We have a donation option accessible on the https://www.aitrace.org/donate site for individual donations.

No grants or institutional money yet since our March 16th 2026 formation.

Pending, Lightcone Commons ($50k ask, decision October), BlueDot Rapid ($3k, submitted August 26th), Awesome Foundation ($1k, two chapters pending),