D. R.
Why this founder surfaced, what the evidence supports today, and the next fact that would most reduce uncertainty.
Why is this person worth attention?
The founder submitted only a company name and deck. The system checks the important claims independently.
The founder submitted only a company name and deck. The system checks the important claims independently.
apply form6 claims verified, 3 contradicted, and 8 still need evidence.
17 claims checked3 expected artifacts are missing. This increases uncertainty; it does not automatically make the founder bad.
1 harmless absences were not penalizedSend the three-condition list and hold the position.
Open claims and receipts →Evidence mapWhat should exist if the founder’s story is true
Expected artifacts
Found artifacts, meaningful gaps, and harmless absences are kept separate.
| Artifact | Status | P(findable) if it existed | Effect on the estimate |
|---|---|---|---|
company domain live Live, transacting, TLS issued 2026-03-22. | found | 0.91n=19 | informs the estimate |
dated changelog Found — and it is the artifact that contradicts the revenue claim. Presence is not the same as corroboration. | found | 0.54n=19 | informs the estimate |
team page Found. Lists three named people against a claim of twelve. We used a team page and a job board here rather than LinkedIn, because we declined LinkedIn in our own not-collected ledger and it would be incoherent to lean on it for the demo's peak moment. | found | 0.66n=19 | informs the estimate |
code host first commit Found, and dated. Confirmation source only — first-commit date and cadence. Never stars, never followers. | found | 0.78n=19 | informs the estimate |
open roles posting Expected at P=0.58 for a company at the stated revenue. None found, on the company domain or on two job boards. Widens the interval by 2.6 and contributes −1.3 to the revenue claim's log-odds. | expected · absent | 0.58n=21 | widens the interval +2.6 — score unchanged |
cap table disclosure Absent but expected at this stage of a diligence conversation. It becomes a gap row in the memo, not a guess. | expected · absent | 0.44n=19 | widens the interval +1.9 — score unchanged |
round terms disclosure Absent. Without a post-money we cannot compute implied ownership against the 8–12% thesis target, and the decision card says exactly that rather than showing a number. | expected · absent | 0.49n=19 | widens the interval +2.2 — score unchanged |
accelerator cohort listing Not expected, and not collected either — accelerator rosters are a declined pedigree channel. Grey for two independent reasons. | not expected for this founder profile | 0.06n=19 | not penalised — costs nothing |
Missing evidence widens the interval and never lowers the score. Hatched rows are absences — the graphite ones our findability prior already predicted for this resource class, so they cost nothing. That asymmetry is what keeps this from re-encoding the network gate: a founder is never marked down for lacking artifacts that someone in their position would not be expected to have.
Detailed founder profileCredibility, build capability and persistent history
Founder evidence profile
Credibility and build capability stay separate.
Two components, both persistent, both append-only. The split is why a lie about revenue does not erase a real build record, and why the decision can be conditional rather than a flat pass.
Credibility 44.1 against build capability 64.8 is the whole shape of this deal. He builds; his statements about what he built do not survive a fetch. Averaging those two numbers would destroy exactly the information the investor needs, which is the same reason we never average the three axes.
AXES DISAGREE — Founder neutral, Market bear, Idea-vs-Market neutral. We don't average them and we don't resolve it.
Decision axesFounder, market and idea-versus-market stay separate
Founder
▬stable[-4.62, 2.18]·4pts25 direct observations at this asof, 6 claims verified and 3 contradicted. Prior weight 19% — that share of this number is the reference class, not the person. The persistent Founder Score is one of four inputs here, not a substitute for the axis. 2 absent-but-expected item(s) widened the interval without lowering the point.
Market
▬stable[-0.41, 0.35]·4ptsCategorical on purpose. There is no number here to average with the other two axes.
6 market row(s) at this asof, stance index -0.772 from the terms listed, plus a sector prior of 0 carried over from 0 previous Market verdict(s) in b2b_fintech_infra. Categorical on purpose: there is no number here, so there is nothing that could be combined with the other two axes even by accident.
Idea vs Market
▬stable[-0.96, 1.56]·4pts23 scored input(s) at this asof against a BEAR market read. Prior weight 18%. The pivot clause reads build capability out of Memory — the one declared cross-axis link in the system.
How the estimate was formedCold-start prior, history and score definitions
Three different things get called a founder score. They are not the same and we label which is which.
This is why a contradiction on Ledgerline moves a number that has a 2024 history attached to it. The score follows him. Not this company — him.
Cold-start bench
How much is direct evidence, how much is a reference-class prior, and what would narrow the range.
Founder score history
Append-only versions preserve the person’s record across ventures.
- 2024-03-1152[31, 73]· Kestrel Ops— First scored. One claim resolved, so the reference class is carrying almost all of it.
- 2024-06-0263[46, 80]· Kestrel Ops— Four of four claims verified against third-party observables.
- 2024-11-1874.5[61, 88]· Kestrel Ops— Eight of nine claims verified. This is the record he arrives with.
- 2025-02-0474.5[61, 88]· Kestrel Ops— Venture wound down. No claim resolved, so the score does not move. Winding a company down is not a credibility event and we refuse to treat it as one.
- 2026-07-1868.2[52.1, 84.3]· Ledgerline— New venture. The score carries over — it belongs to him, not to Kestrel Ops. It drifts down only because seventeen months passed without a resolved claim, which widens rather than punishes.
- 2026-07-1851.6[35.4, 67.8]· Ledgerline— Revenue claim contradicted at log-odds −4.7.
- 2026-07-1846.9[31, 62.8]· Ledgerline— Headcount claim contradicted at −3.4.
- 2026-07-1844.1[29.8, 58.4]· Ledgerline— Founding-date claim contradicted at −2.8. Three contradictions in thirty-eight minutes, against a record of eight verified claims from 2024 that is still sitting in the same line.
Founder–market fit
Domain-exposure evidence with an interval, not a personality guess.
Named in the brief as a soft-skill assessment that should carry a prediction interval.
Evidence basis: Lived-domain-exposure atoms from written technical answers across both ventures — named settlement counterparties, specific reconciliation failure modes, dated integration details. Scored on atom density, never on how fluent the prose is.
Entity resolution
Why separate venture records are believed to belong to the same human.
- same code-host handle across both ventures
- same registrant email domain on both company domains
- first observed 2024-03-11
name — the spelling differs across the two ventures, which is exactly why we never key the merge on a name
Milestones
Timestamped observations from the same ledger.
- 2024-03-11Kestrel Ops incorporatedincorporated
- 2024-08-19Kestrel Ops first revenuefirst revenue
- 2025-02-04Kestrel Ops wound downwound down
- 2026-03-22Ledgerline incorporatedincorporated
- 2026-04-30Ledgerline first release tagshipped
Complete recordEvery remaining field, for auditability
- Person id
- per_dr
- Display name
- D. R.
- Pseudonymized
- false
- Is real person
- false
- Refuter enabled
- true
- Provenance class
- synthetic
- Provenance note
- AUTHORED. This person and both of his ventures were written by us, precisely so that we could plant contradictions and catch them on camera without doing it to a real person. The contradiction fixtures are hosted at real public URLs and the Validator performs a genuine HTTP GET against them.
- Contact status
- public_email
- Channels
- apply_form, github_person, domain_probe
- First observed at
- 2024-03-11T00:00:00Z
- Sector
- b2b_fintech_infra
- Region
- DACH
- Founder type
- technical
- Opportunities
- opp_ledgerline
- Axes disagree
- true
- Axes disagree headline
- AXES DISAGREE — Founder neutral, Market bear, Idea-vs-Market neutral. We don't average them and we don't resolve it.
- Axes disagree detail
- The disagreement that matters is inside the Founder read: credibility is bear at a lower bound of 29.8 while build capability clears the gate at 55.4. Two of three axes point down and the third is neutral, and there is no arithmetic in this system that would turn that into one number. Naming the binding and dissenting axes is the whole output; resolving them would be inventing a view we do not have.
- Cold start bench blocked reason
- Not a cold-start founder. Twenty-three direct observations across two ventures, so the prior weight is 10/(23+10) = 30% and the reference class is doing very little of the work. The bench is the wrong instrument here and rendering it would imply a thinness that does not exist.