Look SidewaysA guided argument against comparison

Coverage ledger · candidates through publication

No record ceiling.
Every gap stays visible.

The project now tracks discovered people before they become profiles. Publication still requires a dated age-relative outcome and independently useful evidence; uncaught private or poorly indexed outcomes remain a known boundary rather than a reason to declare the census complete.

3,578published paths
92.0%three-event trajectories
87.6%two or more listed sources
71.7%pass the search evidence gate

The complete funnel

A profile count is not a coverage claim.

The target is uncapped public-evidence coverage through age 30. The top-0.001% calibration remains methodology-review pending until its exact-age field gates are independently audited.

01 · Discovered universe10,339

Unique candidate identities

Deduplicated across the published corpus and four broad research queues.

02 · Research backlog4,076

Awaiting publication decisions

Queued, researched, or eligible candidates that are not yet published.

03 · Published evidence3,578

Evidence-gated paths

No qualifying person is downsampled to preserve a target record count.

04 · Recall backtest5/5

Known outliers recovered

A five-person sentinel check, including Neil Movva—not a population recall estimate.

Candidate
A discovered identity with source provenance, before an eligibility decision.
Published
A dated age-relative milestone supported by independently useful evidence.
Indexable
A published profile that also clears the stricter search-evidence gate.
Audited
A profile whose listed source links were checked in the latest reachability pass.

Source freshness

Published Look Sideways corpus3,578 · 0d old
Y Combinator company directory4,105 · 7d old
Pantheon and Wikipedia candidate queue3,000 · 35d old
Founder and engineer candidate queue1,071 · 35d old
Creator candidate queue476 · 27d old
Technology candidate queue290 · 35d old

Known lifecycle states

queued3,115
eligible961
ineligible2,685
published3,578

Four different claims

Coverage is a ladder—not one percentage.

A filled field is not a verified fact. A listed source is not an independent audit. The product keeps those levels separate.

01 · Record structure76.6%

Narrative fields present

Milestone, early history, family context, evidence summary, and current position.

02 · Source diversity87.1%

Two source domains

3,115 profiles link evidence across at least two domains.

03 · Publication gate71.7%

Indexable evidence

Two sources, non-low leverage confidence, a milestone, and a trajectory.

04 · External verification41.9%

Source reachability verified

Source URLs have been fetched and verified reachable on 2,740 audited paths. 838 newer paths are not yet audited. Content verification and user comprehension remain pending.

Annotation confidence

The weaker layer is documented starting advantage.

Biographies reliably celebrate output. They are much less consistent about money, introductions, family logistics, private coaching, permission, and social capital.

Starting-advantage evidence

What access or conditions were documented near the beginning?

High798 · 22.3%
Medium1,628 · 45.5%
Low1,152 · 32.2%

Built-leverage evidence

What multiplying capacity was documented later in the path?

High1,129 · 31.6%
Medium1,538 · 43.0%
Low911 · 25.5%

Published composition

The published corpus sits inside a larger queue.

Every published person cleared an age-relative evidence gate. These shares describe the published corpus; candidate and source gaps remain visible above.

Cohort share

Founders / operators1,390 · 38.9%
Athletes1,079 · 30.2%
Creators / artists811 · 22.7%
Researchers / independent engineers298 · 8.3%

Observed career tier

T1653 · 18.3%
T21,254 · 35.0%
T31,498 · 41.9%
T4173 · 4.8%

Tiers summarize documented career recognition. They do not measure human worth, potential, or only the selected early milestone.

The denominator that remains unobservable

This project cannot count every promising path that disappeared.

There is no matched group of ordinary attempts, near-misses, or people with similar backgrounds who never reached a documented milestone. Therefore this dataset cannot estimate success probabilities, causal effects, or “how many failed for every Bill Gates.”

Observed10,339 candidates; 3,578 published
Not observedComparable attempts and near-misses
Valid conclusionDescriptive patterns and path differences
Invalid conclusionPersonal odds or a causal success formula

What would earn an audited v1.0

One gate done. Three remain.

  1. 01
    Double-code a stratified sample.

    Completed 2026-07-31 across every populated cohort and career tier. Tier agreement was 0.723; leverage scores matched exactly 33.8% of the time and landed within one point 91.9% of the time.

  2. 02
    Audit source reachability.

    Last pass 2026-08-08, covering 9,934 source URLs across 2,740 of 3,578 paths. 1,498 paths have every source reachable. 838 paths added since that pass are still unaudited, so this gate is not closed.

  3. 03
    Observe first-time readers.

    Test comprehension and the intended emotional arc without coaching. Requires external participants.

  4. 04
    Add a matched comparison study.

    Only then could population odds or counterfactual claims be investigated.

Inspect the published pathsRead the full methodology →

The releaseLook sideways for information. Never turn another path into a verdict on your own.