The product does not ask a partner to begin with 2,513 rows. It begins with a ranked top 100.
Edited reconstruction of the shipped interface using entirely fictional investor and fund data.
Living Work Library · Project 02
Investor Intelligence
A sparse list of 2,513 investor records becomes a reviewable decision system — without pretending missing evidence is negative.
Built for a venture fund’s partners to prioritize LPs and prepare outreach. Enrichment runs Claude → Apify → Apollo → Clay, and every field keeps the provenance it arrived with.
Scroll, or open the room
SOURCE LIST · 2,513 RECORDS · 13 COLUMNS
FICTIONAL
Present in source
Not available
Enriched with provenance
73.6% of records were missing at least one enrichment-relevant field, and none of the five decision signals existed as a source column. Field diagram uses fictional records.
Movement 01 · source reality
The spreadsheet held records. The evidence had to be added.
2,513 records13 source columns73.6% missing at least one enrichment-relevant fieldnone of the five decision signals existed as a source column
↔
Comparison · 50% enriched
Illustrative reconstruction based on the verified source schema. All investor data shown is fictional. Unavailable evidence stays unknown.
Movements 01–02
The list becomes comparable.
Most records were incomplete, and the evidence needed to prioritize them did not exist in the source schema.
- 01Source list
- 02
Claude· interpretation - 03
Apify· web & Crunchbase - 04
Apollo· multi-source - 05
Clay· multi-source - 06Comparison plane
Claude · interpretation
Apify · web & Crunchbase
Apollo · multi-source
Clay · multi-source
RESOLVED0
PARTIAL0
STILL UNKNOWN0
Stage 1 of 6 · Source list
Designed reconstruction with fictional records. Each stage identity leads with the company’s own official mark, embedded unchanged from its first-party source and shown on a uniform light field so every mark stays legible on the dark ground. The marks identify the enrichment stages actually used, and imply no partnership, sponsorship or endorsement.
Movement 03 · the product judgment
The most important redesign was not improving the model.
It was teaching the product that “unknown” does not mean “unqualified.”
Confirmed
Solid block · full fill
Evidence found, with a named source and a stage attached. Counts toward the ranking.
Partial
Half fill · outlined top
Something was found, but not enough to stand alone. Counts at reduced weight and says so.
Unknown
Dashed outline · empty
No evidence yet. Preserved as unknown rather than collapsed into a confirmed negative. It receives the configured partial-credit treatment and remains visible.
Negative
Struck block · diagonal rule
Evidence against the signal. It receives the confirmed-negative treatment and must be sourced.
Each state is carried by shape, fill pattern and a written label as well as color, so the four remain distinguishable without color vision, in reduced motion, and in a printed page.
Movement 04 · ranking becomes adjustable
The shortlist answers to the criteria, not the other way round.
2,513 records → a ranked top 100 available immediately → configurable weights → a revised shortlist → profile evidence → a reviewable outreach draft. Move the five weights and watch the order change.
The five signals and their default weights are the product’s own. Every investor, fund and score shown here is fictional.
Reviewable shortlist · 8 of a ranked top 100
Order held
Highest-ranked under the selected criteria — a reviewable shortlist, not objective truth.
Movement 05 · evidence becomes action
From shortlist to a draft the GP can reject.
- 01 Ranked top 100
- 02 Adjustable weights
- 03 Profile evidence
- 04 Outreach draft
If the ranking does not reflect the partner’s priorities, the five evidence weights can be changed and the shortlist recalculates — highest-ranked under the selected criteria, not objective truth.
Edited reconstruction of the shipped interface using entirely fictional investor and fund data.
One row opens into its evidence: the fit score, the five components behind it, the analysis, the contact context that exists — and, plainly, what is still unknown.
Edited reconstruction of the shipped interface using entirely fictional investor and fund data.
A first draft, assembled from the evidence gathered for review. The GP inspects, edits or rejects it — the product does not send on their behalf.
Edited reconstruction of the shipped outreach-drafting interface using entirely fictional investor and fund data. Drafts require GP review.
View 1 of 4 · Ranked top 100
Movement 06 · use and feedback
In use by all three partners.
All three partners
use the product to prioritize LPs and prepare outreach.
GP feedback
changed the scoring and filtering logic, including the emerging-manager correction.
Free and low-cost tiers
carried the enrichment architecture through this stage of the build.
“Teja built an LP intelligence platform for our fund largely from scratch, enriching over 2,500 LP records using AI at a remarkably low initial cost. What stood out was how quickly he translated GP feedback into concrete product changes — the scoring and filtering logic was reworked directly off our conversations. He pairs strong technical execution with genuine product instinct, which is unusual this early in a career.”
Movement 06 · the loop
One reviewable path, end to end.
- 012,513 records
- 02Enriched evidence
- 03Ranked top 100
- 04Adjustable priorities
- 05Profile review
- 06Outreach draft
Each step stays inspectable: the evidence behind a rank can be opened, the priorities behind the order can be changed, and the draft at the end waits for a GP to accept, edit or reject it.
Current, and next.
Current · implemented
- All three partners use the product to prioritize LPs and prepare outreach.
- GP feedback changed the scoring and filtering logic.
- The Claude → Apify → Apollo → Clay enrichment waterfall is implemented.
Next · not built
- Paid-tier scaling of the enrichment architecture.
- Broader enrichment coverage.
- A relationship-continuity layer.
- Outreach-state and progress tracking.
- Micro-commitment follow-through.
All investor names, funds, scores, evidence states, analysis text and outreach copy in this artifact are fictional reconstructions in the product’s visual language. The re-ranking model is illustrative. No raw source data, real LP identity, confidential fund data, real outreach content or private contact detail appears here. The four enrichment-stage marks are the vendors’ own first-party asset files, embedded in this page as inline data URIs and rendered unchanged — not redrawn, recolored, cropped or combined — normalized by matched optical size on a shared visual axis rather than identical pixel dimensions, each on the same neutral light field so a dark mark stays legible on the dark ground. The visible vendor name is kept beside every mark. The names identify the enrichment stages actually used and imply no partnership, sponsorship or endorsement. The company names identify the implemented enrichment stages only, and imply no partnership, sponsorship or endorsement.