BRENNEN

Industries / Finance and business services / Finance and banking

Private equity and venture capital

Private equity and venture capital firms raise money from pension funds, endowments and other investors into funds that buy stakes in private companies, from new startups to large established businesses, and sell them some years later.

1946

American Research and Development (ARD) began investing money raised from outside investors in new companies, with the Harvard Business School professor Georges Doriot as its president from that year; the school's library calls it "the first publicly funded venture capital firm".

library.hbs.edunews.harvard.edu

47

Milestones achieved across 7 domains.

13

To watch for: announced, not yet confirmed outside.

14

Open deliverables. Nobody has shipped them.

In one minute

Where it started

In 1946 American Research and Development began investing money raised from the public in new companies, and the $70,000 it put into Digital Equipment Corporation in 1957 was worth $355 million 14 years later. From 1979, when US pension funds were allowed to invest in it, most venture money came through limited partnerships, funds that investors commit to for about ten years. Buyouts, purchases of whole companies paid for mostly with borrowed money, grew from deals Jerome Kohlberg arranged at the bank Bear Stearns from 1965.

Where it stands

In 2025 venture investors put $425 billion into more than 24,000 private companies, about half of it into AI companies, by Crunchbase's count. Buyout funds did $904 billion of deals and sold $717 billion of companies, by Bain's count, yet still held about 32,000 unsold companies worth $3.8 trillion, and the cash they paid back to their investors stayed below 15% of the value of their holdings for a fourth year running.

What is open

Many records this industry runs on are public but scattered: pension funds' commitments and California's fee reports, accelerator terms, research grants that led to private funding, the share terms in startup charters, continuation-fund deals, and what buyouts and SPAC mergers promised against what they delivered. Agents that collect and read documents can turn each into an open table. The other open jobs are public tests for AI: reading fund agreements, working out who is owed what from a fund or a sale, checking founders' claims against public registries, spotting invented customer lists, and interviewing managers. None of them requires anyone to invest or move money.

Domains

AchievedWatch forOpen

Advances to build with

3 research notes

Open in private equity and venture capital

3 of 14 open deliverables

Open · Raising funds

Public commitments in one table

An open, sourced table of the commitments US public pension plans make to private funds, taken from the documents the plans publish: board and investment committee materials, fund lists and annual reports. Each row gives the plan, the fund, its manager, the amount committed, the date, and the document and page it came from. It records public facts only and does not rate funds or managers.

Built with what exists

A web agent such as Browser Use 1.0 to collect board documents (see the Note "Deleting 80% to start over"); pdfplumber and camelot for PDF tables, as in the CalPERS project; a vision-language model such as Claude Fable 5.2 for scanned pages (see the Note "Three working devices"); Python.

Credit · unclaimed

Open · Finding deals

A register of accelerator terms

A public, dated table of startup accelerators worldwide that records, for each program, what it gives and what it takes: cash invested, the equity or instrument taken (such as a SAFE, a simple agreement for future equity that turns into shares later), program length, cohort size, city, sector focus and demo-day date. Each field links to the program page it came from, and the table is re-read every quarter so that changes in terms show over time. It records published terms only and does not rate programs.

Built with what exists

A web agent such as Browser Use 1.0 to collect program pages (see the Note "Deleting 80% to start over"); Claude Fable 5.2 to pull terms into fixed fields (see the Note "Three working devices"); Python; a public Git repository, so each quarter's changes are a dated record.

Credit · unclaimed

Open · Due diligence

Fake users, caught early

A public test of the checks that a buyer, or a neutral vendor, can run on a target's customer list to tell real users from fabricated ones without the buyer seeing personal data. Build a set of invented customer tables: some generated as a realistic population with sign-up times, activity and email domains, and some fabricated in the ways seen in fraud cases, such as a list made by a script or a bought list relabeled as customers. Score each checker on how many fakes it catches at a fixed rate of false alarms. All data is invented.

Built with what exists

SDV (Synthetic Data Vault), an open-source Python library that generates realistic tables; Faker for invented names and addresses; scikit-learn; the detectors of Kindji and co-authors as baselines; Claude Fable 5.2 to write fabrication scripts in varied styles (see the Note "Three working devices").

Credit · unclaimed

Where it sits