LIVE CAPTURE · SESSION ······ · 00:00:00
0 events logged
felineunion.org · investigative · file 03-30 · EN · FR

LifeLog
was real.

Before there was a like button, before there was a feed, before there were smart speakers in your kitchen and accelerometers in your pocket — the Pentagon spent eighteen months building a system to index a human life. Then, on a single morning in February 2004, they killed it. And something else launched the same day.

§ 01 · The Solicitation

"An ontology-based system that captures, stores, and makes accessible the flow of one person's experience."

In May 2003, the Defense Advanced Research Projects Agency posted a bid solicitation titled BAA 03-30. The program manager was a man named Douglas Gage. The premise, on paper, was modest: build a research system that could trace the "threads" of an individual's life — events, states, relationships — and weave them into something searchable.

In practice, the threads were everything. The phone calls a subject dialed. The email they sent and received. The pages they browsed. The TV they watched. The radio they heard. The credit cards they swiped. The places they went, captured by GPS sensors they would be required to wear. Their heart rate. Their step count. Every magazine they opened. Every book they finished. Every breath, according to one early DARPA description, every step made, every place gone.

The grant guidelines specified that the researchers awarded LifeLog contracts would be required to test the system on themselves. Cameras would record them during trips. Biomedical sensors would monitor their bodies. Everything they sent and read would be indexed and made searchable. The output would not be a diary. It would be a model. The system was meant to infer routines, habits, and relationships, and then exploit those patterns to anticipate what the subject would do next.

"To be able to trace the threads of an individual's life in terms of events, states, and relationships." — DARPA BAA 03-30, Program Objective (May 2003)

The Threads · what LifeLog wanted

/01 · COMMUNICATIONS
Every message sent and read
email · IM · SMS · phone calls · faxes · postal mail (scanned) · video conferences
/02 · LOCATION
Every place visited
wearable GPS traces · check-ins · routes · movement patterns · dwell times
/03 · TRANSACTIONS
Every purchase made
credit card payments · receipts · subscriptions · financial records · vendor patterns
/04 · MEDIA
Every channel consumed
books · magazines · newspapers · TV · radio · websites · database queries
/05 · BODY
Every biometric signal
heart rate · steps · sleep · breathing · ambient sensors · physiological state
/06 · RELATIONSHIPS
Every person known
contact graph · meeting cadence · co-location · inferred closeness · social network
/07 · CONTEXT
The ambient stream
broadcast feeds in proximity · documents passed · environmental factors
/08 · INFERENCE
The model on top
preferences · plans · goals · intentions · routine prediction · behavioural forecast
§ 02 · The Demonstration

This is what it looks like when the system works.

LifeLog never reached deployment. But the categories of data it wanted — and the premise that a human being is most legible when their threads are unified into one indexed stream — are not exotic anymore. They are the default behaviour of every modern web page, including this one. The panel below was assembled from your session in the last few seconds. No data leaves your browser. Nothing is uploaded. The point is only to make the surface area visible.

▸ subject capture · session active

One reader. Forty-seven data points.

refresh rate · ~1Hz · scope · local only
Session ID
Local Time
Timezone
Language
Dwell Time
0s
Scroll Depth
0%
Cursor Distance
0 px
Click Count
0
Keystrokes
0
Viewport
Screen
Pixel Ratio
Input Type
User Agent
Platform
Network
Referrer
direct
Visibility
visible
Focus Changes
0
CPU Cores
Device Memory
Color Scheme
Battery
Storage Available
Every one of these signals is already collected, in some form, by an average commercial web page. The 2003 program wanted to add what was missing — body, location, every device, every transaction — and fuse them under one ontology. Twenty-two years later, most of the rest has been added too, just not by the agency that asked first. This panel runs entirely in your browser · no transmission
§ 03 · How It Died

A timeline, compressed.

Click an event to expand. The conventional version of this story flattens it into one coincidence on one day. The longer version is messier, and more interesting, because LifeLog did not die — it was retired from public view.

2001 — 2002 · Pre-history
DARPA begins exploring "total information awareness" and life-logging.
After September 2001, DARPA's Information Awareness Office launches a portfolio of pattern-recognition and behavioural-modeling programs, the most famous of which is Total Information Awareness (TIA). LifeLog is conceived inside the Information Processing Techniques Office, separately from TIA, but in the same intellectual climate: how do you make a person legible to an algorithm?
May 13, 2003
DARPA posts Broad Agency Announcement BAA 03-30.
The program is formally launched. Program manager Douglas Gage solicits proposals for "an ontology-based subsystem that captures, stores, and makes accessible the flow of one person's experience." Proposal deadline: May 7, 2004. Four 18-month contracts to be awarded. Researchers required to instrument themselves.
July 2003
Wired publishes. The public finds out.
Noah Shachtman's reporting in Wired makes the scope public — cameras, GPS, biometric sensors, indexed phone calls, scanned mail. The framing is unflattering. Civil libertarians read the BAA closely. Privacy advocates fire alarms.
December 2003
Congress defunds Total Information Awareness.
TIA is cancelled by congressional vote. LifeLog is administratively distinct, but the political shadow falls across both. Inside DARPA, director Tony Tether begins reading the room.
January 22, 2004
BAA 03-30 is formally withdrawn.
DARPA quietly cancels the Broad Agency Announcement. No proposals will receive awards. Douglas Gage is mid-evaluation when Tether pulls the plug. The agency provides one explanation, on the record: "a change in priorities."
February 4, 2004 · The Coincidence
LifeLog is publicly buried. TheFacebook launches the same day.
Defensetech.org publishes the obituary. DARPA director Tony Tether confirms the program is dead. Spokesperson Jan Walker offers no further comment. On the same calendar day, a Harvard sophomore launches a social network at thefacebook.com. The conspiracy theory writes itself; the truth is harder. Most of LifeLog's program manager's appointment was not renewed. Most of LifeLog's data taxonomy was about to be voluntarily filled out, by hundreds of millions of users, on a platform DARPA had nothing to do with.
September 2004 · Seven months later
ASSIST appears. Same architecture, military application.
DARPA's Advanced Soldier Sensor Information System and Technology program emerges publicly. Soldiers wear cameras, microphones, GPS, and biometric sensors on patrol. Sensor streams are indexed by AI software into "an electronic chronicle of events" for after-action review. Defensetech.org calls it "LifeLog revived." NIST will later publish performance evaluations. The architecture is intact; the subjects have simply been narrowed to enlisted personnel.
2007 — present
The threads get filled in. Voluntarily.
The iPhone ships GPS, accelerometer, and microphone in one pocketable enclosure. Fitness wearables add the biometric layer. Smart speakers add ambient audio. Social platforms add the relationship graph and message log. Credit card companies add the transaction layer. By the late 2010s, every category in DARPA's 2003 BAA exists in commercial form, owned by no single agency, accessible through warrants, subpoenas, and purchase.
2024 — 2026
AI assistants close the loop.
Large language models give the indexed life database the "enduring personalized cognitive assistant" DARPA imagined in the BAA. Memory features, smart-glasses screen recording, agentic browsing, persistent context windows — these are the inference layer that LifeLog's BAA called "exploit these patterns to ease the user's task." The system, distributed across a dozen companies, is functionally complete. The original program manager retired and took up sailing.
§ 04 · February 4, 2004

One day. Two announcements.

The conspiracy version of this story says Facebook is LifeLog. It isn't. Mark Zuckerberg was not a DARPA asset. The Bell-to-Starner consulting line — the one piece of evidence anyone can name — is a single conversation about wearable memory, not a handover. Douglas Gage, LifeLog's program manager, has explicitly denied any link. But the more accurate story is, in its way, stranger. On the same calendar day in 2004, the government conceded that it could not legally build a system to index a population's lives — and a private company shipped the social architecture that, over twenty years, would do most of it anyway.

11:00 ET · the pentagon
DIES.
DARPA director Tony Tether terminates the LifeLog program. Reason given on the record: "a change in priorities."
BAA 03-30 withdrawn · no proposals advanced · Douglas Gage mid-evaluation when notified · ASSIST will appear seven months later under a different office
later that day · cambridge, ma
LAUNCHES.
A Harvard sophomore publishes thefacebook.com as a social directory for students. No connection. No conspiracy. Just calendar.
initial scope: harvard only · expands ivy league within months · adds messaging, photos, news feed across 2005–2006 · platform now indexes 3+ billion accounts
§ 05 · The Inheritance

Everything LifeLog wanted is collected today.

The original BAA itemised eight categories of personal data. None of them are exotic anymore. The list below maps the 2003 specification, column by column, to its twenty-twenty-six commercial equivalent — not as accusation, but as accounting.

DARPA BAA 03-30 · /01
Every email, IM, SMS, phone call, fax, and piece of postal mail.
↓ filled by
Gmail, Outlook, iMessage, WhatsApp, Signal, the carrier-level metadata your phone provider sells, plus the postal scanning service the USPS already runs.
DARPA BAA 03-30 · /02
Continuous GPS traces and movement patterns from a worn sensor.
↓ filled by
The phone in your pocket, plus Google Maps history, Apple Significant Locations, every fitness watch, every car telematics module.
DARPA BAA 03-30 · /03
All credit card payments, subscriptions, and financial records.
↓ filled by
Visa, Mastercard, Plaid, Amazon, Apple Pay, plus the data brokerage layer that resells transaction streams to advertisers.
DARPA BAA 03-30 · /04
Every book read, magazine opened, TV watched, radio heard, site browsed.
↓ filled by
Kindle, Spotify, Netflix, YouTube, plus the cross-site ad-tech graph that watches you read this page right now.
DARPA BAA 03-30 · /05
Continuous biometric capture: heart rate, breathing, steps, sleep.
↓ filled by
Apple Watch, Fitbit, Oura, Whoop, Garmin, plus the smart mattress and the smart ring and the smart scale.
DARPA BAA 03-30 · /06
A complete social graph: who you talk to, how often, how closely.
↓ filled by
Meta, LinkedIn, X, TikTok, with weighted edges inferred from message frequency, photo co-presence, and shared location.
DARPA BAA 03-30 · /07
Ambient broadcast and document context in proximity to the subject.
↓ filled by
Alexa, Siri, Google Assistant, plus the smart-TV ACR module that fingerprints what's playing in the room.
DARPA BAA 03-30 · /08
The inference layer: preferences, plans, goals, predicted behaviour.
↓ filled by
GPT-class models with persistent memory, AI-assistant features in every major OS, agentic browsers, and the ad-prediction stack that already runs at scale.
"Sure the program was canceled, but the idea still lives on." — Newton Lee, former DARPA researcher, quoted in WHYY (2024)
THE ACTIVATION THRESHOLD

Collection becomes discipline when the subject knows the record can act.

A stored life is not yet a completed panopticon. The disciplinary turn arrives when a person understands that conduct can be recorded, joined to identity, interpreted, scored, and returned as a recommendation, price, ranking, eligibility decision, or intervention. At that point the system no longer has to issue every command. Anticipation does part of the work.

01

Wall → access

The boundary is now the account, checkpoint, service, or payment rail that can admit, delay, price, or refuse.

02

Tower → data

No single watcher is required. Sensors, platforms, brokers, and institutions can each contribute partial visibility.

03

File → identity

Events become consequential when attached to a durable person, device, household, vehicle, or relationship graph.

04

Sentence → score

Classification can shape the next encounter before a person knows which record or threshold produced it.

05

Guard → protocol

Rules embedded in software can repeat decisions at scale while making institutional judgment look automatic.

06

Discipline → self-performance

When review is expected, people may withdraw, continually check, or adjust themselves before an official intervenes.

Boundary of the claim. Awareness is not universal, and it does not produce one response. It can also produce resistance, evasion, organizing, and demands for appeal. The test is empirical: who knows what, which systems actually decide, where records cross institutions, and what remedy can interrupt the loop?
§ 06 · What's Worth Remembering

The argument has never been about capability.

In 2003, the United States Department of Defense looked at a system that would index a person's emails, calls, locations, purchases, conversations, body, and relationships — and the political class said no. Not on grounds of feasibility. On grounds of principle. The same political class, twenty-two years later, lives inside that system every day, because it was built by companies instead of agencies, and the constitutional arguments do not apply.

FINDING / 01
The legal architecture has not caught up.

Fourth Amendment protections — and their Canadian equivalents under Section 8 of the Charter — were written for a state that wanted to enter your house. They were not written for a private entity that buys a feed of your movements, your relationships, and your physiology from a dozen vendors and resells it.

FINDING / 02
"A change in priorities" was honest.

Tony Tether's one-line cancellation has been mocked for two decades. But it was true in a way the joke misses. The priority that changed was the locus of capture — from a government program subject to oversight, to a commercial layer subject to terms of service. The capability did not change. Only the venue.

FINDING / 03
The successor was already on the shelf.

ASSIST appeared seven months after LifeLog died, with the same data-fusion architecture, narrowed to soldiers. The technology was not abandoned. It was reassigned. The civilian version was outsourced to the consumer market, where it has been built more thoroughly than DARPA could have funded.

FINDING / 04
Consent is not the same as participation.

The 2003 BAA required informed-consent protocols for human subjects research. The 2026 equivalent — Terms of Service running to tens of thousands of words, unilaterally amended, opt-out impossible without exiting modern life — is a paperwork artefact that resembles consent. It is not the same thing.

FUTURES FILE / EPSTEIN · AI · HUMAN DESIGN
The future they were buying.

LifeLog asked what becomes possible when a system can record a whole person and infer what they will do next. Jeffrey Epstein's network points to the next question: what becomes possible when that inference layer is joined to genetics, reproduction, longevity research, and the ambition to design the people who come next? The same network contains foundational AI research, institutional money, OpenAI capital and governance, transhumanism, and an explicit project to reproduce Epstein's own DNA.

01 / Minsky · 2002

MIT confirms that Epstein gave $100,000 in 2002 to support Marvin Minsky's research. Minsky was not merely adjacent to AI: he co-founded MIT's Artificial Intelligence Laboratory and was one of the field's foundational figures.

02 / Ito and the Media Lab

MIT's independent investigation documents later gifts to the Media Lab after Epstein's conviction, his designation as a disqualified donor, efforts to keep his name out of the record, and unrestricted support directed to Media Lab director Joi Ito. Epstein was not watching the technological future from outside. His money bought continued access inside one of the institutions building it.

03 / The road into OpenAI

Ito operated inside the venture network that turned experimental computing into funded companies. The direct founding record for OpenAI names Peter Thiel among its supporters in 2015. Epstein later maintained contact with Thiel. The route is not a straight corporate org chart; it is the social and financial circuit through which money, people, and technological ideas moved.

04 / Summers and governance

Harvard's records and subsequent reporting document Epstein's long relationship with Lawrence Summers. OpenAI then appointed Summers to its board in November 2023. A figure from Epstein's institutional circle arrived inside the governance of the company most associated with the generative-AI wave.

05 / Thiel, capital, and transhumanism

Released schedules and correspondence document meetings and contact between Epstein and Thiel. Thiel funded OpenAI and has spent decades financing life extension and arguing that technology should break the limits imposed by nature and death. Epstein's access therefore reached directly into both sides of the future under construction: machine intelligence and engineered human longevity.

06 / Reproduction as a project

Epstein wanted to use his New Mexico ranch to impregnate multiple women and “seed the human race” with his DNA. He discussed genetics, artificial intelligence, cryonics, and “superior humans” with scientists. This was not abstract curiosity. It was a reproductive program with himself as the template.

The convergence

AI supplies eugenics with what it never had: population-scale memory, behavioural prediction, automated ranking, genetic analysis, and a feedback system capable of deciding which traits are desirable. LifeLog is the data body. Modern AI is the inference engine. Reproductive technology makes the inference actionable across generations. That is why this network matters—not because a dead financier secretly invented AI, but because his money and relationships reveal which futures powerful people believed were worth funding before the public had language for them.

§ 07 · Hand It Forward

Ten ways to say it out loud.

§ 08 · Origin System

LifeLog was not born in a vacuum.

Before LifeLog there was an institution built to turn speculative science into strategic capability. DARPA was created after Sputnik to prevent—and create—technological surprise. Its program managers fund universities and companies to prove that a technology can work; military agencies, intelligence organizations, and commercial partners can then carry it into operation.

The new supplemental file places LifeLog inside that machinery: DARPA's 1958 origin, its operating model and budget, the sixteen publicly reconstructed information-awareness research efforts surrounding TIA, and the early-1990s software ideas that made a human life imaginable as an operational application.

$4.322BDARPA · FY2026 enacted agency budget
16Related information-awareness R&D efforts identified by CRS
$317MThose efforts · FY2001–FY2003 · not LifeLog alone

LifeLog itself was canceled at the solicitation stage. The portfolio totals are not its price tag. They show something more important: LifeLog was one horse in a funded field of identity, link-analysis, database, translation, decision-support, and sensor-fusion programs. Killing one name did not end the race.

Enter 1992 · open the interactive evidence file →