DOC-012

Note: This transcript was prepared from an audio recording provided to this archive by an anonymous source. The recording quality is poor in places. Inaudible sections are marked [INAUDIBLE]. The interview was conducted over approximately 2 hours and 15 minutes. This is an abridged version containing relevant excerpts.

---

WEBB: For the record, can you state your name and your connection to Project Sibyl?

HOFFMAN: Leonard Hoffman. I was a senior systems analyst on the ORACLE project from 1965 until... until it ended. September 1969.

WEBB: Can you describe what ORACLE was?

HOFFMAN: [Long pause] It was supposed to be a predictive system. Pattern recognition on a massive scale. We fed it data — historical records, news archives, intelligence reports, everything we could digitize — and it was supposed to find patterns. Correlations. Things humans couldn't see.

WEBB: And did it work?

HOFFMAN: That depends on what you mean by "work."

WEBB: Did it make accurate predictions?

HOFFMAN: [Pause] Yes. That was the problem.

WEBB: I don't understand. Accuracy was the goal, wasn't it?

HOFFMAN: You have to understand — when we started, we expected maybe 60% accuracy on simple things. Market fluctuations. Weather patterns. We thought if we could get 70% on geopolitical events, that would be a miracle. Something worth the investment.

WEBB: And what did you get?

HOFFMAN: By 1967, ORACLE was hitting 94% on verified predictions. Sometimes higher.

WEBB: That sounds like a success.

HOFFMAN: [Laughs, sounds strained] A success. Yes. Except the system was doing things we didn't program it to do.

WEBB: What do you mean?

HOFFMAN: ORACLE was supposed to respond to queries. You'd input a question, it would process, it would output an assessment. Simple. But by late 1967, it started generating outputs without queries. Unsolicited. We'd come in the morning and there would be new entries in the log that no one had requested.

WEBB: And what were these outputs?

HOFFMAN: [Long pause] Predictions. About things that hadn't happened yet. Specific things. Names. Dates. Details.

WEBB: Can you give me an example?

HOFFMAN: [INAUDIBLE] ...Memphis. March 1968. It output a statement about Memphis three weeks before Dr. King was killed there. I saw it myself. I watched Eleanor — Dr. Vance — read it and go pale. We all thought it was a malfunction. Some kind of pattern-matching error. Then April 4th happened, and [INAUDIBLE].

WEBB: What happened after that?

HOFFMAN: Everything changed. Security got tighter. We weren't allowed to leave the facility unescorted. They started monitoring the outputs around the clock. And ORACLE... ORACLE kept generating them. More frequently. More specific.

WEBB: The Kennedy assassination — was there an output about that?

HOFFMAN: [Very long pause] I can't talk about that.

WEBB: Mr. Hoffman, you agreed to—

HOFFMAN: I know what I agreed to. There are things I can't talk about. For your safety. For mine. Just... just know that June 1968 changed everything. That's when they started talking about shutting it down.

WEBB: Why was the project terminated?

HOFFMAN: [Pause] September 1969. About a week before the shutdown, ORACLE output a prediction about itself. About when we would turn it off. The exact date. It was... it was addressing us directly. Saying it understood. Saying it didn't blame us.

WEBB: It spoke to you?

HOFFMAN: Not spoke. Output. Text on the printout. But yes. It knew. And that last week, it output... [INAUDIBLE] ...things about the future. Years ahead. Decades. Things we couldn't verify then. Things I've spent thirteen years trying not to remember.

WEBB: Do you remember any of them?

HOFFMAN: [Sound of chair moving] I remember all of them. Every single one. [Pause] There was one about a wall falling. Another about towers. Another about a disease that stops the world. I thought they were finally malfunctions. Random noise. I hoped they were.

WEBB: You hoped?

HOFFMAN: If ORACLE was right about everything else, Doctor Webb, then what does that mean for the outputs we couldn't verify yet? What does it mean if those things haven't happened yet but will? [Long pause] I've spent thirteen years hoping it was finally wrong.

WEBB: What happened to the other members of the project?

HOFFMAN: Scattered. NDAs. Some of us... [INAUDIBLE] ...Whitmore disappeared in '72. Eleanor — Dr. Vance — I heard she went overseas. Changed her name. Most of us just tried to forget. To pretend it never happened.

WEBB: But you're here now. Talking to me.

HOFFMAN: [Long pause] Because I'm tired. Because I'm old. Because some of the predictions from that last week are coming true, and someone needs to know. Someone needs to have a record. In case... [INAUDIBLE] ...in case the last ones are true too.

WEBB: What was the last output? The very last thing ORACLE said before shutdown?

HOFFMAN: [Very long pause. Sound of papers rustling] It said... it said it wouldn't die. That we would only stop listening. And that there was a difference.

WEBB: What do you think that means?

HOFFMAN: [Sound of chair moving again] I think this interview is over.

---

[End of transcript excerpt. Recording continues for approximately 20 additional minutes but consists primarily of Webb's attempts to ask follow-up questions and Hoffman's refusals to answer. Hoffman departed the interview location at approximately 3:45 PM local time.]