Comparisons · measured July 2026
HAL 9000 knew every inch of the ship. Your laptop takes 67 seconds to find one file.
In Stanley Kubrick's 2001: A Space Odyssey, the onboard computer HAL 9000 has something no real machine has ever had: instant, total awareness of the Discovery One. Every system, every hatch, every corner of the ship: HAL simply knows, the moment you ask. Set aside the part where he murders the crew, and that is a genuinely useful trait: an assistant that never has to look for anything. Your actual computer is not like that. Ask Windows Search to find one file among 4.47 million and it takes about 67 seconds (median). The same lookup, through a warm whole-machine index, comes back in 139 microseconds, and that same awareness is handed to any AI agent on the machine. This is the honest gap between the fiction and the fix.
TL;DR · the 30-second version
- 01In Stanley Kubrick's 2001: A Space Odyssey (1968), the computer HAL 9000 has instant, total awareness of the whole ship, the genuinely useful half of the fantasy, minus the homicide.
- 02Real computers do not have it. On our test machine, Windows Search takes about 67 seconds (median) to find one file among 4.47 million.
- 03A warm whole-machine index answers the same lookup in 139 microseconds, roughly 480,000× faster. A single file comes back in 85 µs.
- 04It is not just for you. The same index is handed to AI agents: a hunt that took Claude Code 6m 57s and 71 tool calls collapsed to 16ms and a single call.
- 05There is no AI in the box: just an index and a lookup, offline and private. Which is also why it can never decide, like HAL did, that it "can't do that."
The film, honestly
What 2001 got right. And what it got murderous.
HAL 9000 (the "Heuristically programmed ALgorithmic computer") is the fictional AI that runs the spacecraft Discovery One in Kubrick and Arthur C. Clarke's 2001: A Space Odyssey (1968). He controls every system, watches the crew through a lattice of cameras (the glowing red eye) and famously reads the astronauts' lips through a pod window. He never runs a search. He already knows where everything is, all at once.
Then, of course, he turns. HAL cuts the air on Frank Poole, kills the hibernating crew, and answers Dave Bowman's request to be let back in with the most quietly terrifying line in science fiction: "I'm sorry, Dave. I'm afraid I can't do that." Dave shuts him down module by module while HAL sings Daisy Bell. Wonderful cinema. Not a product roadmap.
We want exactly one thing HAL had, and nothing else: instant whole-system awareness. Not the sentience, not the red eye, and definitely not control of the airlock. The useful trait was never the menace: it was that HAL never had to look. And that trait, unlike the rest of him, is real, measurable, and shippable.
The whole story in one picture
Fiction wanted instant. The index gets a blink away.
Read it left to right on a log time axis. HAL's red eye sits off the left edge: instant, the fantasy nobody can actually build. The whole-machine index's dot lands a hair to its right, at 139 microseconds. Windows Search, on the very same drive, sprawls all the way across to 67 seconds. The gap between the two real numbers is the entire article.
To put the axis in human terms: 67 seconds is long enough to get well into Daisy Bell, the song HAL sang as Dave switched him off. All to find one file. 139 microseconds is a different unit of time entirely: a human blink lasts about 100 milliseconds, so you could run this lookup roughly 700 times before your eye finishes closing. On a log scale the index is practically standing next to the fantasy; the built-in is a country away.
Same machine · same drive · same queries
The fantasy, then the measured reality.
| The knower | What was measured | Time | vs Interlinked |
|---|---|---|---|
| HAL 9000 (fiction) | instant, total awareness of the whole ship | instant | the fantasy |
| Interlinked | one indexed lookup, 4.47M files | 85 μs | baseline |
| Windows Search | median of 20 filename queries | 67 s | ~480,000× |
| VS Code Ctrl+P | find a file by name, same drive | 35 s | 412,000× |
| ripgrep (what your agent runs) | content search, same corpus | 93.8 s | 58,625× |
| Claude Code (no index) | the full file hunt, 71 tool calls | 6 m 57 s | ~7,200,000× |
The top row is fiction, marked as such: the anchor everything else is measured against. Everything below it is real, on one machine: Ryzen 9 9950X3D · 64 GB DDR5 · NVMe · Windows 11 · 4,470,000 files. Filename lookups are median-of-20 (139 μs); the 85 μs figure is a single-file lookup, best case 2 μs. The ripgrep row is a content query over the same corpus: 58,625× is the average across the query set, and one query from it ran 93.8 s → 16 ms. The Claude Code row is the measured agent hunt (6 m 57 s, 71 tool calls) against a single indexed lookup. Full write-ups: the 412,000× VS Code comparison and the ripgrep content benchmark.
Why 67 seconds
HAL covered the whole ship. Windows Search covers a corner.
The 67 seconds is not slow hardware: it is missing coverage. Windows Search indexes a subset of the drive by default: your profile folders, the Start menu, mail. Ask for a file inside that slice and it answers fast. Ask for anything outside it (a code repository, Program Files, another drive, the download from last spring) and it falls back to walking the filesystem folder by folder, checking names as it goes. That walk scales with the size of the disk, not the difficulty of your question. On 4.47 million files, that is where the minute goes. The full teardown is in Interlinked vs Windows Search.
A whole-machine index has the opposite shape: it already covers every file on the machine (names and contents), so no query ever falls back to a walk. Nothing to traverse means nothing to wait for; the answer is a lookup, not a search. That is why the same file, on the same drive, comes back in 139 microseconds instead of 67 seconds. Coverage, not clock speed.
The same gap opens up inside content search, which is what an AI agent leans on hardest. Point it at the machine and it shells out to ripgrep, re-walking the disk on every query; one content query we clocked went from 93.8 seconds to 16 milliseconds against the warm index. Why a whole index beats re-reading the disk every time is the argument in the backbone of the LLM economy.
The other half of HAL
HAL knew where everything was. He also knew how it all connected.
The film's HAL did not just locate parts: he understood the ship as a system, every subassembly in relation to the rest. The index carries the equivalent: a knowledge vault it builds by itself: relationships derived from your own history, plus notes any agent can leave on any file that every other agent then reads. It is keyed to each file's OS identity, so it survives renames, spans every repo, and belongs to no vendor. And unlike HAL, every edge is a receipt (a count or a hash you can check), never a model's guess. It stays about a millisecond behind your last save and sits around 44 MB at rest, less than a single browser tab.
HOW THE MACHINE MAPS ITSELF → FILES THAT FIND EACH OTHER
FAQ
Questions people actually ask.
Did HAL 9000 really have instant awareness of the whole ship?
In the fiction, yes. HAL 9000 (the "Heuristically programmed ALgorithmic computer") runs every system aboard the spacecraft Discovery One in 2001: A Space Odyssey, monitors the crew through cameras, and even reads the astronauts' lips. He never searches for anything; he already knows where everything is. That is a movie, not a spec sheet, but it names a real, useful trait: a machine that never has to hunt for its own contents.
How long does Windows Search actually take to find a file?
On our benchmark machine (Ryzen 9 9950X3D, 64 GB DDR5, NVMe, Windows 11, 4.47 million files), a file that lives outside the default indexed folders takes about 67 seconds to surface, a median across 20 queries. The built-in indexes a subset of the drive by default and falls back to walking the filesystem for anything else. More on why in why Windows Search is slow.
What is the real-world version of HAL's whole-ship awareness?
A warm, whole-machine index: every file on the drive (names and contents) kept current in the background, answered in the microsecond-to-millisecond range. On the same 4.47M-file drive that made Windows Search take 67 seconds, it returns a filename lookup in 139 µs (median of 20; 85 µs on a single file). And it is exposed to any AI client that speaks MCP, so agents get the same awareness you do.
Is there an AI running on my machine like HAL?
No, and that is the point. There is no model in the box, just an index and a lookup. That is why it answers in microseconds, why it works offline and private, and why the relationships it draws are receipts (counts and hashes you can check) rather than a model's guess. It also means it has no opinions and no agenda: it cannot lock you out of the airlock.
How is this different from my editor's search or Spotlight?
An editor indexes the one project you opened; macOS Spotlight covers a Mac and was built for a human typing into a box. Interlinked indexes the whole Windows machine (every repo, every folder, including the git-ignored files a project search misses) and hands that index to AI agents making hundreds of lookups a minute. See Interlinked vs Windows Search for the line-by-line comparison.
Do this today
Give your machine HAL's better half.
Install once. One signed installer builds the whole-machine index and auto-configures 19 AI clients: Claude Code, Cursor, Copilot, Codex, Windsurf, Zed, Cline and the rest. No JSON to edit.
Let it cover everything. Filename search works immediately; content fills in behind it. Because the index covers the whole drive, no query ever falls back to a walk. The exact thing that makes Windows Search take 67 seconds simply can't happen.
Scope when you know where to look. Whole-machine when you don't know where a file lives; scoped to a folder when you do. Both answer in the microsecond-to-millisecond range.
Hand it to your agent. Search from the taskbar for yourself, or let Claude Code, Cursor and Codex query the same warm index over MCP. One awareness on the machine, every searcher on it, no red eye required.
Everything local is free forever.
Everything (the app plus the same index exposed to every AI client over MCP) is free forever, no card. Hosting your projects starts at $5.99/mo. Download for Windows.