Explainers · NTFS in plain terms · July 2026
Your PC wrote down every file it has. Finding one still takes a minute.
NTFS (the filesystem Windows has run on since the Windows NT line) keeps a running catalog of every file on the drive: its name, its size, where it lives, when it last changed. The operating system updates that catalog automatically, as a side effect of normal use. So your machine already knows, right now, exactly what it holds. The slow part was never the knowing: it's the tools that go looking anyway. Here is how NTFS tracks your files, and why a catalog the OS maintains for free is the key to searching a whole drive in microseconds.
TL;DR · Key takeaways
NTFS already keeps the list. The filesystem maintains a live, whole-drive catalog of every file (name, size, location, timestamp), updated automatically as you work. Your PC already knows what it holds; nothing needs to scan.
Slow search is self-inflicted. Most tools ignore that catalog and re-walk the folder tree on every query, which is why finding one file can still take a minute (Windows Search: 67 s).
Read it and search is instant. A single-file lookup returns in 85 µs (a 139 µs median across 4.47M files), about 412,000× faster than VS Code and ~480,000× faster than Windows Search.
Contents, not just names. Content queries run in ~7 to 9 ms: a search that takes ripgrep 93.8 s returns here in 16 ms (58,625×), and an agent file-hunt of 6m57s over 71 tool calls collapses to one 16 ms call (~7,200,000×, ~58% of the context window down to almost none).
Light, fresh, private. Idle around 44 MB, searchable ~1 ms after you hit save, fully offline. Everything local is free, forever: the app, whole-machine search, MCP. Hosting starts at $5.99/mo.
The filesystem under Windows
What NTFS is, in one page.
NTFS stands for New Technology File System. It arrived with Windows NT in 1993 and has been the default on consumer Windows since XP. A filesystem's job is to turn a flat expanse of disk into named files and folders and remember where every byte lives. NTFS does that, plus four things worth knowing, one of which quietly makes fast search possible.
A master catalog
One central record listing every file and folder on the volume. It is how the OS answers 'what is here?' without re-reading the disk: the list is already written down.
Rich metadata per file
Timestamps, size, and access permissions (who is allowed to read or write it) travel with each file, not just its bytes. That is why Windows can enforce who sees what.
Crash-safe by design
NTFS is a journaling filesystem: it notes what it is about to change before it changes it. Pull the plug mid-write and the drive recovers to a consistent state instead of corrupting.
Built for big
Volumes and files far larger than a spinning disk of 1993 ever held, plus optional per-file compression and encryption. The format aged well because it left itself room.
Most of this you never see. The one that matters for search is the first: NTFS keeps a catalog, and it keeps it current.
How the OS knows what exists
The filesystem keeps a catalog. You get it for free.
Every time you create, save, rename, move, or delete a file, the operating system updates that one central record. You never asked it to: it is simply part of how the drive works. The upshot is quietly powerful: at any instant, the machine already holds a complete, current list of every file it has. Nothing needs to scan the disk to produce that list, because the disk has been keeping it all along.
Read it top to bottom. Ordinary file operations keep the catalog current at no extra cost. That is the whole trick. Then there are two ways to answer "where's my file?" read the list the OS already compiled (microseconds), or ignore it and re-open every folder yourself (seconds to minutes). Almost every everyday search tool does the second.
The part everyone gets halfway
Two famous tools tried to fix this. Both stopped short.
If the catalog is already there, why is search slow? Because most tools do not read a catalog at all: they walk the folder tree from scratch on every query. find, grep, ripgrep, a directory listing, VS Code's file picker: each one re-opens folder after folder, every time you ask.
Two well-known products set out to end that. Spotlight on macOS and Windows Search each pre-build their own index so a lookup can read instead of walk, exactly the right instinct. But both stop short in the same two ways. Each indexes only a slice of the drive (your user folders, the file types it recognizes), so anything outside that slice drops back to a live disk walk. And each answers only one kind of caller: a person typing into a search box. An AI agent, which does not type into search boxes, was never a caller either tool imagined.
Spotlight
Indexes your home folder and recognized file types; answers ⌘-Space and the mdfind terminal command. Fast inside its scope, but macOS-only, and there is no standard interface a cross-platform agent can call.
Windows Search
Indexes user folders and the Start menu, then falls back to a directory walk for the rest, and merges local hits with web results. Measured 67 seconds, median, on the drive below.
Same drive · same question · find one file
The same lookup, five ways.
| The tool | How it answers | Time · 4.47M files | vs Interlinked |
|---|---|---|---|
| Interlinked | reads a live, whole-drive index | 85 µs | 1× |
| Windows Search | indexes a slice, walks the rest | 67 s | ~480,000× |
| VS Code (Ctrl+P) | walks the folder you opened | 35 s | 412,000× |
| ripgrep (content) | walks the disk, scans the bytes | 93.8 s | content query |
| Claude Code (no index) | shells out, lists directories | 3 m 13 s | 2,276,000× |
Methodology: Ryzen 9 9950X3D · 64 GB DDR5 · NVMe · Windows 11 · 4,470,000 files. Interlinked, Windows Search, VS Code and Claude Code rows are single-file lookups; the multipliers are the published head-to-head figures (85 µs single-file lookup versus VS Code's 35 s = 412,000×; the median find across 20 queries is 139 µs, versus Windows Search's 67 s ≈ 480,000×). The ripgrep row is a content query over the same corpus: the same query it drives takes 93.8 s there and 16 ms here, 58,625× across the set. Spotlight is macOS-only and cannot be benchmarked on this Windows machine; its limits here are coverage and the missing agent interface, not raw speed inside its own scope.
Reading the catalog, all the way
Now point the idea at the whole machine. Names and contents.
Take the instinct those two tools got right and remove the two compromises. Index every file on every volume, not a slice. Index what is inside them too, not only their names. Keep it fresh the instant you hit save. And expose it through a single call any agent can make. That is the entire design: an index that is always there, always current, covering the whole drive.
On the same Windows machine that is roughly 480,000× faster than Windows Search on a filename lookup: 139 microseconds instead of 67 seconds. Not because the disk got faster. Because nothing walked it.
Why it matters now
The catalog was always there. Now every agent can read it.
The OS has kept a faithful list of your files for three decades. The missing piece was a fast, whole-machine, always-fresh index that anything could query, including the AI agents that now do real work on your machine. That is what Interlinked is: whole-machine filename and content search, offline and private, answered in microseconds, behind one call that 19 AI clients (Claude Code, Cursor, Copilot, Codex, Cline and the rest) already share. No model, no embeddings, no cloud: an index and a match. And the same index carries a knowledge vault: short notes and relationships that live on your files and folders, keyed to their identity so they survive renames, and are fed back to the next agent automatically.
THE EXPLAINERS SERIES: the plumbing under fast local search, in plain terms.
Do this today
Read the catalog your PC already keeps.
Install once. One signed installer auto-configures 19 AI clients: Claude Code, Cursor, Copilot, Codex, Cline and the rest. No JSON editing, no per-tool wiring.
Let it index. Filename search works immediately; content search fills in behind it. At rest the whole engine sits around 44 MB, and the index on disk stays under 1% of the drive.
Ask from any agent. Whole-machine when you don't know where something lives, scoped to a folder when you do. Either way the answer comes back in microseconds, machine-wide, offline.
Everything local is free forever, on all your devices, no card. Hosting starts at $5.99/mo. Download for Windows.