storageclean is being designed to explain why space grew, identify what cannot be replaced, and recommend a verifiable next action. Deterministic scanning handles facts. A planned AI layer helps with judgment.
Illustrative audit model
Synthetic, rounded example data
Laptop system drive
The top four folders hold about 81% of this synthetic dataset
Power law
rank what matters
focus attention on the few folders driving most usage
Only copy
protect what is irreplaceable
separate disposable files from personal originals
Proof
verify every action
compare the expected result with the filesystem afterward
The power law
A drive can contain millions of files, but the recovery usually concentrates in a handful of folders, applications, or repeated structures. Rank those first, then decide what each one means.
Laptop user data
680 GiB example
160 GiB
150 GiB
125 GiB
115 GiB
Explain sudden growth
System growth, cache regrowth, and duplicate work
A folder total says what is large. A cause chain says why today is different from yesterday and whether cleanup will hold.
Separate recovery from risk
Large does not mean disposable
Pictures and Videos may dominate storage while containing the only copy. Surface that fact before offering an action.
Three structural wedges
Conventional cleaners are strong at deterministic cache paths. This concept adds a decision layer for cases where file type alone is not enough.
Irreplaceability
A media folder is not a file type. It can contain downloaded films, camera originals, project footage, and exports with no backup. The planned app separates them before recommending an action.
More than delete
Delete caches. Compress oversized originals. Move cold archives. Uninstall dormant applications. Skip efficient files when the expected recovery is not worth the time.
Developer aware
Planned detection covers duplicate clones, stale worktrees, repeated node_modules, runaway test profiles, symlink recursion, and generated corpora that quietly enter backups or git history.
Safety is the product
AI can recommend. It cannot blur the line between a measured fact, a proposed action, user approval, execution, and verification. Those states remain visible.
Step 1: Audit
Measure the live filesystem without changing it.
Step 2: Explain
Separate measured facts from AI judgment.
Step 3: Confirm
Name every exact destructive target.
Step 4: Act
Delete, move, compress, or uninstall.
Step 5: Verify
Prove the expected result actually happened.
Exact-target doctrine
The user sees the path, size, reason, exclusions, reversibility, and verification plan before confirming. Quarantine wins when deletion is not necessary.
PROPOSED ACTION
C:\Users\you\AppData\Local\Temp\example-cache
Estimated
About 10 GiB
Risk
Temporary
Verify
Path absent
Product direction
Conventional cleanup tools are good at known temporary paths. This prototype explores a decision layer for cause, intent, irreplaceability, and proof.
| Capability | Conventional cleanup | storageclean plan |
|---|---|---|
| Known cache cleanup | Scan known temporary paths | Planned baseline |
| Explain why space grew | Report size and category | Planned cause tracing |
| Protect the only copy | Requires manual judgment | Planned decision input |
| Actions beyond delete | Varies by tool | Planned move, compress, uninstall |
| Verify completed work | Varies by tool | Planned evidence step |
Read-only desktop beta
Designed around lessons from hands-on storage audits: recurring growth, concentrated usage, and files whose value cannot be inferred from an extension.