How We Keep Your Computer Safe
Letting software fix your computer requires trust. Here is exactly how we earn it — no marketing fluff, just the real architecture.
Our Safety Promise
- Every fix is checked three separate ways before it touches your system
- A full snapshot of your system is saved before every change, so we can undo anything
- Every fix package is cryptographically signed — tampered packages are rejected instantly
- You always have the final say — nothing runs without your explicit approval
- Every action is logged in an immutable audit trail you can review anytime
Three Layers of Protection
Before any fix runs on your computer, it passes through three independent safety checks. If any layer says "no," the fix is blocked. No exceptions, no overrides.
Pattern Guard
Instant check — under 1 millisecond
A blocklist of known dangerous commands. Things like "delete everything" or "wipe the hard drive" are caught immediately and rejected. Think of it as a bouncer at the door — obvious troublemakers never get past the entrance.
Impact Analysis
Structural check — under 10 milliseconds
The command is broken down into parts and analyzed for what it actually affects. How many files? Which system services? Can the change be undone? A "blast radius" score from 0 to 100 determines whether extra precautions are needed. High-impact changes require a system snapshot first.
AI Safety Review
Context-aware check — under 2 seconds
A dedicated safety AI (separate from the one that proposed the fix) reviews the command in context. It considers: what problem is being fixed, what the command does, what could go wrong, and whether there is a simpler alternative. If it is not confident the command is safe, it either asks you for permission or blocks it outright.
Automatic Snapshots — Your Safety Net
Before every fix, we take a complete snapshot of your system using your operating system's built-in technology. If anything goes wrong — or if you simply change your mind — we can roll your computer back to exactly how it was before.
Automatic Rollback
If the fix fails or you're not satisfied, one click rolls your computer back to exactly how it was before. Automatic health monitoring is on the roadmap.
Post-Fix Verification
After every fix, we re-scan to confirm the problem is actually resolved. If it is not, the fix is marked as failed and rolled back.
Cryptographic Verification
Every fix package that runs on your computer is digitally signed. Before execution, the signature is verified. If someone tampers with a fix — even changing a single character — the signature check fails and the fix is rejected.
We use Sigstore, the same open-source signing infrastructure used by Kubernetes, npm, PyPI, and other critical software ecosystems. Every signature is recorded in a public transparency log, so anyone can independently verify the integrity of our fix packages.
Trust Tiers — Not All Fixes Are Equal
Different fixes have different levels of validation. We are transparent about this so you can make informed decisions.
Tier 1 — Vendor Certified
Created and signed by the software vendor themselves (e.g., Microsoft, Apple, Adobe). Audited by our platform. The highest level of trust.
Tier 2 — Community Validated
Created by community experts and validated through automated testing, sandbox execution, and real-world success rates. Earns Bronze, Silver, Gold, or Platinum status based on track record. You approve each fix before it runs.
Tier 3 — AI Generated (Experimental)
When no certified fix exists, our AI can propose a solution. These are clearly labeled as experimental, tested in a sandbox first, and require your explicit approval with full risk disclosure. Opt-in only — disabled by default.
Complete Audit Trail
Every action is logged: what was scanned, what was found, what fix was proposed, whether you approved it, what happened during execution, and whether the problem was resolved. These logs are immutable — no one, including us, can alter or delete them.
This is not just good engineering — it is a legal requirement under California AB 316 and the upcoming EU Product Liability Directive. We maintain these records as evidence of reasonable care in our AI-assisted repair process.
You Are Always in Control
Our AI diagnoses problems and proposes fixes. But nothing happens to your computer without your approval. You can review what will change, approve or reject it, and undo any fix at any time.