what we are building next.
one research pipeline, three products. the research that found the issues becomes an app that prevents them, and a model that runs on the phone itself.
scan more of the long tail.
the pilot proved the method. next is scale: more countries, more sectors, and the everyday apps that run no security program. human verification and coordinated disclosure stay at the centre.
live · scaling- re-scan every reported app on each new release and publish fix rates
- record dismissed candidates so the triage model learns from real false positives
- publish anonymised case studies of fixed issues with the owner's agreement
an app that checks your other apps.
an android app that examines the other apps on your phone the way we examine them in the lab: what they bundle, what they can take, where they send it. a plain-language verdict first, the full static report on tap. nothing leaves the device.
in design- safe to use · use with care · avoid, explained in one screen
- a researcher mode with findings, call paths and masvs mapping
- no apk ever leaves the phone unless you choose a deeper cloud review
a small model that runs on the phone.
a small, security-tuned model that reads the scanner's findings and does four narrow jobs: explain, rate, dismiss false positives, and write the verdict. trained on our own corpus of confirmed findings and the rejections that taught us what not to report.
research- about one gigabyte, runs on a mid-range phone
- never discovers flaws itself; the deterministic rules do that
- published by hash so anyone can verify what they run
now, next, later.
milestones, not dates. each gate is a measurement: the engine reproduces the lab’s findings, the model beats plain rules, a scan finishes on a mid-range phone in minutes.
now
- scale the scanning pipeline across new regions
- deepen national-cert partnerships
- publish anonymised findings and fix rates here
next
- monkguard private beta with the bug-bounty community
- severity framing agreed with vendors
- a hall of fame for owners who fix fast
later
- the on-device model, verifiable by hash
- developer self-serve review before release
- open methodology, rules catalogue and findings schema