The Playbook

Ten ideas stolen from the best of the AI security market, filtered through Quint’s thesis, and converted into shippable specs

The Playbook

What this is. A deliberate inventory of the best ideas in AI security — surveyed across 22 competitors, ranked by leverage, and rewritten as concrete specs Quint can ship. Every page here answers three questions: who did it first, why it matters, and what Quint’s version looks like.

What this isn’t. A roadmap commitment. The priorities stack shifts with design-partner conversations and the divergence-detector sprint. Treat each page as an option with a known cost and a known payoff, not a scheduled deliverable.

Why a playbook

The AI security market produced six acquisitions totaling ~$1.8B between March and September 2025. Between March and May 2026, three incumbents — CrowdStrike, Microsoft, Palo Alto — announced endpoint-level AI agent security in six weeks. The category is being defined in real time, by someone, whether Quint participates or not.

The risk isn’t missing an idea. The risk is scattering attention across ten ideas and shipping none. This playbook exists so Quint can pick deliberately.

The three filters

Every page in the playbook was stress-tested against three rules:

1. Does it deepen the intent-vs-truth moat, or chase someone else’s game? Adding a WAF = chasing F5. Adding an AIBOM = deepening our visibility story.
2. Does it build the behavioral corpus faster? Anything that gets more agent sessions into Quint’s data path is worth doing. Anything else is a distraction.
3. Can a one-person engineering team actually ship it? Bounty programs are cheap. Federated learning isn’t.

Ideas that failed all three filters are not in this playbook. What survived:

The ten

Shadow AI Discovery

Inspired by: CrowdStrike AIDR + Microsoft Agent 365. Every CISO’s first question is “what’s running?” — answer it as a standalone SKU before selling detection.

Agent Bill of Materials

Inspired by: HiddenLayer + CycloneDX ML-BOM. Per-session AIBOM export (JSON + PDF + Ed25519 signed). Satisfies EU AI Act Annex IV, NIST AI RMF, ISO 42001.

Security Graph + Attack Paths

Inspired by: Wiz Security Graph. 14 node types × 16 edge types. The divergence overlay is the moat — no other vendor has both streams to render.

Category Naming

Recommended: Commit to “Agentic AI Security” (per Gartner’s April 2026 Hype Cycle). Use “AI Agent EDR” as the sales soundbite only.

SKU Packaging

Inspired by: SentinelOne/Prompt Security + Snyk. Three SKUs: Quint Code ($18/dev), **Quint Desktop** ($12/endpoint), Quint Agentic ($25/agent +$2k platform).

Closed-Loop Signatures

Inspired by: F5 AI Remediate + CrowdStrike IoC distribution. Divergence detection → FlowMatrix distillation → fleet push in <30s. Most of the architecture already exists.

Quint Labs (Research)

Inspired by: Noma Security + Wiz Research + Unit 42. First research brand targeting CVE disclosures in Cursor/Claude Code, MCP servers, and incumbent blind spots.

Quint Lab (CTF)

Inspired by: Lakera’s Gandalf. “Escape the Watchdog” — invert Gandalf. Player tries to exfiltrate a canary past Quint’s divergence detector. 10 levels. Lead capture at level 4.

IDE Extension

Inspired by: Snyk + GitGuardian. Thin VS Code / Cursor extension talking to the local daemon. Status bar + inline risk diagnostics + approval modal.

Air-Gapped Deployment

Inspired by: TrojAI + Lakera self-hosted. Three tiers: BYOC (Terraform), on-prem (Helm), true air-gap (Zarf bundle). Defense primes first.

The priority stack

Each idea has an impact/effort profile. Ranked (highest leverage first):

#	Idea	Effort	Impact	Blocks What
1	Shadow AI Discovery	~16 eng-days	Every first sales conversation	Nothing — ship it while divergence detector is in flight
2	Category naming commitment	0 eng-days	Every pitch, every landing page	Blocks brand/SEO work
3	Quint Labs — first research piece	2-3 days writing	Investor credibility + organic top-of-funnel	Hamza availability
4	SKU repackaging	0 eng-days	Every design partner conversation	Landing page rewrite
5	Agent Bill of Materials	~2 weeks	EU AI Act / SOC 2 sales	Compliance-driven deals
6	Security Graph (MVP)	~2.5 weeks	The demo moment after divergence card	Design partner validation
7	IDE extension (VS Code)	~5 weeks	Developer love + daily engagement	After design partner signed
8	Closed-loop signatures	~5-7 weeks	The fleet-learning narrative	After Stage 1 exits shadow
9	Quint Lab CTF	~5 weeks	Viral top-of-funnel, research credibility	Hamza bandwidth, post-first-DP
10	Air-gapped (Phase 0)	~2 weeks	Optionality for defense primes	Only if government deal is in flight

What NOT to steal

This list matters as much as the one above. Explicitly rejected:

• Inline LLM gateway / prompt rewriting — SentinelOne owns this via Prompt Security. Not winnable from our position.
• AI-SPM / cloud posture scanning — Wiz already won this. Don’t touch.
• Model weight scanning — Protect AI, HiddenLayer, TrojAI own the model artifact layer. Different threat model.
• Pure governance SaaS — Credo AI’s compliance dashboards without technical scanning are a dead-end for a security product.
• SASE / network proxy — WitnessAI, Cato, Zscaler own the network tier. Quint’s wedge is on-endpoint.
• Windows-first expansion — Microsoft Agent 365 bundles “endpoint AI governance” into E5. Competing on Microsoft’s home field is a losing game. Linux/K8s before Windows.
• GNN in production — No trained weights exist. The 500K labeled sessions required don’t exist. Ship when the data does, not before.

How to read this playbook

Each page follows the same structure:

1. What the market is doing — who built it, how it looks, what made it work.
2. Why Quint’s version is different — the intent-vs-truth thesis applied to this idea.
3. Concrete spec — schema, architecture, file list, where it lives in the stack.
4. Effort + sequencing — eng-days, blocking dependencies, first-week sprint where applicable.
5. Open questions — things that need a design partner or product decision before shipping.

These are living documents. When a design partner signs, or the divergence detector ships, or an idea rotates from “roadmap” to “in flight,” the page changes with it.

The one thing

If forced to pick a single page from this playbook to ship first, it’s Shadow AI Discovery. Every competing pitch starts with “we catch attacks.” CISOs don’t buy attacks. They buy visibility. Shadow AI Discovery is the Trojan horse for the divergence detector. Sell the discovery. Upsell the detection.