00 — Mission Refresh, 2026

Read this before the original vision-mission doc. This document expands the “protect seniors from fraud” mission into a 2026-relevant, AI-era mission. Both documents are load-bearing — neither replaces the other.

The 2026 Mission Statement

Glance gives every human back the time and attention that the automated communication world is taking from them — by putting a trusted intelligence between you and the noise.

Concretely:

• For seniors at risk: keep the fraud and impersonation out — by Companion approval, not by AI guesswork alone.
• For busy professionals: keep the AI-generated noise out — sales SDRs, AI cold outreach, AI-summarized newsletters, AI-rephrased phishing — none of it reaches you unless someone you trust (or your own learned preferences) says “yes, this person, this time.”
• For families: connect the right humans to each other — your kids, your bank, your doctor, your dentist — without exposing any of them to the open-internet sewage.

The unifying thread is time-and-attention recovery in a world where the marginal cost of sending a personalized-looking message has dropped to zero. That world arrived between 2024 and 2026. Glance is the answer to it.

Why this mission, why now

Three irreversible 2024–2026 shifts created the opening:

1. AI made personalized communication free

A fraudster used to need a copywriter, a translator, a domain, and a reason to target you. ChatGPT, Claude, and open-source LLMs reduced all four to “free + 30 seconds of prompting.” Phishing volume is up ~1,200% YoY (2025 Anthropic threat report); the sophistication ceiling collapsed.

2. Big-tech inboxes ran out of detection signal

Gmail, Outlook, and Apple Mail rely on sender reputation + content heuristics + ML scoring. All three signals degrade against AI-generated mail because:

• The sender is real (compromised mailbox or freshly-warmed domain)
• The content is grammatically perfect and tonally appropriate
• The ML model was trained on yesterday's adversarial distribution

The big platforms can ship faster models, but they can't ship a trust graph — they don't know who your daughter, dentist, or accountant is. Glance does, because the user (or their Companion) tells us.

3. The cost of “missing one” went up

A 2010 missed phishing email cost a few dollars. A 2026 missed phishing email — synthetic-voice follow-up, deep-fake video confirmation, a real bank account number scraped from the LinkedIn breach — costs five-to-six figures and, for the elderly, often costs the relationship between a parent and their adult kids (“how could you fall for that”). Insurance won't price the risk. Banks won't reimburse it. Big tech won't fix it. Someone has to be the bouncer. Glance is that bouncer.

The “save time and connect humans” framing

The old framing — “fraud protection for seniors” — was true but narrow. It implied defense. The 2026 framing is defense + curation:

The expansion does not dilute the senior-protection mission — it makes it more sellable, because the same product also serves the senior's adult child, who is the actual buyer.

The strategic moat: trust graph + zero-knowledge + on-device ML

Three load-bearing differentiators against AI-first competitors:

1. The Circle of Trust is training data

Every Companion approval is a high-quality labeled signal — far higher quality than any third-party reputation feed or ML inference, because it was made by a human who knows the recipient personally. Glance's local LightGBM model trains on this signal continuously. Sublime Security, Abnormal Security, and Microsoft Defender don't have it because their customer is the IT department, not the family.

2. Zero-knowledge by construction

• 24-hour Train-Distribute-Destroy loop on threat metadata
• Email bodies scanned in memory, never persisted
• Per-user encryption keys; Glance can't read what it just verdicted
• No bank connections. Ever.

This is provable in code and audit-able by anyone. Sublime, Abnormal, Material — all of them retain message contents for “investigation.” We don't, and won't.

3. On-device + sub-50ms inference

The local LightGBM ONNX model (<50MB, V1=100%, F1=0.985) runs in the user's browser/PWA. It's not “secure cloud AI” — it's “AI that doesn't see your email.” The mobile-app strategy (Apple Wallet pass, Google Wallet pass, iOS MDM profile) means we never need a binary on the App Store, which means we ship faster than any competitor, including Microsoft.

What stays out — unchanged

Every “what we will not build” item from the original mission still holds. The expansion adds adjacency, not scope creep:

• No corporate inbox security. We are not Proofpoint.
• No general-purpose productivity tool. We are not Superhuman or Notion Mail.
• No AI-generated replies sent without explicit user consent. Glance does not pretend to be you.
• No bank connections. Ever.
• No data resale. Threat patterns flow into the network and are wiped.

How this mission shows up in the next 90 days

What the April 2026 competitor research changed

Key risks and mitigations identified from the 18-competitor analysis (April 2026):

1. Apple Intelligence ships "Family Mail Approval" — Catastrophic for iOS-only segment. Mitigation: Own cross-provider coverage early; build the family graph Apple cannot replicate (Apple's graph is single-Apple-ID).
2. Microsoft bundles Defender into M365 Personal/Family ($10/mo) — High for Outlook segment. Mitigation: Cross-provider + family + zero-knowledge wedge holds; M365 Family is small share of senior email.
3. AI-generated phishing surges past 95% — Already at 82.6%. Mitigation: This helps Glance vs. naive heuristic vendors. Lean in.
4. Anthropic ships "Claude email assistant" — Medium probability in 18mo. Mitigation: Glance's value is the trust graph, not the AI. Even the best AI doesn't know who your daughter is.
5. Apple Intelligence becomes consumer default for "on-device privacy" — High probability. Mitigation: Combine on-device WITH the uniquely differentiated Family Circle + cross-provider stack — the combination is the moat.