For the clinician a patient hands a report to
A summary of your patient's own CGM data — and nothing you didn't already decide.
Eating Glasses turns a patient's continuous glucose history into a clear, cited, observations-only summary for their visit with you. It surfaces patterns. It does not recommend a dose, diagnose a condition, or predict glucose — those stay with you.
A general-wellness tool operating in the FDA general-wellness enforcement-discretion lane, not a medical device. If you're reading this after a patient's appointment, this page exists so you can verify it in a few minutes.
Where the line is
What it does, and what it never does
The distinction is the whole design, so it leads — not the fine print. Every report carries this footer verbatim: "Generated from CGM data for information only; contains no treatment recommendations."
Never
- Calculates or suggests an insulin dose, carbohydrate ratio, correction factor, or basal setting
- Diagnoses any condition
- Predicts a future glucose value, or issues real-time alarms
- Tells a patient to change their therapy
Does
- Computes the consensus CGM metrics
- Surfaces ranked, time-stamped past patterns
- Writes a plain-language question the patient brings to you
- Re-measures the same metric after a wellness change the patient chose to try
Methodology
Deterministic, on-device, auditable
Patterns are computed by a fixed, rules-based engine that runs on the patient's phone. The same data always produces the same report. Glucose is never sent to a large-language model or any third party — there is no generative step in the analysis, and so nothing to hallucinate.
The metrics follow the standard consensus definitions, with the Glucose Management Indicator computed as GMI = 3.31 + 0.02392 × mean mg/dL and CV = SD / mean × 100%. The cutpoints and the GMI constant are pinned as unit-tested constants, so the app, the PDF, and this page never diverge.
What it cannot see: it has no insulin, pump, or meal data unless the patient supplies it, and it does not infer dosing. When CGM coverage is too sparse to summarize, it says so rather than guessing.
Evidence
Every clinical claim has a primary source
Each reference point is tied to a primary citation with the population noted (type 1 vs. general). The consensus targets and the GMI definition follow the international CGM consensus (Battelino et al., Diabetes Care, 2019) and the Glucose Management Indicator (Bergenstal et al., Diabetes Care, 2018), alongside the current ADA Standards of Care.
Citations last verified: [pending final re-verification before publish]. Population and edition are noted on each so you can confirm the source supports the exact claim.
The artifact
The one-pager your patient brings
| Metric | Value | Consensus target |
|---|---|---|
| Time in Range (70–180) | 88% | >70% |
| Time in Tight Range (70–140) | 71% | — |
| Time Below Range (<70) | 2.2% | <4% |
| <54 | 0.3% | <1% |
| Time Above Range (>180) | 9.8% | <25% |
| >250 | 2% | <5% |
| Mean glucose | 128 mg/dL | — |
| GMI | 6.4% | — |
| Coefficient of variation | 32% | ≤36% |
Time in Tight Range (70–140) has no established consensus target — an emerging metric; the dash reflects that, not a missing value.
Observations, ranked and time-stamped
- Readings below 70 on 24 of 90 nights, onset clustered around 03:00; three sustained below 54.
- Evening runs higher: a median near 173 mg/dL around 8pm versus about 120 mg/dL across the day, with the widest day-to-day spread in that window.
- A pre-wake rise of roughly 35 mg/dL on about half of mornings.
In practice
How your patients use it
It reads CGM glucose from Apple Health automatically — no logging — and turns it into the profile above plus one focused question and this one-pager. The intent is narrow: the 15-minute visit can start at the decision, not the data review.
When a patient tries a general wellness change, the app shows the same metric before and after. It is an association, n-of-1, not adjusted for confounders; it describes what changed or differed, never what "worked," and it never contradicts your plan.
Data, privacy & security
The liability question, answered plainly
Glucose is read from Apple Health and analyzed on the patient's device; it is never sent to a large-language model and is not stored on our servers. Data in transit and at rest is encrypted, records are row-level isolated per account, the patient controls all sharing, and account deletion is available in-app.
HIPAA does not apply to a direct-to-consumer app like Eating Glasses — so we do not claim it. We are built to meet the requirements of the FTC Health Breach Notification Rule and the Washington My Health My Data Act. See the Consumer Health Data Privacy Policy and the Privacy Policy.
Spot an inaccuracy? Reference a report by its share link rather than emailing patient-identifiable information.
Accountability
Who's behind it
Eating Glasses LLC is an independent company in Washington, DC. Its founder, Justin Silver, has lived with type 1 diabetes for decades (Dexcom G7, Omnipod 5) and has spent his career building the clinical systems that hospitals and digital-health companies run on. Eating Glasses is built to the standard that combination demands: evidence-cited, deterministic, and clear about its limits.
Methodology authored by the founder; formal clinical review is in progress. Decades of living with type 1, and a career building clinical software, are not a substitute for clinical authority, and we do not present them as one. Methodology last reviewed: [date at publish]. Contact: hello@eatingglasses.com.
If it's useful to your patients
There's nothing to sign up for
No clinician account, no portal, no referral program, and no incentive of any kind — the report belongs to the patient and is shared at their discretion. If you'd like to point patients to it, a line that works:
"There's a wellness app called Eating Glasses that turns your CGM into a plain-language summary you can bring to our visits. It doesn't touch your insulin or your doses, and your glucose stays on your phone."
You're welcome to print this page for the exam room. To flag an inaccuracy or ask a question: hello@eatingglasses.com — please reference a report by its share link rather than any patient-identifiable detail.