GlycemicGPT Docs
GlycemicGPTDaily use

Daily Briefs

AI-generated summaries of your day, automatically.

A daily brief is an AI-generated summary of what happened with your glucose, insulin, and patterns over the past day. Today briefs run on a daily cadence; weekly / longer cadences are a roadmap item but not shipped yet. They show up in your dashboard automatically -- you don't have to ask for them.

Briefs are informational summaries, not medical advice. They highlight patterns the AI noticed -- often things worth discussing with your healthcare provider. They are not a clinical assessment.

What's in a brief

The AI brief generator passes your data (Time in Range, average glucose, low / high counts, Control-IQ correction count, total insulin, optional pump profile, optional IoB context) to your configured AI provider with a prompt asking for a plain-language summary that touches on patterns, post-meal spikes, overnight trends, and meaningful Control-IQ activity.

Because the brief is AI-generated prose rather than a fixed-template report, exact sections are not guaranteed -- the AI structures the response based on what your data actually shows. A typical brief mentions:

  • Time-in-Range performance for the day
  • Notable highs and lows -- when they happened and possible context
  • Insulin / Control-IQ patterns the AI noticed
  • Overnight behavior if there's anything interesting to note
  • Suggestions for follow-up when the AI sees a pattern worth pulling on

Briefs are written in plain language, not clinical jargon -- they're meant to be easy to read.

When briefs are generated

You configure this in Settings → Briefs:

  • Enable / disable the daily brief
  • Time of day -- when the daily brief gets generated (default: morning)
  • Timezone -- so the time-of-day setting reflects your local time
  • Delivery channel -- in-app (web only), Telegram (if configured), or both

If the brief is set to "morning" and there's no glucose data for the previous day, the brief will say so -- the AI can't summarize what isn't there.

Where to read them

  • In the dashboard -- a Briefs panel on the home screen and a dedicated page
  • Telegram (if configured) -- the full brief delivered as a message

Note on push notifications: brief delivery to a phone via push notification (rather than Telegram) is a roadmap item, not shipping today. If you want briefs on your phone today, the path is Telegram.

Manually generating a brief

If you want a brief outside the schedule (e.g., to summarize a specific period before an endo appointment), go to Briefs → Generate brief and pick the date range. The platform queues the request and the brief appears within a minute or two.

Briefs are different from AGP

GlycemicGPT does render an AGP (Ambulatory Glucose Profile) chart on the home dashboard -- percentile bands by hour-of-day across a configurable window. AGP is the standardized clinical chart your endocrinologist most consistently knows how to read, and you can see one in GlycemicGPT today.

Daily briefs are a different artifact: AI-written prose summaries of recent activity, focused on what's interesting or unusual rather than the structured clinical picture AGP gives. They have different jobs:

  • AGP (on the dashboard) -- deterministic percentile bands, fixed statistics, the clinical-format view
  • Briefs -- prose for self-reflection, with the AI chat behind them for follow-up questions

You'll often want both: AGP for "how am I doing structurally over time," briefs for "what stood out yesterday and is there a thread to pull on."

Note: a printable AGP-style report (the kind you'd hand to your endocrinologist) is a roadmap item -- the dashboard AGP is the visualization, but the standardized exported / printable report format is still being built.

Brief quality depends on data quality

A brief is only as useful as the data the AI has to work with:

  • Sparse glucose data -- if your CGM was in warmup or had connection gaps, the brief will be thinner
  • No insulin data -- if you're using GlycemicGPT for monitoring only and don't have a pump connected, briefs focus on glucose patterns and don't mention insulin
  • Recent setup -- briefs that try to identify patterns need at least a few days of historical data

The platform does what it can with what it has -- briefs explicitly note when data was missing.

Privacy

  • Briefs are stored on your platform's database alongside your other data
  • The AI provider you configured generates briefs by reading your data and writing the summary -- the data goes to the provider during generation (see Privacy)
  • Briefs are not shared with anyone unless you've linked a caregiver who has brief-read permission (see Caregivers)

Disabling briefs

If you'd rather not have automated briefs, go to Settings → Briefs and toggle them off. You can still generate briefs manually when you want them.

Why is my brief wrong / weird / missing things?

Same root causes as any AI quality issue:

  • Sparse / weird data -- if your CGM had a 6-hour gap, the brief might say "no overnight data available"
  • Smaller models hallucinate more -- using a premium model (Claude Opus, GPT-4-class) gives sharper briefs
  • The AI just gets it wrong sometimes -- AI is not infallible. Use briefs as starting points for conversations with your endo, not gospel.

A hallucination-feedback mechanism (so you can flag a bad brief and have it regenerated from a fresh session) is on the roadmap -- see ROADMAP.md §Phase 1 AI Engine 2.0.

On this page