Remi, the coach

Remi is the coach inside Peptide AI. Ask it anything about your protocol. It answers from your own doses, labs and wearable data, and it cites the research behind each compound. It never produces a number it did not read from your log, and it never recommends a dose. That line is deliberate, and it is where most AI health apps fail.

How it works, step by step

1. Ask in plain language. "Has my resting heart rate changed since the last dose step?" or "What does the research on this compound actually show?"
2. Remi reads your log. Doses from the dose log, results from bloodwork and data from the wearable integrations.
3. It answers with sources. Numbers about you come from your log. Claims about a compound come with the research behind them.
4. It stops at the dose. Questions about what dose to take go to you and your prescriber.

Why the two rules exist

Never a number it did not read. General-purpose language models are fluent, and fluency is not accuracy. A 2023 study in Cureus asked ChatGPT-3.5 to write 30 short medical papers with references and then checked every reference.

That was an older model writing without access to any data, and newer systems do better. The lesson still applies. A model asked about your HbA1c with nothing to read will produce a plausible number. Remi is built so that any figure about you comes from your own log, and if the log does not have it, the answer says so.

Never a dose. Doses depend on things an app cannot see: your full medical history, other medicines, the exact product, and an examination. The FDA's 2024 alert on compounded semaglutide dosing errors showed how badly dose arithmetic can go even between patients and providers. Remi can explain a titration schedule from the label, show you your own history, and point to the research. It will not choose a number for you.

The kind of question it is for

What happens to your data

The privacy policy is specific. Remi is powered by Anthropic's Claude API. When you use it, your messages, any body metrics or protocol and dosing details relevant to the question, and the conversation history may be sent to Anthropic for processing. Your email address, account credentials and payment information are not. The policy states that the API is used under zero-retention terms that prevent your data being used to train Anthropic's models. You can choose not to use Remi, in which case none of that data is sent. The privacy and data page covers the rest.

Who it is for

• People with weeks of logs who want to ask questions of them rather than scroll through charts.
• Anyone who wants the research on a compound summarised with the citation attached.
• People preparing for an appointment who want their own history pulled together first.
• Anyone who has read a confident claim about a compound online and wants to check it against the research and the evidence grade.

How it connects to the rest of the app

Remi reads everything the rest of the app records: the dose log and protocol, labs, wearables, nutrition, supplements, training and body scans. It cites the research from the compound library, including the evidence grade. The more complete the log, the more useful the answers.

Ask Remi one question you would otherwise have asked a forum, and check whether the answer cites its source and uses your own numbers. That is the test it is built to pass. The app is on the download page.