Nutrition (Beta)

About the Beta tag

The short version

Nutrition is live for your clients to use. The Beta tag is on because we are still expanding the Australian chain and supplement coverage, and the food matcher is still learning from corrections. Everything works. The tag will come off when we have covered enough of the AU restaurant and meal-prep scene that an average Australian client can log a week of eating without hitting a "not in the list yet" message.

Why we built our own food list

Most nutrition apps your clients have tried (MyFitnessPal, Cronometer, MacroFactor) pull their numbers from the US Department of Agriculture database, the National Cancer Center database, or whatever the wider community has crowdsourced. Useful for an American client. Less useful for one who shops at Coles, drinks a flat white, and orders Lite n' Easy.

We started from AUSNUT, the official FSANZ Australian food composition database, and built outward. Right now the list pulls from four sources:

- AUSNUT (FSANZ) for verified generic foods. About 1,600 entries. Same data dietitians use. - The Open Food Facts AU subset for barcoded brands. About 49,000 entries tagged for Australia (Coles, Woolworths, Aldi private labels and most major Australian grocery brands). - A hand-curated AU chain list. 160 entries across GYG, McDonald's, KFC, Hungry Jack's, Subway, My Muscle Chef, Lite n' Easy, Bulk Nutrients, MyProtein, and True Protein. - 16 canonical staples (bread, rice, eggs, milk, chicken and a few others) with verified macros so common words land on a clean entry.

About 51,000 foods on day one. We add more every week.

What clients see

Your clients get a Nutrition card on their portal with five ways to log food:

- clipboard-list-linear Recents. One tap to repeat a meal they logged before. - magnifer-linear Search. Type a food name and pick from the list. - barcode-scan Barcode. Scan packaged food on mobile (native scanner on iOS and Android). - camera-bold Photo (label or meal). AI reads a nutrition panel, or estimates portions from a meal photo. - chat-square-2-bold Tell Repley. Plain English, like "two eggs on toast and a flat white".

Each logged item shows on the Today card with a pen-linear pencil to edit the portion and a trash-bin-trash-linear bin to remove it. Macros tick into the rings as your client logs. See How clients log food for the full walkthrough of each method.

What you see as a trainer

- Nutrition tiles on your client dashboard. Total active plans, average adherence, protein hit rate. - Generate plan. Ask Repley or use the Plan Generator. Review the macros and assign in about a minute. - Food Database tab in your Library. Browse the food list, see what your clients have been logging, and spot the foods they searched for that we did not cover yet (great signal for what to ask us to add). - Per-client nutrition view. Open any client and review their logs alongside their workouts. See the Trainer guide for the full workflow.

How the food matcher learns

When a client types "bread" we want them on plain sliced bread, not a brioche. When they type "chicken" we want raw chicken breast, not a Hungry Jack's burger. The food matcher uses a commonality score so common staples win against specialty entries.

Every time a client edits or removes a log that the AI proposed, we record what they typed, what the system picked, and what they changed it to. We use that to:

1. Boost frequently-corrected foods so they win the next match. 2. Add missing entries for queries that keep getting fixed. 3. Update synonyms so "snags" matches sausages, "spud" matches potato, and so on.

This is the long game. The more clients log, the sharper the food matcher gets.

What to expect

- Search and barcode are sharp. AUSNUT-backed entries are verified government data, and our barcode coverage is strong on Coles, Woolworths, Aldi, Cadbury, Arnotts and the rest. - Meal photos work well on a single plate where the food is clearly visible. They are less sharp on mixed dishes (a curry where everything is sauced together). The AI proposes, your client confirms before saving. - Specialty items might land on a different but equally valid entry. "Naan" and "lavash" both pick bread; the source might be Open Food Facts rather than AUSNUT. Same nutrition, different row. - Branded items work better when the brand is named. "Bulk Nutrients whey" lands cleanly. Generic "whey protein" might pick a competing brand. - Restaurant meals are most accurate when the chain is in the message. "McMuffin" or "Big Mac" hit the right McDonald's entry.

What if a client's food is not in the list yet

They can still log it three ways:

- flash-linear Quick add. Just calories and protein. Takes about 5 seconds. - camera-bold Photo label. Snap the nutrition panel and the AI reads it. Works on almost anything with a back-of-pack panel. - chat-square-2-bold Tell Repley. Describe it in plain English and the AI estimates the macros.

The Food Database tab shows you the foods clients are searching for that we do not cover yet. We use that list to decide what to add next.

Privacy

The AI sees a redacted version of what your client typed. No names, emails, or phone numbers go to OpenAI. We track AI cost per business with a monthly cap so a runaway logger cannot rack up a bill.

When the Beta tag comes off

We will drop it when:

1. The hand-curated chain coverage gets to about 30 of the major AU restaurant, cafe, and meal-prep brands (currently 10). 2. Real-user correction data shows the food matcher hitting the right entry on the first try more than 90% of the time. 3. We have at least four weeks of those numbers in front of us.

No date promise. The shorter the path from "your client opens the app" to "the meal they ate is in the list, accurate, ready to log", the closer we are.