Drink Buddy
One is the memory. One is the evidence.
In the database the two images look symmetrical: same table, same upload path, both required. To the person taking them they are nothing alike, and almost every trade-off in the capture flow falls out of admitting that rather than designing around it.
One artefact, one measurement
The schema stores two images against one check-in and gives them equal billing. Everything downstream refuses to treat them equally, and that refusal is the design.
The environment image is the social object. It is what a friend reacts to, what you scroll back to, and the thing that still has value long after you have forgotten what was in the glass. Visible compression on it is a product defect.
The drink image is a measurement. Its only job is to be legible enough to confirm what you were holding. Nobody has ever missed a drink close-up, and the product budgets it accordingly.
Where the asymmetry shows up
Same table, same upload path, opposite treatment on every axis that costs something.
| Axis | Environment image | Drink image |
|---|---|---|
| What it is to you | The memory | Proof of what was in the glass |
| Emotional weight | High — this is the photo | Near zero |
| Quality budget | The larger share; visible compression is a defect | Only enough to be legible |
| Capture style | Self-timing countdown, wider frame, worth a retake | Deliberate shutter press, as close to instant as possible |
| Retention posture | The last thing that should ever be deleted | A legitimate candidate for downscaling |
| If recognition returns | Never sent — enforced by a single-image contract | The only image that would be sent |
Why the drink comes first
The order is not arbitrary, and it is the reason the second shot can be a countdown rather than a pose.
The drink, on the rear camera by default.
A deliberate shutter press, framed at the glass, cup, can or bottle. You can flip the camera if that suits the table better.
The moment, on a countdown.
A wider frame with softer quality gates. The countdown exists because you are usually in this one, and a shutter button you have to reach for produces a worse photograph than a timer you can stand behind.
Review, where correction happens.
Both images, the resolved drink, the caption and the location control in one place. If you find yourself waiting here, something upstream failed — the uploads were supposed to finish underneath this screen.
Publish.
The slot is spent at this moment and the check-in becomes visible to your accepted friends. Nothing before this point is visible to anyone.
The consequence nobody designs for on purpose
Rear drink then front moment is the common path, and it means a face in most photographs. That is a heavier privacy posture than a product of this size usually takes on, and it is acknowledged rather than discovered later.
A photograph of your friends is published to a circle one of them did not choose.
The audience is only your accepted friends — there is no public tier and no share link. Anyone in the frame who is also in the app can see and react; nobody outside it can reach the image at all.
You want the moment photograph without being in it.
Flip the camera before the second capture, or point it at the table. Neither the flow nor any gate requires a face.
You assume a photograph of a bar identifies the bar.
Not from the file. Location metadata is removed on the device before upload, so any place information attached to a check-in comes from the separate, opt-in location control rather than from the image itself.
What this page leaves to other pages
Named so that silence is not read as an answer.
- How the drink itself gets identified, which is a manual selection and not an inference from either photograph.
- Who can see a published check-in, and what a non-friend can reach.
- Whether a coordinate is attached at all, and who is allowed to see it.
- How long the images are kept, which is covered alongside account deletion.
- Whether any automatic moderation looks at the environment image. It does not, and the reasoning belongs with the visibility model.
Questions
Can I check in with only one photograph?
No. One environment image and one drink image are both required to publish, and there is no reduced path that skips either.
Is my photograph sent somewhere to work out what the drink is?
No. Automatic recognition was cancelled in full — no provider is configured, no credential exists and no call is made. The drink is chosen by hand from a catalogue.
Does the photograph carry the location it was taken at?
Not by the time it leaves your phone. Location metadata is stripped on the device before upload, and on Android the encoder produces none in the first place. Any coordinate on a check-in comes from the separate location control.
Can I replace a photograph after publishing?
No. The caption and the drink identity remain editable for twenty-four hours, but the images are fixed, because they are the evidence the record rests on.