Originally published on Medium ↗
How AI Visibility Studio turned a dog care idea into a working product, what testing taught us, and why we’re learning before introducing a paid tier.
You’ve just come back from a walk. Your dog ate less than usual this morning. There’s a medication to remember tonight, and somewhere on your phone is a photo you meant to show the vet.
Dog care involves plenty of small details. Remembering them all is another job.
We’ve been building My Dog Pal at AI Visibility Studio to help owners keep those details together. Meals, walks, routines, observations, records, and memories, organised around the dog they belong to.
We’ve developed it with Codex, OpenAI’s coding agent. We lead product direction, design review, acceptance testing, and release decisions. Codex carries out substantial implementation work and helps us investigate and test changes.
That collaboration has taken the app through successive builds and into private beta. Along the way, we’ve simplified everyday tasks, corrected misleading calculations, protected existing records, and held back features that needed further testing.
My Dog Pal is now accepting private-beta requests from Android and iPhone users. Before introducing a paid tier, we want to understand how it fits into people’s lives.
An app built around everyday care
The useful moments are often ordinary.
An owner records a meal, completes a scheduled care task, saves a walk, or notes a change in appetite. Later, they can return to that history instead of trying to reconstruct the week from memory.
My Dog Pal brings those activities together through daily check-ins, care logging, schedules, reminders, and recorded history. Development has also included weight records, food-product scanning, reusable walk routes, and records for veterinary appointments.
There’s room for the enjoyable side of living with a dog, too. Photos, Stories, memories, and Teddy’s stickers give the app a personality beyond its forms and charts.
Teddy is the inspiration behind My Dog Pal and a familiar face throughout its artwork. We wanted the product to feel welcoming while remaining clear about what its records can tell you.
Dog profiles and core care records stay on the device by default, and the current core experience does not require an account. The app helps owners remember and review what happened. It doesn’t diagnose conditions or replace veterinary advice.
How we worked with Codex
The process usually started with a practical request.
Make daily check-ins easier to find. Simplify the home screen. Help an owner return to a saved walk. Make a chart easier to understand.
Codex inspected the existing code and worked through the implementation. We reviewed the result and tried it on a real iPhone.
That often revealed something a screenshot couldn’t. A screen could look tidy but feel crowded in use. An action could work but take too many steps. A feature could exist without being easy to discover.
We worked through those details repeatedly.
Our development process included focused changes, automated checks, simulator testing, and installation on a physical device. We kept a separate development app so candidates could be reviewed alongside the TestFlight version.
Approved releases were tied to a specific code revision, giving us a record of what had been tested and what went into each build.
Codex also maintained written handoffs: checks completed, known limitations, and the next work required. That helped us distinguish between a successful build, a reviewed experience, and a candidate ready for wider testing.
How the product grew
The release history shows the app developing in stages.
By late July 2026, the work included food and weight records, barcode scanning, catalogue lookup, and improvements to the home screen. Build 5 went to TestFlight on 29 July.
Early August brought further work on reminders, walks, onboarding, rewards, and layouts for larger text. Later builds developed the schedule and diary, reusable walk routes, and the ability to follow a previous route.
By the September 4 delivery of build 14, we had refined daily care again, corrected chart behaviour, and developed Teddy’s Stories, stickers, and PNG exports.
That build’s handoff recorded 808 passing app tests, alongside other code checks and successful iOS builds. Android bundle exports had also passed, although those checks are separate from testing interactions on an Android device.
The milestones show progress. The decisions behind them explain more about how we work.
Making daily check-ins easier
Daily check-ins needed to be straightforward enough to repeat.
We made common observations available directly, moved additional categories behind an optional disclosure, and removed a required second page. We also made the check-in more prominent on the home screen, with the day’s schedule close by.
The aim was to let someone record what they had noticed without making them complete an unnecessarily long form.
That required care with the defaults. If an owner recorded appetite and left the other categories blank, the unanswered categories stayed unobserved. They weren’t automatically marked “normal.”
We also worked on the save process. If saving failed, the draft needed to remain available for another attempt. The app shouldn’t announce success before the information had been saved.
These decisions connect product design with data quality. A quicker flow still needs to preserve what the owner actually meant.
The chart that made us look more closely
One of the clearest examples of the value of testing came from a chart audit.
A water-intake comparison could mix millilitres with the number of entries someone had logged. Treat records had a similar issue, where known quantities could be compared with entries whose amounts were unknown.
The result could be a plausible-looking percentage based on incompatible measurements.
We corrected the calculations so those measurements weren’t compared. When the records couldn’t support a percentage, the app stopped displaying one.
The review also led to changes in how weight readings were ordered and how chart points and event markers aligned with dates.
We then checked the behaviour through the app. In one simulator test, we logged 250 ml of water for the previous day, relaunched, and confirmed that it remained on the correct date. Today stayed empty.
That distinction matters. An empty day means nothing was recorded. It doesn’t mean the dog drank nothing.
The same principle applies across My Dog Pal: make recorded information useful without suggesting that the app knows more than the owner entered.
Keeping existing information safe through changes
As features developed, we also had to preserve what was already saved.
One correction addressed photos whose stored paths referred to an older iOS app location after an update. We changed how the app located those files, reviewed the candidate on a real phone, and included the fix in build 9.
The test instructions were specific: update without deleting the app, then confirm that photos remained after launch and relaunch.
Pet profiles, care history, schedules, and household information required the same attention. Each new version had to account for the records people had already created.
Deciding what was ready
Family sharing has been a separate development and testing effort.
We implemented functionality and confirmed specific compatibility fixes on a real device, including preservation of an existing household board. That did not establish the complete experience across two phones.
Invitations, sign-in, offline changes, conflicting edits, and removal of access needed their own checks.
Family therefore remains unavailable in the current beta while that work continues.
For us, this is part of product delivery. We need to be able to explain what has been verified and where the evidence is still incomplete before making a feature available more widely.
Why we’re beta testing before charging
We want to understand which parts of My Dog Pal become useful habits.
Can a new owner find where to start? Is recording a meal straightforward? Does a check-in ask too much? Can someone retrieve an observation they entered last week?
Which features do people return to? Where do they hesitate? What feels unnecessary?
Alongside those questions, beta testing gives us more opportunities to examine reminders, interrupted walks, larger text, updates, and repeated use across devices.
Automated checks help us test specific behaviours. Reviewing builds on a real phone adds another layer. Other people using the app bring different routines, expectations, and ways of doing things.
We’re keeping the beta free while we learn.
That doesn’t establish a permanent pricing policy, and free signups won’t prove willingness to pay. We want the eventual commercial offer to reflect something people value, supported by a product we can operate and improve responsibly.
What this shows about AI Visibility Studio
My Dog Pal brings several parts of our work together: shaping requirements, designing the experience, directing AI-assisted implementation, reviewing technical behaviour, and managing release decisions.
Working with Codex gave us a way to move repeatedly between an idea, a working implementation, and a tested revision. Our responsibility was to give that work direction and evaluate the result.
The evidence is in the product decisions: simpler check-ins, more careful charts, preserved records, and clear limits around unfinished validation.
At AI Visibility Studio, our wider work covers website strategy, design, development, and search and AI visibility. My Dog Pal adds a practical product-development case study to that work, showing how we use AI and how we check what it produces.
There is more to learn. The next stage is to listen to people using the app, improve the parts that get in their way, and understand what earns a place in their daily routine.
If you’d like to help shape that next stage, request private beta access to My Dog Pal.
The current beta is free, and Android and iPhone testers are welcome. Invitations are limited and sent in small groups. If invited, you’ll receive setup instructions for your platform; iPhone testers use TestFlight.