What reached users, operators or an upstream project — the outcome, not just the code.
- Grow Journal product website and product identity; iPhone application remains in development and is not represented as released
TECHNICAL CASE STUDY / NATIVE IOS · DATA · AI
Grow Journal
A native journal where AI answers are grounded in the grower's own records.
An iPhone and iPad application built around plant histories, quick logging, photos, environment records, analytics and an AI coach that can cite the underlying journal evidence.
Grow Journal — visual identity and journal experience
PRODUCT SCREEN GALLERY / IPHONE
Inside the product.
Real iPhone application captures from Grow Journal, showing the Today dashboard, plant records, grow replay timeline and analytics. These are product screens from the project, not concept mock-ups.
DELIVERY EVIDENCE / ENGINEERING MATURITY
What reached the real world.
Delivered outcomes come first. The remaining categories show what is implemented, what has test evidence, what is deployed and what is still staged or planned.
Capabilities evidenced in source code or project documentation.
- SwiftUI iPhone/iPad project with structured plant history and PlantMemory evidence model
- Read-only AI record tools and auditable check/export components are present in the project
- Separate watchOS companion target and offline WatchConnectivity queue are implemented in source
Automated test coverage or documented test evidence; not a blanket passing-build guarantee.
- No current passing automated test run is claimed on this page
Live product surfaces or historically operated deployments, with scope stated.
- Grow Journal product website is live; this does not imply the iPhone app is publicly released
Known release boundaries, future work or areas not claimed as complete.
- Apple Watch companion is staged for version 1.1 and excluded from the iPhone 1.0 bundle
- iPhone application remains in development
THE PROBLEM
Why this needed
to be engineered.
A useful journal needs to preserve the history behind each plant, not just collect disconnected notes. AI guidance becomes risky when it cannot show which records it used or when users cannot inspect and correct its prior checks.
Build a native Apple-platform experience with structured plant and journal models, local persistence, photo and environment histories, and an AI coach whose tools read from the user's records. Keep the AI's evidence, model and prompt version in an auditable history.
SYSTEM ARCHITECTURE
How the pieces fit together.
01
SwiftUI iPhone and iPad application with dedicated plant, journal, analytics and tools views
02
Core Data persistence and migration components, with Cloud record/sync components in the Xcode project
03
PlantMemory evidence layer combining dated readings, notes, stages, irrigation, treatments and photos
04
Read-only AI tools for searching records and plant history; AI check history preserves evidence, model and prompt version with optional answer evaluation
05
A separate watchOS 10+ companion target with WatchConnectivity bridge to the iPhone app
06
Watch companion snapshot carries today’s tasks and each plant’s day and stage; offline actions queue locally and sync when the phone is nearby
TECHNOLOGY CHOICES
Tools chosen for the job.
Native Apple UX
SwiftUI is used across the core app and supporting views. The project also contains a separate watchOS 10+ companion target and App Intents.
Watch companion with phone-owned data
WatchConnectivity sends the day’s tasks and plant stage to the wrist. Users can complete tasks, log watering and enter pH with the Digital Crown; the watch keeps an offline copy and queues changes for the iPhone to save.
Release scope kept explicit
The latest project change keeps the watch target and source code in the project but removes it from the iPhone 1.0 app bundle. The companion is staged for a 1.1 release rather than being represented as shipped in 1.0.
Structured plant memory
PlantMemory assembles dated records into a citable timeline so the AI coach can explain what changed and point back to source entries.
Read-only AI tools
The coach's record-search and history tools are read-only, with each tool stating its source rather than silently mutating user data.
Integrity and portability
Backup, report and export paths use SHA-256 manifests; the JSON export is documented as an open format.
ENGINEERING CHALLENGES
Problems solved along the way.
Evidence-grounded answers
The AI needs access to relevant historical records while making it possible for a grower to inspect the evidence behind an answer.
Auditable AI behaviour
Problem checks preserve evidence, model and prompt version, and allow users to mark answers right, partly right or wrong.
Longitudinal data
Plant histories combine photos, stage changes, readings and journal notes across the full lifecycle, so the UI and data model must work over months rather than a single session.
Data ownership
Open-format export and checksummed backups support portability and restore verification.
Companion-device synchronisation
Watch actions must remain useful offline without creating a second source of truth. The phone remains the persistence authority and receives queued watch actions through WatchConnectivity.
SOURCE & VERIFICATION
Inspect the work.
Private source repository (access may be restricted)↗Plant memory model↗AI coach audit trail and export commit↗WatchConnectivity bridge and offline queue↗Apple Watch companion implementation↗1.0 release-scope adjustment for watch target↗Product website↗
The repository is private, so public visitors may not be able to inspect its source. The Apple Watch companion is implemented in the project but deliberately not embedded in the iPhone 1.0 release; it is staged for 1.1. Screenshots should be verified against the current iPhone/iPad build before publication; this page does not substitute generated mockups for real app screens.
NEXT / START A PROJECT
Have a similar challenge?
Tell me what you are trying to build. We can scope the right next step.