Warpath

• 

  Warpath iPhone screen captured without a device frame

• 

  Warpath app icon, 1024 x 1024

Inspiration

Deadlift progress is difficult to review when the only record is a feeling from the set. Warpath started from a simpler idea: make the camera repeatable, review the pull afterwards, and turn that evidence into one useful thing to try next time.

What it does

Warpath is a record-then-analyze deadlift app for iPhone. It guides a repeatable floor-mounted side view, records the set, saves the user-confirmed load and reps, and analyzes the video while the lifter rests. The review combines replay, body and estimated bar-path overlays, per-rep floor-to-lockout timing, and one evidence-backed next-session steer.

The product is intentionally retrospective. It does not provide feedback during a set, diagnose injury risk, or infer stance width or toe angle from a side view. The user confirms the load before it is saved, and uncertain observations are marked or omitted.

How we built it

We built the prototype in Swift and SwiftUI. AVFoundation handles capture, Vision provides body landmarks, and deterministic analysis turns complete frames into bar-motion and timing observations. Local persistence keeps the video, raw observations, and interpretations available for review. RevenueCat is wired to the Warpath Pro entitlement boundary.

Challenges we ran into

The hardest parts were making floor-mounted capture repeatable, defining a quality gate for framing and complete frames, and keeping the raw recording available when analysis logic changes. We also had to be explicit about what a side view can and cannot support, and to abstain when an observation is incomplete or low confidence.

Accomplishments that we're proud of

We focused the product on a complete post-set loop instead of live coaching: capture, rest, review, and one next-session experiment. The same domain model supports the non-LiDAR flow, while optional depth remains a capability rather than a requirement.

What we learned

Good product guidance begins with evidence boundaries. Provenance, confidence, capture quality, and analysis version belong with every derived observation. A user-confirmed load is more trustworthy than an automatic guess, and a clear limitation is better than a confident but unsupported coaching claim.

What's next for Warpath

Next we will validate the capture and analysis thresholds on physical devices with a controlled dataset, then prepare the first public iPhone build. We will keep the core promise narrow: record every deadlift, see what changed, and know what to try next.

Built With

• avfoundation
• ios
• revenuecat
• swift
• swiftui
• vision