07 / PRODUCT / FIELD GUIDE / LOCAL HISTORY

Lakeglass

What did the lake give you?

Role
Concept / Product Design / Research / Development
Built
2026
Core idea
Read a find from observable evidence, then keep inference separate from known provenance.
FIELD READINGLM-042

LIKELY READING

Pale aqua

Likely utility-bottle glass, softened by prolonged Lake Michigan tumbling.

Physical clue
Thin / curved wall
Occurrence
Uncommon
Context
Chicago shoreline
Provenance
Recorded separately

Identification is based on observable clues, not a photo match.

Why I made it

Beach glass invites a very specific kind of curiosity: you can hold a fragment in your hand and know that it had a life before the lake, but the evidence is incomplete. I wanted a tool that helped read those clues without turning uncertainty into a false answer. Lakeglass treats a find more like field evidence than an image-search problem.

The identification model

The flow starts with what can actually be observed: shoreline, color, thickness, form, opacity, surface condition, markings, and any diagnostic feature that survived tumbling. Those clues are combined into a likely object type, historical context, and occurrence reading. Color matters, but it never gets to make the decision by itself.

Provenance, not prediction

The app keeps known provenance — where and when something was found, plus the collector's note — separate from inferred history. A fragment can have a strong physical reading while its original maker, date, or route into Lake Michigan remains unresolved. That distinction is part of the product rather than fine print hidden after the answer.

Research-backed on purpose

Regional and beach-glass research informs the field reading, and the result explains why a hypothesis was made. Lakeglass surfaces evidence, occurrence, local shoreline context, and the research basis so the user can understand the reasoning instead of receiving a single black-box label.

No camera identification

There is deliberately no photo upload or camera-identification step. A photograph can make an uncertain fragment look more knowable than it is, while useful clues like thickness, curvature, texture, and a partial mark are easier to assess in hand. The interaction asks the user to look more closely rather than outsource the observation.

Local-first and offline

Lakeglass is an installable PWA with a local IndexedDB specimen archive, import/export, and service-worker support for offline use. There are no accounts, no server-side collection database, and no photo storage. The field guide can go to the shoreline, and the collection stays with the person doing the finding.

Design direction

The interface borrows from field guides and collection labels: soft paper tones, Lake Michigan blue-greens, serif headings, monospaced evidence labels, and translucent specimen shapes. For the portfolio, I kept those signals inside the same editorial/brutalist system as the rest of the work so Lakeglass feels distinct without becoming a separate visual world.