Concept | Product design and design system

Verso

Verso

Verso

A light-mode editorial desk that turns AI conversations into published posts, keeps a citation on every draft, and calibrates voice per platform.

A light-mode editorial desk that turns AI conversations into published posts, keeps a citation on every draft, and calibrates voice per platform.

A light-mode editorial desk that turns AI conversations into published posts, keeps a citation on every draft, and calibrates voice per platform.

Role

Concept | Product design | Design system

Timeline

2026 | Self-directed concept

Stack

Fraunces + Inter + JetBrains Mono | warm paper, single indigo accent

Platform

Web

Context

A growing number of founders and creators now do their best thinking inside AI chats, then have to reconstruct posts from those threads by hand. Tools in this space prove the capture-to-calendar loop works, but they lean on comforting abstractions: an idea gets filed with no memory of where it came from, every platform receives the same generic draft, and the calendar reassures you it is balanced without saying what that means.

Problem

Three honest gaps sit underneath the polish. Captured ideas lose their provenance, so a week later you cannot tell which conversation a claim came from or whether it was your point or the model’s. One voice is stretched across LinkedIn, X, and Instagram, which reads as nobody’s voice. And a soft balance metaphor hides the only question a creator actually has: am I about to run dry, and where.

Approach

Verso is a concept for the same space with a different posture: an editorial desk on warm paper rather than a clinical dashboard. Every draft carries a persistent citation chip that links back to the exact message range that seeded it. Voice is calibrated per platform against your own published history and shown as a live match readout while you write. The calendar is replaced by a runway that states queue health in literal numbers, including the uncomfortable ones. Delivered as three screens plus a semantic token system, clearly labeled concept work.

Outcome

A self-directed concept that reframes the category around provenance and honesty, and gives the portfolio deliberate range beyond dark UI: a warm light-mode system with a single reserved accent, an editorial serif for the writing and a grotesque for the tool, and tabular numerals wherever a number carries a decision.

Decisions | Reasoning

Warm paper light-mode system (page #F7F5EF, cards #FDFCF9) rather than pure clinical white.

Warm paper light-mode system (page #F7F5EF, cards #FDFCF9) rather than pure clinical white.

The reference canvas is #ffffff, correct for a utility app but generic and hard on the eyes across long editing sessions. A creator spends real time reading and rewriting prose here, so the surface should feel like a manuscript on a desk. Warm off-white also gives the portfolio a distinct light-mode identity and lets a single cool accent read as unmistakably intentional against it.

The reference canvas is #ffffff, correct for a utility app but generic and hard on the eyes across long editing sessions. A creator spends real time reading and rewriting prose here, so the surface should feel like a manuscript on a desk. Warm off-white also gives the portfolio a distinct light-mode identity and lets a single cool accent read as unmistakably intentional against it.

One reserved accent, proof indigo #3A2FD4, assigned only to provenance links and the single primary action per view.

One reserved accent, proof indigo #3A2FD4, assigned only to provenance links and the single primary action per view.

The reference reserves one accent (orange) for scheduling. I keep the discipline but re-role the accent to the product's core promise: since provenance links are literally hyperlinks back to a source, indigo reads as traceable and honest, the web's own convention for this leads somewhere real. Everything else stays warm grayscale, so the accent always means citation or commit, never decoration.

The reference reserves one accent (orange) for scheduling. I keep the discipline but re-role the accent to the product's core promise: since provenance links are literally hyperlinks back to a source, indigo reads as traceable and honest, the web's own convention for this leads somewhere real. Everything else stays warm grayscale, so the accent always means citation or commit, never decoration.

Two-tier type: Fraunces for user content and headings, Inter for tool chrome, JetBrains Mono for citation ids and timestamps.

Two-tier type: Fraunces for user content and headings, Inter for tool chrome, JetBrains Mono for citation ids and timestamps.

The reference is all-sans, which flattens the writing into form fields. Splitting the writing (serif) from the tool (sans) makes drafts feel like finished prose and reinforces the voice-calibration idea, the words on screen already look like your voice. Mono is reserved strictly for machine-truth metadata (source ranges, times), so the eye learns that mono equals provenance receipt.

The reference is all-sans, which flattens the writing into form fields. Splitting the writing (serif) from the tool (sans) makes drafts feel like finished prose and reinforces the voice-calibration idea, the words on screen already look like your voice. Mono is reserved strictly for machine-truth metadata (source ranges, times), so the eye learns that mono equals provenance receipt.

Provenance is a persistent citation chip visible at rest on every capture and draft, never a hidden menu.

Provenance is a persistent citation chip visible at rest on every capture and draft, never a hidden menu.

If keeping the receipt is the headline improvement over the reference, it cannot be buried. A draft is never shown without its source, expandable in one click into the original two-message exchange. This turns a compliance-sounding feature into a trust cue you feel on every card, and quietly enables recall (why did I write this) and dedupe against already-published points.

If keeping the receipt is the headline improvement over the reference, it cannot be buried. A draft is never shown without its source, expandable in one click into the original two-message exchange. This turns a compliance-sounding feature into a trust cue you feel on every card, and quietly enables recall (why did I write this) and dedupe against already-published points.

Replace the balance metaphor with literal queue-health numbers, including bad news, using tabular numerals and no green all-clear state.

Replace the balance metaphor with literal queue-health numbers, including bad news, using tabular numerals and no green all-clear state.

Balancing the week is comfort, not information. Verso states runway in days, names the platforms that go quiet and when, and counts stale drafts, so the creator can act. Tabular numerals keep those figures aligned and jitter-free across the calendar. Refusing a reassuring green check is the honest choice: the system commits to a legible truth even when the truth is you run dry Thursday.

Balancing the week is comfort, not information. Verso states runway in days, names the platforms that go quiet and when, and counts stale drafts, so the creator can act. Tabular numerals keep those figures aligned and jitter-free across the calendar. Refusing a reassuring green check is the honest choice: the system commits to a legible truth even when the truth is you run dry Thursday.

Voice is calibrated per platform against the user's own published history and surfaced as a live match score with plain-language reasons, not a rewrite button.

Voice is calibrated per platform against the user's own published history and surfaced as a live match score with plain-language reasons, not a rewrite button.

The reference knows the best time to post per platform but not how each platform should sound. Verso learns from posts you already published, then shows an 82 out of 100 match with legible drivers (your LinkedIn posts open with a claim 78% of the time; this one opens with a question). Keeping the human in the loop, coaching rather than auto-rewriting, respects that the voice belongs to the user, which is the whole point of the score.

The reference knows the best time to post per platform but not how each platform should sound. Verso learns from posts you already published, then shows an 82 out of 100 match with legible drivers (your LinkedIn posts open with a claim 78% of the time; this one opens with a question). Keeping the human in the loop, coaching rather than auto-rewriting, respects that the voice belongs to the user, which is the whole point of the score.

Results

A self-directed concept, not a shipped product: reasoning and system choices are the deliverable.