Product UI, display, editorial, monospace, licensing, and delivery tradeoffs. All research.
The Thorbis call
Research the range. Ship the disciplined pair. Interowns product UI and Geist Mono owns code-like detail. Every other family remains comparison evidence. This page does not authorize a third face.
01 · InterfaceInter
Quiet at density. Clear at field size.
02 · TechnicalGeist Mono
IDs, money, shortcuts, and code.
19families studied
12open families
4bounded roles
2shipped faces
07 families
Product UI families
Interface type must stay legible at compact sizes, cover real product languages, preserve stable metrics, and disappear behind the work.
Reference
Product UI
Inter
Shipped
Purpose-built interface type with a large x-height, broad language coverage, variable weight, and optical sizing.
Open·Variable weight + optical size · current UI family
Thorbis call
Keep for controls, labels, body copy, and responsive product UI.
The canonical Type skill owns Thorbis law. These sourced findings cover licensing, payload, fallback metrics, scripts, zoom, and user-controlled spacing; they remain research until that owner adopts them.
01
Role precedes family choice. UI text, long-form reading, display, and monospaced data have different constraints. Comparisons become meaningful when each family is evaluated inside one named role.
02
Web rights are separate from desktop rights. Web adoption requires confirmed embedding, traffic or seat limits, applications, redistribution, derivatives, and fallbacks. System availability does not prove web rights.
03
WOFF2 and subsetting reduce payload. Subsetting to real language coverage and removing unused weights reduces transfer. A variable font can replace several styles, while one static cut may still be smaller for a one-style role.
04
Preload helps only critical above-fold text. Preconnect can help an unavoidable font origin, and self-hosting can help when the license and cache strategy allow it. Preloading below-the-fold fonts adds work without protecting first render.
05
Font swap and fallback metrics are one problem. font-display choices work with fallback matching. size-adjust plus ascent, descent, and line-gap overrides reduce reflow when the web font arrives.
06
User reading settings expose fragile layouts. Robust layouts survive browser zoom, user styles, narrow widths, 1.5× line height, 2× paragraph spacing, .12em letter spacing, and .16em word spacing without clipping.
07
Hostile strings reveal false confidence. Pairs such as 0/O, 1/l/I, and rn/m—plus punctuation, currency, fractions, long names, diacritics, non-Latin scripts, and tabular numbers—expose weaknesses at real interface sizes.
How to read the evidence
01
Interface
Inter remains the approved UI face; the other sans families are comparison evidence.
02
Technical
Geist Mono remains the approved face for code-like detail, IDs, money, and shortcuts.
03
Display
Expressive sans and serif specimens document contrast; they do not approve a third face.
04
Accessibility
Atkinson Hyperlegible Next records useful legibility evidence without changing the stack.
Delivery and accessibility sources
Each card links to its foundry, maintainer, or distributor. These standards and browser references support the delivery findings.
Reviewed 2026-09-06. These examples reinforce the existing owners; they do not introduce another visual style or claim the product behavior is implemented. Read the 26-source review and access limits.
Keep operational text steady and unambiguous
Changing digits, shifting labels, and brittle wrapping make money and status harder to inspect.
Evaluate numeric alignment where amounts are compared, stable action labels during requests, and wrapping with real content. Treat optical adjustments as controlled system decisions. Keep the existing font families and semantic type roles.
Thorbis example. An invoice comparison aligns changing amounts without hiding their currency or precision. A long property name wraps without displacing the next action. Save remains the action; the adjacent status explains what happened.
Verify. Try zero, negative and large amounts, long names, missing values, fallback fonts, text expansion, and browser zoom. Compare readability at actual size rather than approving a magnified specimen alone.
Do not import. No blanket balanced wrapping on data rows, copied display face, custom page-local type scale, or status copy that claims more than the system knows.