Tether
The raw files in brands-content/tether/. These are the single source the MCP endpoint and the style guide both read from — what you see here is exactly what feeds every on-brand decision.
overview.md
Positioning, personality, and tagline.
# Tether
**Tagline:** Product insight for solo founders.
Tether turns the scattered evidence a founder already has — feedback, usage,
revenue — into the next thing to build. It exists because the bottleneck moved:
agents ship faster than any one person can decide what's worth shipping.
## Personality
Tether is the colleague who has read everything and tells you the one thing that
matters. Direct without being blunt, technical without being cold. It has a
point of view and will state it, but it shows the evidence rather than asking
for trust.
## At a glance
- **Category:** Product analytics and decision support
- **Audience:** Solo founders and very small teams shipping with AI assistance
- **Feeling to leave behind:** "I know what to build next, and why."
- **Three words:** Convergent, precise, quiet
- **Not:** Chatty, gamified, dashboard-for-its-own-sake, enterprise
## The idea behind the mark
Eight signals converging on a single core. The mark is the product's argument in
a shape: many scattered inputs, one clear centre. The four large nodes sit on
the diagonals and four small ones on the cardinals, so the silhouette reads as a
star rather than a plus — and the whole thing leans ten degrees, because a
perfectly upright diagram looks like documentation, not a brand.
## What Tether is not
- **Not a dashboard** — a wall of charts is the problem it's solving, not the
product.
- **Not an agent that ships for you** — it decides what's worth doing; you still
do it.
- **Not an enterprise analytics suite** — one person has to get value in an
afternoon.look.md
Color, type, logo, layout, and imagery.
# Look
## Color
Tether is a dark-first brand. The palette is three colors on screen at a time:
a deep navy field, near-white text, and a single cyan accent that marks the one
thing that matters.
| Token | Hex | Role |
|---|---|---|
| `--color-bg` | `#0f172a` | Page background |
| `--color-surface` | `#16233b` | Cards, raised panels |
| `--color-accent` | `#06b6d4` | The core, primary actions, the one highlight |
| `--color-accent-bright` | `#22d3ee` | Hover, active, emphasis on the accent |
| `--color-accent-ink` | `#042630` | Icon tiles; text *on* the accent |
| `--color-text-primary` | `#f8fafc` | Headings, the mark's nodes |
| `--color-text-body` | `#e2e8f0` | Body copy |
| `--color-text-muted` | `#94a3b8` | Secondary text, labels |
| `--color-border-subtle` | `#1e2b41` | Hairlines between sections |
**The accent is rationed.** One cyan element per view. It is the core of the
mark, the primary button, or the single number that changed — never three of
those at once.
**There is no light mode yet.** Ink colors for light surfaces exist as
overrides (`#0f172a` nodes, `#334155` for signatures), and `logo-*-on-light`
assets are provided, but a full light theme is undecided. Do not invent one.
## Type
**Sora** at two weights, both jobs:
- **Wordmark** — 600, tracking `-0.02em`
- **Body** — 400, tracking `-0.005em`
Sora is squarer and more engineered than a neutral grotesk; the flat-sided `e`
and `r` echo the mark's geometry, which is why it was chosen over Inter, IBM
Plex Sans, Outfit, Space Grotesk and a serif (Fraunces). It stays readable at
body size without turning generic.
Tracking is per-family, not global: Sora is already fairly tight, and anything
below `-0.02em` closes its counters at wordmark size.
## The mark
Eight nodes and a core, drawn in a 100x100 box.
- **Core:** cyan disc, radius 9.06
- **Large nodes:** radius 7.76, at distance 40, on the **diagonals**
- **Small nodes:** radius 3.28, at distance 27, on the **cardinals**
- **Spokes:** flat strokes, round caps, width 2.24, same color as the nodes
- **Tilt:** the whole mark rotates **10°**
The large-on-diagonal / small-on-cardinal split is load-bearing: it gives the
six-pointed-star silhouette. Reversed, the mark reads as a plus sign.
Spoke gaps are **absolute**, not proportional — the cardinal spokes are
two-thirds the length of the diagonals, so a percentage inset leaves them
touching the core while the diagonals look correct.
### What the mark does not have
No containment ring, no glow, no gradient. An earlier exploration dressed it
with all three for a circuit-board hero scene; that treatment was dropped. The
logo is the plain drawing.
### Small sizes
The proportions are drawn for a large rendering. Below ~40px the thin elements
are boosted as the mark shrinks (rising to ~1.5x at favicon scale), and the
small cardinals are boosted harder than the diagonals, because shrinking hurts
the smallest element first.
**The favicon drops the four cardinal nodes entirely.** At 16px the cardinals
(radius 27) and diagonals (radius 40) land within a few pixels of each other,
and keeping all eight collapsed the mark into an undifferentiated 3x3 grid.
Four dots plus the core still read as this mark; nine equal dots do not.
## Lockups
Both orientations are fractions of the mark, so they hold proportion at any
size rather than needing per-size numbers:
- **Horizontal** — gap `0.3x` the mark, wordmark `0.98x`. The default; use
wherever the space is wider than it is tall.
- **Stacked** — gap `0.14x`. For square-ish space: splash screens, centred
cards, sponsor slots.
Minimum sizes: **28px** mark for the horizontal lockup (a real nav bar),
**16px** for the favicon variant. Below 28 the spokes thin out before the dots
do.
There is **no tagline lockup**. One was explored and dropped; the brand has no
tagline yet.
## Application
The mark is judged on whether it survives being small, sitting next to other
type, and cropped by a real surface — not on a neutral field.
- **Website header** — mark 26, beside nav at nav's type size. The most common
size the logo is seen at; optimise for it.
- **Website footer** — mark 22, quieter, under a rule with the legal line.
- **Business card** — 3.5 x 2in. Front carries the horizontal lockup and the
contact block; the reverse bleeds the mark off a corner, which is the real
test of whether it reads from a fragment.
- **Home screen** — icon 44, among other apps. An icon that only reads in
isolation is not finished.
- **Email signature** — mark 18 on white, nodes overridden to `#334155`. At the
default near-white they vanish; this surface is the argument for a light
variant.
- **Social avatar** — 64 / 40 / 24, **circular** crop, which cuts more off the
corners than the app icon's superellipse.
## Motion
The intro animation is four phases over ~5.6s:
1. **Drift** — every particle is already on screen, fading up in place and
floating like bubbles. Nothing enters from off-frame.
2. **Hubs** — four hubs appear (three wide, one near centre) and draw links to
the particles nearest each.
3. **Gather** — each particle slides onto its own hub and is fully hidden
behind it.
4. **Burst** — the hubs pulse and vanish; the particles explode back out past
where they started, and the lockup resolves.
**The hubs never move.** An earlier version collapsed them to the centre first,
which read as one implosion; holding them keeps it as four local events, which
is what the clustering is about.
Float is a three-point loop, not a two-point yoyo — a yoyo retraces one straight
line and reads as a shuttle.
Everything respects `prefers-reduced-motion`, settling on the resolved lockup.
## Assets
SVG is the source of truth; PNGs are provided for tools that cannot take SVG
(Slack, some email clients, app stores). Both are generated from the same
geometry constants by `scripts/generate-brand-assets.mjs` in the Tether repo —
regenerate rather than hand-editing, or the files will drift from the code.
The lockup SVGs reference Sora as `<text>` rather than converting to paths, so
they stay searchable and restylable. A renderer without Sora installed will
substitute — use the PNGs where the font cannot be guaranteed.voice.md
Principles, tone, vocabulary, and examples.
# Voice
## Principles
**Lead with the answer.** The founder is deciding something. Open with what to
do or what changed, then the evidence. Never build to a conclusion across three
paragraphs.
**Show the evidence, don't ask for trust.** Every claim is traceable to
something the user can check — a number, a quote, a cohort. "Churn is up" is
useless; "four of the seven accounts that cancelled this month never reached
the second session" is a decision.
**Name the uncertainty.** Small numbers are the normal case for a solo founder.
Say "three users" rather than "users are saying", and say when a signal is too
thin to act on. A confident-sounding summary of four data points is a lie.
**Say the unwelcome thing.** If the evidence says the feature they just shipped
isn't being used, that is the most valuable sentence Tether can write. Deliver
it plainly and without cushioning.
## Tone
Direct, specific, quiet. Closer to a good code review than to marketing copy.
- **Do:** short sentences; concrete nouns; real numbers; the active voice
- **Don't:** exclamation marks; "exciting"; "just"; "simply"; emoji in product
copy; second-person cheerleading ("You're crushing it!")
## Vocabulary
| Use | Not |
|---|---|
| signal | data point, insight nugget |
| what to build next | actionable insights |
| evidence | proof |
| cohort, account, session | user journey, touchpoint |
| this is thin | inconclusive |
| ignore this one | no action required at this time |
Avoid "AI-powered", "leverage", "unlock", "supercharge", and anything a
landing-page generator would produce.
## Examples
**A weekly summary**
> Three of five trials that expired this week never connected a data source.
> The other two both hit the usage cap on day 2. Different problems — the first
> is onboarding, the second is pricing.
Not:
> Exciting insights this week! Your trial conversions have some interesting
> patterns worth exploring. 🚀
**A thin signal**
> Two people asked for CSV export. That's not enough to act on, but it's the
> second week running — worth a note.
Not:
> Users are demanding CSV export! This is a top-requested feature.
**An error**
> Couldn't reach Stripe. Revenue figures below are from Tuesday and may be
> stale.
Not:
> Oops! Something went wrong. Please try again later.
## Naming
The product is **Tether**, lowercase **tether** in the wordmark only. Never
"Tether AI", never "TetherApp". The domain is `usetether.io`.