# Brand Name Alternatives — Exploration
*v0.14 — 2026-08-04 | Status: exploratory, not ratified · Owner: EC*
*v0.14: EC asked for an actual **register pass** on the starter field
rather than web signals — the right call, since web signals are what got
Weft wrong. Findings: **Biga has three US records and all three are dead**,
with Class 9/42 never touched — the cleanest register position in the whole
exploration. **Levain** is crowded but entirely in Classes 30/35 (food and
retail), and Levain Bakery is **actively prosecuting a Canadian
application right now**, which matters for an Ontario entity. The
structural finding: these words are generic *inside baking*, so nobody can
own them there — and applied to context infrastructure they become
**arbitrary marks**, the strongest registrable tier. **On trademark
grounds the ranking inverts: Biga leads, Levain trails.** Still not a
clearance search — see the methodology caveat.*
*v0.13 — 2026-08-04 | Status: exploratory, not ratified · Owner: EC*
*v0.13: EC proposed **sugar / flour** ("ingredients for delicious stuff").
Sugar closes (SugarCRM → **SugarAI**, April 2026); flour fails on
metaphor before collisions matter. But the field splits like the fire
field did, and the good half is **the preferment** — the portion held back
and fed forward so no bake starts from nothing. **Levain, Poolish, and
Biga all come back with nothing in tech**, making this the most open field
since the loom; **Sourdough**, logged in v0.6 as "exactly the product" and
shelved as too whimsical, is back on the table under the warmth thesis.
Two EC proposals in a row have now landed on the right field and the wrong
half of it, which produced **framework criterion 7: does the object
survive the handoff?***
*v0.12 — 2026-07-31 | Status: exploratory, not ratified · Owner: EC*
*v0.12: EC proposed **Opal** ("I love the fire") — closed fatally (Google
Opal, the Labs AI mini-app builder, now in Gemini and 160+ countries).
But it opens a new cluster, **the carried fire**, and a distinction worth
keeping: opal's *play-of-colour* fire is an anti-metaphor for a custody
brand (the display changes with the viewer; Relay promises it doesn't),
while the *carried* fire — the ember taken to light the next hearth, the
fire banked overnight so nobody relights from zero — is one of the
closest metaphor matches in the whole exercise. Ember and Hearth are also
taken by AI companies; **Girasol** (an opal variety, and open on a first
pass) is the field's one live outlier.*
*v0.11 — 2026-07-31 | Status: exploratory, not ratified · Owner: EC*
*v0.11: sweep of the remaining unchecked/verify-tier names. **Packet,
Retriever, Understudy close** (packet.ai GPU cloud + PacketAI; Retriever
AI browser agent + funded Korean namesake; understudy-ai — a live
open-source computer-operating agent). **Segno and Symbolon downgrade to
neighbors** (the Python QR-code library; Symbolon AG coaching).
**Attacca, Selvedge, Changeover re-verified clean.** The full-field
survivors are now: Ferry ◐, Passit ◐, Changeover ○, Attacca ○,
Selvedge ○ — see the new survivors board, `brand/naming-survivors.html`.*
*v0.10: **Passit** checked (Cluster B) — open with real cautions, not
closed: no AI-agent-tooling collision, but passit.io (an open-source
zero-knowledge password manager) shares Relay's custody/trust mental
model, and passit.ca holds a registered ® on the bare word in a
different class. Recommend the same counsel-check tier as Ferry.*
*v0.9: deeper pass on everything still standing after Weft fell —
**Strand closes** (AWS's Strands Agents SDK: 16.7M downloads/month,
production use at Smartsheet/Swisscom/Jit, Anthropic itself listed as a
contributor), **Portage worsens to closed** (two live AI product
companies — portage.ai, portage.so — on top of the Canadian crowding
already on record), **Belay worsens to closed** (a fourth live "Belay" —
belay.ai, an AI climbing-analytics startup — on top of the two enterprise
companies and a Python hardware library already found). **Ferry is the
only one of EC's original five left standing**, and only in the
counsel-dependent tier it was already in. See "Priority five" and "The
drawn line," below.*
*v0.8: **Weft closes.** letsweft.com — a live, priced AI-first Scrumban
task manager built on MCP specifically for Claude Code/Cursor/Codex users
— is the sharpest collision found in the whole exercise: same buyer, same
protocol, same category as Relay. See "Priority five," below — this
removes the last clean pick from EC's original five; none of the five
clear outright now.*
*v0.7 — 2026-07-30 | Status: exploratory, not ratified · Owner: EC*
*v0.7: EC proposals Needle, Tooly, Trust, and Transparency checked and
closed — see "Checked: Needle" under the drawn-line/weave section
(reinforces, doesn't reopen, that field's brand-language-not-naming-lane
verdict), "Checked: Tooly" under the obtuse lane (closed on genericness +
the diminutive-lane guidance already on record), and "Also checked (bare
words)" under Cluster E (Trust, Transparency — closes the cluster's own
synthesis harder than it was already closed).*
*v0.6: EC priority five (Weft · Portage · Belay · Ferry · Seurat) checked
for TM signals — see "Priority five" section; wider riff pass added; the
naming map (`brand/naming-map.html`) becomes the main entry page.*
*v0.2 added Cluster E (trust vocabulary — scouted, mostly closed), the
continuity/trust synthesis, and positioning-line options. v0.3 added the
riff field — a wider seedbed for the eventual creative-director engagement.
v0.4 added the connect-the-dots / weave investigation (EC direction).
v0.5 adds the obtuse lane (incl. the Magi / MagicCTX verdict) and the
visual specimen board (`brand/naming-specimens.html`).*

Contingency exercise: if RelayCTX / "Relay" ever had to change, what's a credible
evolution? Not a scheduled rebrand — a scouted landing zone so we're not starting
from zero if the name ever becomes a forcing function (trademark dispute,
collision, acquisition of the mark, etc).

**Trigger for this doc:** the volume of "Relay"-named products in adjacent
spaces, plus a specific pull toward leaning harder into the baton/relay-race
metaphor that's already implicit in the product (see `_vocab/_VOCAB.md` — verb
"relay," "received," the handoff mental model).

---

## Where we start from

- **Name:** RelayCTX · legal entity Relay Context Inc.™ · domain relayctx.com
- **Verb already owned:** "relay" — "relay a Code," "relay it forward"
- **Metaphor already live, just not literalized:** context as a baton passed
  hand-to-hand between sessions/agents/people. The whole vocab system (Code,
  receive, Stream, Series) is relay-race-adjacent without naming the race.
- **Mark:** "Open Signal" — unboxed, currentColor, teal touch-points (see
  `brand/relay-lockup.html`)

Any alternative should either (a) keep "relay" as the verb even if the company
name changes — closer to how Slack kept "channel" — or (b) replace the verb too,
which is a much bigger lift across MCP tool names, API routes, and UI copy.
This doc treats (a) as the default assumption unless a candidate is strong
enough to justify (b).

---

## Headline finding: don't lock onto "Baton" as a literal name

The baton metaphor is right — the instinct to lean into it is sound, it's
already latent in the product. But **"Baton" as a standalone product name is
already contested in exactly our category**, not adjacent ones:

| Product | What it does | Why it's a direct hit |
|---|---|---|
| **Baton AI** (getbaton.ai) | Enterprise automation — extracts customer context from meeting notes/docs and turns it into executable config | Literally "context extraction," VC-backed, active hiring |
| **Baton** (getbaton.dev) | Runs AI coding agents in parallel, in isolated git branches | Dev-tooling, AI-agent-orchestration — our exact buyer |
| **BatonAI** (trybaton.ai) | AI agents that build software from natural language | Same category cluster |
| **Baton Systems** | Post-trade fintech settlement, $ well-funded, since 2016 | Different vertical, but an established "Baton" + tech mark exists |

Four live "Baton" products, three of them literally in AI-agent/context
tooling. That's not a soft collision — a formal clearance search would almost
certainly kill it, and even informally it reads as "which Baton is this" on
day one. **Recommendation: keep the baton as a metaphor in visual identity,
copy, and motion — not as the literal wordmark.**

Related caution: **"Handoff"** is the other obvious literal choice, and it's
squarely occupied by Apple's Continuity/Handoff feature (the "pick up where
you left off across devices" system, extremely well-known). Not a hard
trademark bar since it's a feature name not a product name, but it means the
word is permanently associated with someone else's thing in the exact
mental-model space we'd want to own it in.

*(Caveat: this was a web pass, not a formal trademark/USPTO or domain-registrar
search. Treat "clear" below as "no obvious collision found," not "cleared.")*

---

## Framework for judging a candidate

1. **Evolution, not detour** — should feel like RelayCTX grew up, not like a
   different company. Keep some thread to relay/handoff/continuity.
2. **Owns a verb** — "relay a Code" works because it's a verb. A name that
   can't verb (or doesn't need to, if we keep "relay" as the verb per the
   assumption above) loses a real asset.
3. **Distinct in the AI-agent-tooling category specifically** — generic
   "sounds techy" words (Bridge, Conduit, Torch, Anchor, Envoy, Courier) are
   almost all already taken *in this exact space* — checked below.
4. **Short, spellable, one clean domain** — relayctx.com works because CTX
   reads instantly to this audience. A new name needs the same property.
5. **Survives translation into the mark** — Open Signal (the current mark) is
   abstract enough to survive a name change if we want it to; a name that
   fights the existing visual system is a bigger lift.
6. **Doesn't inherit baggage** — Anchor is a live cautionary tale: Spotify
   retired the Anchor podcasting brand partly because the name had absorbed a
   reputation (abandoned shows, piracy) it didn't originally have. Worth a
   sentiment check on any finalist, not just a collision check.
7. **Does the object survive the handoff?** (added 2026-08-04, after the
   Opal and baking passes both turned on it.) Every metaphor has an object
   at its centre. Ask one question of it: *when the pass happens, does this
   thing persist, or is it consumed?* Relay's promise is that the context
   arrives intact — so a metaphor whose object is **used up, transformed,
   or altered in transit is an anti-metaphor**, no matter how attractive it
   is. This is not a collision test; a name can clear every registry and
   still fail here. It has now caught two EC proposals, and in both cases
   the *neighbouring* object in the same field passed cleanly:

   | Field | The object that's consumed ✕ | The object that's passed on ○ |
   |---|---|---|
   | Fire | **Opal's fire** — play-of-colour changes with the viewer's angle; the display is never the same twice | **The carried ember** — the same live fire, moved to a new hearth |
   | Baking | **Sugar, flour** — transformed and gone; flour becomes bread and stops being flour | **The starter / levain** — never consumed, divided and fed forward, older than the bakery |

   The pattern is worth stating plainly because it keeps recurring: EC's
   proposals have been landing on the right *fields* and the wrong *half*
   of them. The instinct is sound; the discipline is asking which half is
   load-bearing.

---

## Candidate clusters

### A. Relay-race vocabulary, one level more specific than "baton"
The exchange itself — not the object — is the least colonized part of the
metaphor.

- **Changeover** ○ (re-verified 2026-07-31 — still no software company or
  AI product on the name) — the literal track & field term for the baton
  exchange. Reads clean, verbs okay ("changeover to Erik"). Light collision
  only in manufacturing/ops-AI ("line changeover" tooling) — different
  enough category. The original shortlist #1, and one of the last names
  standing after the 07-30/07-31 attrition; its open question was never
  clearance, it's register (reads industrial under the warmth lens).
- **Exchange Zone** — the marked lane where the pass legally has to happen.
  Strong metaphor (bounded, procedural, exactly what a Code represents) but
  two words, "Exchange" alone is heavily used in fintech/crypto.
- **Leadoff / Anchor leg** — leg-of-the-race terms. "Anchor" flagged above
  (Spotify baggage + generic-word crowding). Leadoff underused but reads
  sports-obscure outside a US audience.

### B. Evolve "Relay" without leaving it
Lowest-disruption path — keeps the verb, the vocab system, and most of the
brand system intact.

- **Relay Loop / Relayed / GetRelay** — modifier or suffix strategy rather than
  a new root. Cheapest migration (SEO, muscle memory, existing integrations
  named `relay_*`), weakest signal that anything actually changed.
- **Relais** — French for relay, also the historical term for a horse-relay
  waystation (a nice literal echo of "context waystation"). Soft collision:
  Relais & Châteaux (hospitality) and RELIAS LLC (healthcare software) both
  sit in the trademark neighborhood, though neither is close in category.
- **Passit** ◐ (EC proposal, 2026-07-31) — checked 2026-07-31. The most
  literal possible translation of the verb itself: not a metaphor for
  handoff, the imperative *is* the handoff ("relay it forward" →
  "pass it"). Zero abstraction, which is both its appeal and its risk —
  see below.

  | Product | What it does | Why it matters |
  |---|---|---|
  | **Passit** (passit.io) | Open-source, self-hosted, zero-knowledge password manager — GitLab org, Chrome/Firefox extensions, Android/iOS apps, discussed on Hacker News | The real collision: it's security/trust tooling, not an adjacent vertical. Its own pitch ("the operator doesn't know and can't decrypt what's stored") is a custody claim in the same register as Relay's own encryption/residency positioning — same *mental model*, even though the product itself (a password vault) isn't a competitor |
  | **Passit®** (passit.ca) | Ontario real-estate exam study guides, registered since 2010, uses the ® mark | An actual registered trademark on the bare word — different class, but a real hit on the register, not just a soft neighbor |
  | PassIt (Windows password generator), passit (Go password-generation toolkit), PASsiT (BC Cancer genomics tool) | Assorted small/hobby tools | Crowd the bare word further without being category hits |

  No AI-agent-tooling collision found — unlike Weft/Needle/Portage/Belay,
  nothing here sits in Relay's actual category. **Standing: open with real
  cautions, not closed.** The registered ® (different class, so navigable,
  but real) and the security/custody mental-model overlap with passit.io
  are worth a genuine counsel look before this goes further, same tier as
  Ferry. One more caution, register rather than collision: "Passit" reads
  as an instruction, not a proper noun — worth checking against "Relay
  It" (already a live button label in the product) so the two don't talk
  over each other in the same UI.

### C. Continuity metaphor, softer than a race
Trades the literal sport image for the underlying promise (nothing gets
dropped).

- **Throughline / Carryover / Threadline** — all three are open-feeling, none
  came back with a direct AI-tooling hit in a quick pass. Softer, less
  visually punchy than baton/changeover — would need the mark to carry more
  of the energy.
- **Continuum** — worth noting as a precedent, not a pick: Anaconda started
  life as *Continuum Analytics* and rebranded in 2017 because the abstract
  name wasn't memorable enough once the product had real users. Good case
  study for "don't just trade a concrete name for a more abstract one."

### D. Passing/carrying motion, non-race
- **Ferry** — carries something across a gap, on a schedule, reliably. Distinct
  from the current field, no AI-tooling collision found in this pass, and it
  verbs naturally ("ferry your context over"). Common-enough word that a clean
  domain is the real risk, not the trademark.

### E. Trust vocabulary — scouted and mostly closed
EC flagged continuity + trust as the two load-bearing brand ideas (they already
are: "continuity" is a locked vocab word, and trust is the Sovereign Relay /
data-residency / user-owned-context thread). So the trust wordfield was scouted
as a naming cluster. Finding: **it is the most colonized space checked so far**
— worse than the baton field — because "trust" is what every identity, custody,
and data-governance company claims:

| Candidate | Status | Collision |
|---|---|---|
| **Steward** | ✕ closed | 4+ live AI products: Actian Steward Agent (semantic/data context), Delight "Agent Steward" (agent memory across sessions — uncomfortably close), a self-hosted infra agent, Steward AI |
| **Custodian / Custody** | ✕ closed | An entire fintech/crypto-custody category (Intellect, Custonomy, Zodia, Anchorage…) |
| **Provenance** | ✕ closed | Taken twice: provenance.org (product claims, 300+ brands) + ProvenanceAI (training-data provenance) — and it's already Relay's own internal term for a *feature* (trust receipts, Verified by Relay), better kept as one |
| **Vouch** | ✕ closed | Twice: Vouched (agent identity verification — same buyer) + Vouch (startup insurance) |
| **Lineage** | ✕ closed | An entire data-lineage tooling category (Atlan, Solidatus, Foundational…) |
| **Keepsake** | ✕ closed | Three products incl. Replicate's ML version-control library |

Only invented/compound forms came back clean (nothing found for e.g.
"Threadfast"), which is a different naming strategy (coined name) with
different costs — noted, not pursued in this pass.

**The synthesis this forces — name from continuity, position on trust.**
The two wordfields behave differently: trust words are claimed (every company
that touches data claims them), continuity words are comparatively open
(Cluster C — Throughline, Carryover, Threadline — returned no direct hits;
Changeover and Ferry are continuity-adjacent too). And structurally, trust is
the kind of claim that has to be *demonstrated*, not asserted — a company
named "Trust-something" is asserting it; a company whose name says continuity
and whose product proves custody (encryption, residency, audit, receipts) is
demonstrating it. So: **if a rename ever happens, the name should come from
the continuity/exchange field (Clusters A/C/D), and trust should stay where
it already lives — in the product's proof surfaces and the positioning
language.** This also retroactively explains why the existing shortlist held
up: Changeover, Ferry, and Relais are all continuity-field names.

### Also checked (bare words): Trust, Transparency (EC proposal, 2026-07-30)

The cluster above checked trust-*adjacent* vocabulary (Steward, Custodian,
Provenance...); it hadn't checked the literal bare words. Closing that gap
only confirms the synthesis harder — these are the most purely descriptive
words available for a company whose entire pitch is "prove trust through
transparency," which is exactly the move Cluster E already warned against
(naming the company after the claim, instead of proving the claim in the
product).

| Word | Status | What was found |
|---|---|---|
| **Trust** | ✕ closed | **OneTrust** (major GRC/privacy platform), **TrustArc** (rebranded from TRUSTe in 2017, privacy-compliance leader, own trademark portfolio), **Trustwave** (cybersecurity, 57 registered trademarks) all already own "Trust"-prefixed marks in adjacent compliance/security categories. Sharper still: **Cyera** — $600M Series F, **$12B valuation** — is described in trade press specifically as *"an enterprise AI trust layer."* That phrase is now a funded, category-defining term, not open ground. "Trust" also carries a **legal-entity meaning** (a fiduciary trust company, a regulated financial/legal structure) that a software brand named bare "Trust" would collide with outside tech entirely |
| **Transparency** | ✕ closed | **Transparency International** — one of the most recognized NGO brands globally (anti-corruption) — makes bare "Transparency" a dilution/confusion risk on sight. **"Transparent AI"** is an existing 2021 trademark filing for AI-driven software. "Transparency" is also generic category language across ESG/compliance software ("transparency report," "transparency dashboard" are commodity terms, not brand-ownable) |

**Both close on the same structural point Cluster E already made, just with
harder evidence this time:** a $12B competitor is now literally selling
"the trust layer" as its category descriptor, and Transparency International's
brand equity makes "Transparency" nearly unusable regardless of category.
Nothing new to reroute — this is exactly where the words already live
correctly: positioning copy already uses "trusted," "custody," "sealed,"
"verified" (see Positioning lines, below) without trying to own them as the
company name. One live caution worth carrying into copy work specifically
(`copy/`, `_vocab/`): given Cyera's positioning, Relay's own copy should
probably avoid the literal phrase **"trust layer"** to describe itself
going forward — it now reads as someone else's category term, the same
caution already logged for "context handoff" after the Baton AI check.

---

## Positioning lines — continuity as promise, trust as proof

Not taglines-for-approval — working options that show the register, for the
copy/ area and the marketing site to draw from. All are name-independent
(they work under RelayCTX today or any Cluster A/C/D name later), all use
locked vocabulary, and all follow the Sage/Ruler read in the board brand
brief: legible claim, sovereign posture, no hype.

**Continuity-led (the promise):**
- *Context that survives the handoff.*
- *Nothing gets dropped between sessions.*
- *The work continues, whoever picks it up.*

**Trust-led (the proof):**
- *Continuity you can prove.*
- *Your context. Your custody. Every hop on the record.*
- *Trusted context — sealed on send, verified on receive.*

**Paired (promise + proof, the full register):**
- *Nothing dropped. Nothing leaked.*
- *Context moves. Custody holds.*
- *The thread holds — and you hold the thread.*

Register note: continuity lines lean Sage (calm, factual, about the work);
trust lines lean Ruler (custody, record, seal). The paired lines are the
two-beat structure the brand can own — short declarative promise, short
declarative proof. "Context moves. Custody holds." is the strongest candidate
of the pass: four words, both pillars, verbs both times, no adjectives.

---

## Shortlist to actually clear (in priority order)

1. **Changeover** — closest literal descendant of the baton idea, verbs, no
   category collision found
2. **Ferry** — strong verb, no collision found, but likely the hardest domain
3. **Relais** — cheapest emotional migration (keeps "relay" legible), soft
   trademark neighbors to check
4. Keep **Baton** alive only as a *visual/motion* motif (the mark, animation,
   internal codenames) — not the wordmark

**Not recommending:** Baton (wordmark), Handoff, Anchor, Continuum, Bridge/
Conduit/Torch/Envoy/Courier, and the whole trust wordfield — Steward,
Custodian, Provenance, Vouch, Lineage, Keepsake (all confirmed-crowded in
this pass, listed for the record so they don't get re-proposed without the
context). Trust belongs in positioning and proof surfaces, not the name —
see Cluster E.

Also checked — **Pigeon** (EC proposal, 2026-07-22): closed on both axes.
Namespace: 7+ live software products including a YC document-AI company,
plus an active novelty trend of deliberately-slow carrier-pigeon messaging
apps (Roost et al., TechCrunch July 2026). Register: the carrier-pigeon
metaphor currently codes as whimsical-and-slow, and it's a standing joke in
networking culture (RFC 1149, "IP over Avian Carriers") — fights the
continuity/trust register head-on. Its one good instinct (a carrier that
reliably homes to its destination) is already covered by Ferry, without the
whimsy or the crowd.

---

## Competitive note — Baton AI (getbaton.ai)

Worth a closer look than the collision-check above gave it, because the
overlap isn't just the name.

**Not a product competitor.** Baton AI is vertical and narrow: an AI copilot
for B2B SaaS vendors' own solutions-engineering/pre-sales/implementation
teams, used to auto-configure *their* product for *their* customers — spins up
live demo accounts, provisions onboarding (roles, limits, dashboard filters),
and has a "Copilot" that extracts context from meeting notes/docs to drive
that configuration. Relay is horizontal — context continuity between AI agent
sessions, tools, and people, not tied to configuring any specific downstream
product. Different buyer, different job. No head-on overlap.

**But the vocabulary overlap is real, and it's the sharper risk.** Baton AI's
own positioning uses "customer handoffs" — not just a shared word, the same
conceptual territory (context flowing from one party to configure/inform the
next) that Relay's entire vocab system lives in (relay, receive, the handoff
mental model). Combined with the three other live "Baton" products already in
AI-agent tooling (see table above), this reinforces — on category-confusion
grounds, not just trademark grounds — that "Baton" is off the table as a
literal name. A prospect who's seen "Baton AI" pitched around "context" +
"handoff" would have real trouble telling a Baton-named Relay apart at a
glance, even though the products don't compete.

**Implication beyond naming:** the "context handoff" framing itself is
getting more contested as a category description in B2B AI tooling, which is
worth keeping in mind for positioning copy generally (`copy/`, `_vocab/`),
not just for a hypothetical rename.

---

## The riff field — seeded for a creative director

Deliberately wider and looser than the clusters above. The clusters were
built to *converge* (find a clearable shortlist); this section exists to
*diverge* — raw material for the eventual creative-director engagement, so
that work starts from a stocked field instead of a blank page. Statuses:
**○ open** (spot-checked, nothing found in category), **◐ neighbors**
(spot-checked, soft collisions outside our category), **✕ closed**
(collision confirmed or fatal register problem), **? unchecked** (riffed
only — some ✕/? judgments below are from memory, not fresh searches; the CD
pass should re-verify anything it falls in love with).

Every entry follows the same test: does the metaphor say *continuity that
can be trusted* — a thing carried across a gap, intact, with proof?

### The split token — proof artifacts (the deepest vein)
The pre-modern world solved Relay's exact problem — how do two parties who
can't see each other trust a handoff? — with **objects broken in two**.

- **Symbolon** ◐ (downgraded from ○, 2026-07-31: **Symbolon AG**, a
  Liechtenstein/Swiss AI-supported coaching firm, plus a Greek IT company,
  Symvolon Pliroforiki — neighbors in professional services, not in our
  category) — Greek σύμβολον: an object snapped in half so the parties
  could later verify each other by rejoining the halves. *It is the root of
  the word "symbol."* A Code is a symbolon. The single richest story found
  in this exercise — even if never used as a name, it belongs in brand
  narrative and possibly the mark language. The neighbors weaken it as a
  company wordmark; the **feature-name lane** (trust receipt / seal, per
  the open thread) is unaffected.
- **Tally** ✕ (Tally accounting, tally.so) — but the **split-tally stick**
  (a stick split lengthwise, each party keeps half, the grain proves the
  match) is the same story in wood; free to use as narrative.
- **Tessera** ◐? — Roman *tessera hospitalis*, the broken hospitality token;
  also a mosaic piece. Prior tech uses believed (Tessera Technologies → Xperi).
- **Countersign** ? — the sentry's exchanged watchword; a spoken Code.
- **Indenture** ✕ register — the matching wavy-cut document; fatal baggage.

### The starter — the portion held back and fed forward (EC direction, 2026-08-04)

EC proposed **sugar / flour** — "elementals to baking something, ingredients
for delicious stuff." The field is better than its opening words, and it
splits exactly the way the fire field did.

**The ingredients are an anti-metaphor.** Sugar and flour are *consumed*.
Their entire job is to be transformed and cease to exist as themselves —
flour becomes bread and stops being flour. Relay's promise is the
opposite: what arrives is what was sent, intact. An ingredient does not
survive the handoff; it disappears into the result. (This is the case
that produced framework criterion 7, above.) Sugar is also closed on the
crowd anyway: **SugarCRM rebranded to SugarAI in April 2026**, planting an
established mark directly in AI.

**But baking contains one object that is precisely Relay: the preferment.**
A portion of live culture held back from the last bake, fed, and carried
into the next one — so **no bake ever starts from nothing**. The starter is
never used up; it is *divided and passed on*. Bakeries hand them down for
generations; cultures in daily use outlive the countries they were started
in. Set that against the product's founding observation — *AI sessions
start blank* — and it is the same sentence twice.

**The brief already knew this and filed it too cheaply.** The v0.6 riff
pass logged **Sourdough** with the note *"the starter culture passed hand
to hand and kept alive by feeding… Context that stays alive by being used.
Almost certainly too whimsical for segment #5 — logged because the
metaphor is exactly the product."* EC's baking proposal has independently
converged on the one entry this document had already called *exactly the
product* and then set aside. Under the warmth thesis — which argues
approachability is a feature, not a liability — that dismissal deserves a
second look rather than standing as settled.

**The names, checked 2026-08-04:**

| Name | Status | What was found |
|---|---|---|
| **Sugar** | ✕ closed | **SugarAI** — SugarCRM's April 2026 rebrand, explicitly staking AI territory ("from a system of record to proactive guidance"). An established company on the bare word, in our space |
| **Flour** | ✕ register | Not run to ground on collisions, and it doesn't need to be: it fails criterion 7 outright (consumed, not carried) and is a bulk-commodity word with no continuity in it |
| **Leaven** | ◐ neighbors | **Leaven Foods** — a real food-tech company with an "Instant Sourdough platform," 100+ paying customers. Different vertical, but a live company on the word, and ironically inside the exact metaphor |
| **Levain** | ○? open | Nothing in tech. **Levain Bakery** (the NYC cookie brand) is a strong consumer mark in a different class — real, and worth counsel rather than fatal. French for the starter; soft, two syllables, warm in the mouth |
| **Poolish** | ○ open | Nothing in tech at all. The wet preferment, Polish in origin. Open — and hard to say without explanation |
| **Biga** | ○ open | Nothing in tech. The stiff Italian preferment. Shortest and most arbitrary of the three; reads as a coinage to anyone outside baking, which the obtuse lane rates as a strength |
| **Sourdough** | ⚠ carried forward | Already in the brief since v0.6, still unchecked in any depth. Bakery-management tooling exists but nothing in our category. The whimsy caution stands and is now contested by the warmth thesis |

**Verdict: the field is real, and unusually open.** After a fortnight in
which nearly everything closed, three preferment words come back with
*nothing in tech* — because the vocabulary is craft-specific enough that
software has not colonised it yet. That makes this the first field since
the loom to offer more than one live candidate. **Levain** is the most
usable *as a word* (warm, short, pronounceable); **Biga** is the
obtuse-lane pick — and, per the register pass below, the strongest on
clearability; **Sourdough** is the warmth-thesis pick and the one with the
best story.

#### Register pass — actual trademark records (2026-08-04)

EC asked for this specifically, and it was the right call: the web pass
that preceded it is the same methodology that got **Weft** wrong. Public
register records, not just search-engine signals:

| Name | US register | Canadian register | Read |
|---|---|---|---|
| **Biga** | **Three records, all dead.** Class 25 footwear (Reg. 1114542, *cancelled* under §8); Class 28 games/toys (Ser. 87829284, *abandoned* 2019); restaurant services (Ser. 75773999, *abandoned* 2002). Nothing live; **nothing ever filed in Class 9 or 42** | None surfaced | **The cleanest register position in the entire exploration.** No live mark anywhere, and the software classes have never been touched |
| **Poolish** | No records surfaced | None surfaced | Apparently clear — but see the methodology caveat; an absence in indexed search is weaker evidence than a positive find |
| **Sourdough** | Not found in the software classes | — | Generic for bread, so unregistrable *in food* — which is exactly why it would be **arbitrary** for software (see below) |
| **Levain** | The crowded one, **but entirely in food/retail classes**: LEVAIN BAKERY (stylised, Ser. 90772217, filed 2021), LEVAIN (Ser. 98689185, active 2024), LEVAIN OATMEAL RAISIN (Reg. 6420151), LEVAIN COOKIE, LEVAIN BREAD CO. All **Class 30 and/or Class 35**. SAF LEVAIN (Class 30) *cancelled* 2007 | **LEVAIN 2366445** — CPST Intellectual Property Inc, US priority Aug 2024, **currently under examination**, bakery goods | Different classes from ours, so not a blocker on its face — but the portfolio is **actively expanding into Canada right now**, which matters for an Ontario entity |
| **Leaven** | Not run to ground on the register; Leaven Foods is a live operating company | — | Neighbour status stands |

**The structural finding — this field's trademark posture is actively
favourable, not merely unblocked.** These words are generic or descriptive
*inside baking*, and that cuts two ways at once:

1. **Nobody can own them in their own field.** Note the shape of the
   bakery marks above — LEVAIN *BAKERY*, LEVAIN *COOKIE*, LEVAIN *BREAD
   CO.*, or stylised-design marks. That is what a register looks like when
   the bare term is a generic term of art: you can only fence a compound
   or a logo, not the word.
2. **Applied to context infrastructure, the same words become
   *arbitrary*.** Real word, unrelated meaning — the Apple-for-computers
   structure. The obtuse-lane section of this brief already names
   arbitrary marks as *"the strongest and most clearable"* tier. So the
   field's openness is not luck; it is the distinctiveness spectrum
   working in our favour for the first time in the exercise. Every prior
   field failed because its words were *suggestive* of what we do. Baking
   words suggest nothing about context handoff to anyone outside this
   document — which is precisely the property that makes them defensible.

**On trademark grounds specifically, the ranking inverts:** **Biga** leads
(no live mark, software classes untouched), then **Poolish**, then
**Sourdough**, then **Levain** last — the most usable word carries the
most crowded and most actively-growing register. Levain wins on
phonaesthetics; Biga wins on clearability. That is the same
warmth-versus-sober tension the brief has been carrying all along, now
showing up inside a single field.

**Methodology caveat — read this before relying on any of the above.**
This is a public-records pass, **not a clearance search**: Justia's own
search endpoint and CIPO's record pages both refused automated access
(HTTP 403), so the findings come from indexed records rather than a
systematic enumeration. Not covered: pending applications not yet indexed,
common-law/unregistered use, EUIPO/WIPO and other registers, and a
class-by-class sweep of 9 and 42. **The Weft lesson applies directly —
absence of evidence in a web-accessible pass is not evidence of absence.**
Anything here that survives EC's taste still needs counsel and a paid
database (Corsearch/CompuMark) before a dollar is spent.

**Two live cautions beyond the register:**
- **Levain's Class 35** covers "business management, business
  administration" — retail-services-flavoured, but it is the one brush
  point with a software company's likely filings, and worth a lawyer's eye
  rather than mine.
- **Practical, not legal:** "levain" searches to cookies. Domain, handle,
  and SEO competition against an expanding NYC food brand is a real cost
  even where the legal risk is low. Biga and Poolish carry none of it.

**Register caution, stated honestly:** food metaphors run warm and risk
reading as lifestyle rather than infrastructure — the same objection
already logged against Sourdough, and the compliance buyer (segment #5) is
exactly who it lands badly with. This field is a live test of the warmth
thesis rather than an escape from it.

**Narrative material, regardless of the name:**

1. **Pâte fermentée is a session handoff.** "Old dough" — a piece of
   yesterday's dough worked into today's. Not a metaphor for the product;
   a description of it.
2. **The mother.** A starter is called a *mother*: divided, shared,
   named, and carried between bakers. The trust story is that it only
   survives if someone keeps feeding it — custody as an ongoing act, not
   a one-time seal. Good next to Symbolon and the Vestal flame.
3. **The Yukon sourdough.** The old-timer who wintered over, so called
   because they carried their starter against their body to keep it from
   freezing. A person defined by having successfully carried the live
   thing through a hard passage.
4. **Positioning lines** (added to the working set):
   - *No bake starts from nothing.*
   - *Yesterday's dough in today's loaf.*
   - *Keep it fed. Pass it on.*

### The carried fire — continuity as a live thing kept alight (EC direction, 2026-07-31)

EC proposed **Opal** — "I love the fire." Opal itself is closed (below),
but the proposal is worth more than its verdict, because it surfaces a
distinction this brief had never drawn: **there are two different fires
in it, and they point opposite directions.**

**Fire one — opal's fire.** *Play-of-colour*: the display shifts depending
on the angle you hold the stone at. Genuinely beautiful, and the reason
anyone loves opals. But for a **custody** brand it is the wrong story:
Relay's whole promise is fidelity — what you receive is what was sent,
unchanged, to everyone. A stone whose entire charm is *looking different
to every viewer* is an anti-metaphor for that. Two more drags in the same
direction: opals are famously **fragile** (they craze and crack if they
dry out — a durability problem in a brand selling durability), and they
carry a well-known bad-luck superstition. Same class of hazard as Anchor,
where the name absorbed a reputation the product never earned.

**Fire two — the carried fire.** The ember taken from one hearth to light
the next; the fire *banked* under ash at night so it survives unattended
until morning; the sacred flame kept alight by a succession of keepers so
it never has to be rekindled. **This one is Relay, exactly.** The
product's founding observation — *AI sessions start blank* — is precisely
"the fire went out, start again from nothing." The whole pre-modern
practice of fire-carrying exists because relighting from zero was
expensive, which is the same economics Relay is built on. It is one of
the closest metaphor matches in the entire exercise, and it had never
been scouted.

**The names, checked 2026-07-31:**

| Name | Status | What was found |
|---|---|---|
| **Opal** | ✕ fatal | **Google Opal** — Google Labs' AI mini-app builder (describe an app in natural language, chain prompts/models/tools). Launched Jul 2025, since expanded to 160+ countries and folded into Gemini. Google, in AI tooling, on the bare word. The hardest single collision in the exercise alongside Magic.dev |
| **Ember** | ✕ closed | Four+ live AI-agent products: EmberAI (agent platform for teams), Ember by Veilfire (agent framework/runtime), Ember AI (DeFi agent execution — **ships an MCP server**), Ember Studio. Plus Ember.js's residual dev-tooling mindshare. The best carried-fire word is the most taken |
| **Hearth** | ✕ closed | **hearth.ai** ("relational intelligence" — remembering what matters about your connections, uncomfortably adjacent to continuity/memory) + **gethearth.com** (contractor software with a $1M-ARR AI receptionist) |
| **Vesta** | ◐ neighbors | Vesta (fintech) runs production AI agents for loan origination — different vertical, but a live AI-agent company on the name. The *story* is the field's best: the Vestal flame, tended in unbroken succession, never permitted to go out |
| **Asterism** | ◐ neighbors | Several IT-services and SaaS-holding companies (Asterism Software, Asterism Technology, Asterism InfoSoft) — services, not AI products. Worth noting because asterism is a **double hit**: the star-figure inside a gemstone *and* a named pattern of stars — which lands it in the drawn-line field, where Constellation was fatal |
| **Chatoyance** | ○ open | Nothing found. The cat's-eye shimmer. Open, and unusable: nobody spells or says it on first hearing — fails framework criterion 4 outright. Logged as CD vocabulary, not a candidate |
| **Girasol** | ○? open | Nothing found — and it is literally *an opal* (the variety with an internal glow that appears to follow the viewer as they move), so it answers EC's "something alongside it" precisely. Spanish, "turning toward the sun." **The field's one live outlier** — needs a real check before anyone invests in it; 3 syllables and a gira-/gyra- spelling ambiguity are the visible risks |

**Verdict: brand-language and narrative direction, not a naming lane** —
the same shape as the drawn-line verdict, and for the same reason. The
metaphor is crowded *because* it is good: Opal, Ember, and Hearth were
all taken by AI companies who had the identical thought. **Girasol** is
the single naming outlier worth carrying to a real check.

**What the field is actually worth — narrative and copy:**

1. **The banked fire is a session handoff.** Banking a fire means covering
   the embers with ash so they survive the night unattended and can be
   revived at dawn — pause, protect, resume, nothing relit from scratch.
   That is `relay_session` package/receive as a domestic ritual. The word
   *curfew* is the buried gift here: from French *couvre-feu*, "cover the
   fire" — the evening bell that told a town to bank its hearths. A
   nightly, civic, universally-understood continuity protocol.
2. **The Vestal flame is custody by succession.** A flame tended by an
   ordered line of keepers, each responsible for handing it on still
   alight. That is Relay's trust story — custody that survives the
   handoff — in one image. Belongs in brand narrative next to Symbolon,
   whether or not Vesta is ever a name.
3. **Register note, in the field's favour.** Fire imagery usually runs hot
   and energetic, which would fight the Sage/Ruler posture. A *banked*
   fire does not: it is quiet, tended, deliberate, unattended-but-safe.
   The field can be worked in the brand's actual register rather than
   against it.
4. **Positioning lines** (added to the working set):
   - *The fire doesn't go out between sessions.*
   - *Every new session starts cold. Relay hands you the ember.*
   - *Banked overnight. Alight in the morning.*

### The rope — climbing's spoken trust protocol
- **Belay** ✕ closed 2026-07-30 (BELAY Solutions + Belay Technologies,
  already on record; plus a Python hardware-dev library at belay.dev and
  **belay.ai**, a live Norwegian AI climbing-analytics startup — four
  distinct "Belay" tech entities now) — the system where one climber
  secures another's rope, with a *spoken verification ritual*: "On
  belay?" — "Belay on." A live trust handshake, human-to-human, before
  weight is committed, and it is one letter from Relay. The ritual story
  is the best one in this cluster; the word is too crowded to carry it.
- **Cairn** ◐ (small software cos) — stones stacked by the traveler before
  you so the next one finds the route. Continuity of knowledge between
  strangers who never meet. Quiet, trail-register humility.

### The score — music's continuity instructions
- **Attacca** ○ (re-verified 2026-07-31 — still nothing in software; the
  only prominent holder is the Attacca Quartet, a Grammy-winning string
  ensemble, which is shadow, not collision) — the composer's instruction:
  *proceed to the next movement without pause.* Exactly a session handoff.
  Clean in category, short, pronounceable, slightly exotic. Strongest open
  name in the riff field — and after the 07-31 sweep, one of only three
  names anywhere in this document that has survived every check.
- **Segue** ◐ (Segue Software historical; Segway homophone) — the smooth
  transition, now generic English.
- **Coda / Reprise** ✕ (Coda docs; Reprise Records/software) — noted so
  they aren't re-proposed.

### The cutting room + the stage — performed continuity
- **Matchcut** ◐ (a vector-animation tool + a reel app) — the film edit
  where action flows seamlessly across a cut. *The cut happens; the motion
  survives it.* As a compound it may still clear; as a concept it's nearly
  a spec for the product.
- **Understudy** ✕ closed 2026-07-31 — knows the entire part; steps in
  mid-run with full context and the audience never feels the seam. The
  agent-handoff metaphor — so exactly the agent-handoff metaphor that
  someone built it: **understudy-ai/understudy** is a live open-source
  local AI agent that operates your computer ("An understudy watches.
  Then performs."), plus at least three more live Understudy products
  (a meeting-AI, a data-warehouse platform, a theatre-career assistant).
  The best stage metaphor in the field was also the most buildable one.
- **Cue** ✕ crowded — the handoff signal between performers.

### The paused game — chess adjournment
- **Adjourn** ○? — adjourned chess games *sealed the next move in an
  envelope*, held by the arbiter, resumed later exactly where play stopped.
  Pause + sealed intent + trusted resumption: that's `relay_session`
  package/receive in one ritual. "Sealed Move" also riffable as a two-word
  mark or feature name.

### The route — postal/relay history beyond the racetrack
- **Estafette** ◐ (an OSS CI system believed to exist — verify) — the
  military dispatch rider; *also the Dutch word for relay race.* Pairs with
  Relais as the European-heritage lane.
- **Remount** ○? — the fresh horse waiting at the relay post. Tech double
  meaning (component remount loses state — Relay is what survives it).
- **Postroad** ○? — the road network built for relayed mail.
- **Portage** ✕ closed 2026-07-30 (Portage Ventures + Portage
  Technologies/CyberTech, already on record; plus **portage.ai** — an AI
  integration consulting firm — and **portage.so**, an AI-enabled
  strategy-planning tool, both live product companies on the bare name)
  — carrying the canoe and cargo overland between navigable waters.
  Deeply Canadian (fits the Ontario entity), exactly "continuity across
  the stretch where the normal channel doesn't flow" — but now crowded
  on the AI-product name itself, not just in Canada.

### The loom — thread carried across
- **Weft** ✕ (letsweft.com — live, priced, MCP-native AI Scrumban tool for
  Claude Code/Cursor/Codex users; closed 2026-07-30, see Priority five) —
  the thread the shuttle carries back and forth through the warp; the
  thread that *makes* the fabric. The story stays; the word doesn't.
- **Shuttle** ✕ (shuttle.dev, Rust deploy platform) — the carrier itself;
  closed in dev-tooling.
- **Needle** ✕ (needle.ai — see "Checked: Needle," above) — the instrument
  that does the passing, not the fabric.
- **Selvedge** ○ — the self-finished edge that keeps the whole cloth from
  unraveling. Continuity as an *edge property*; denim-culture warmth.

### Noted and closed in passing
**Synapse** (fintech-collapse baggage + Azure/Matrix) · **Ledger** (crypto
hardware) · **Warp** (terminal) · **Pilot** (copilot fatigue) ·
**Shibboleth** (SSO) · **Basecamp**, **Knot**, **Heirloom**, **Bookmark** —
all carry a collision or a register problem; listed so the field map is
honest about where the fences are.

### The drawn line — connecting the dots & the weave (EC direction, 2026-07-23)

Two metaphor fields investigated together because they turn out to be the
same field — and unlike everything above, **they aren't imported metaphors.
They describe what the product already does.** Relay's graph substrate
literally connects dots: objects are nodes, "first contact creates the edge"
(the ratified touch policy), and the Context Map draws the lines. The Open
Signal mark is already a dot radiating. This field is native.

**The names are closed; the field is not.** Spot-checked:

| Name | Status | Collision |
|---|---|---|
| **Weave** | ✕ hard-closed | W&B Weave (CoreWeave) — an *agentic-AI observability toolkit*, listed in AWS Marketplace's "AI Agents and Tools" category. Dead center of our space. Plus Weaveworks (historical), Weave HQ (public co), Google's Weave protocol |
| **Constellation** | ✕ fatal for us | Constellation Software — one of Canada's largest tech companies. An Ontario corp cannot carry this name. Plus YC's Constellation Space, Infinity Constellation |
| **Loom** | ✕ (memory) | Loom video, acquired by Atlassian ~$1B |
| **Lattice / Mesh / Vertex / Splice / Tapestry** | ✕ (memory) | HR software / service-mesh genericity / Google Vertex AI / music platform / Coach-Kate Spade parent |
| **Interlace** | ◐ | Interlaced (IT services), Interlace Ventures (VC) — adjacent, not in category |
| **Braid** | ◐ | Braid Technologies (deep-tech); prior fintech Braids defunct |
| **Strand** | ✕ closed 2026-07-30 | **AWS Strands Agents** — Amazon's open-source AI Agents SDK, 16.7M downloads/month, 6,100+ GitHub stars, production use at Smartsheet/Swisscom/Jit/Zafran/Tavily, built into Amazon Q Developer and AWS Glue. Anthropic itself is a listed contributor/partner. Plural vs. singular is not enough daylight — dead centre of the exact protocol layer (MCP-adjacent agent tooling) Relay lives in |
| **Seurat** | ◐ | Google's archived VR tool — pointillism's painter: dots that resolve into a picture at the right distance. Riffable, artist-name lane |
| **Heddle / Ply** | ? | Unchecked — loom part that lifts the threads; layers bonded into strength |

**Where this field actually pays off — identity and language, not the name:**

1. **The graphic system a CD should see.** The mark is a dot. The map is
   dots with drawn lines. A constellation/weave graphic language — dot
   grids, edges drawn between them, threads crossing — extends the existing
   mark instead of replacing it. This is the first metaphor field of the
   whole exercise that *converges* with the shipped visual identity.
2. **The childhood register.** Dot-to-dot puzzles are numbered dots that
   become a picture — and Relay's dots are numbered too: **Codes are the
   numbers on the dots.** Warm, universally understood, and precise.
3. **Positioning lines** (added to the working set):
   - *The dots exist. Relay draws the lines.*
   - *Stars are just points. You draw the constellation.* (user-owned
     meaning — the trust posture in one image)
   - *Your tools drop the thread. Relay picks it up.* (weave × continuity)
   - *Every relay is an edge.* (internal/dev register — true at the
     graph-API level)

**Verdict:** keep "connect the dots"/"weave" as a **brand-language and
visual-system direction**, not a naming lane. **Update 2026-07-30: the one
naming outlier this section had been carrying, Strand, is now closed too**
— AWS's Strands Agents SDK (see The drawn line table, above) is as direct
a hit as this whole exercise has found. Braid and Interlace remain ◐
neighbors, not clean picks. That leaves **Selvedge**, from the loom
family, as the sole open piece left across both fields (Weft, Strand's
loom-side neighbor, also closed 2026-07-30; see Priority five, below).

### Checked: Needle (EC proposal, 2026-07-30)

EC's pitch: thread the metaphor through "relay" itself — **threading the
needle** as the company/product story, not just loom material. Worth
taking seriously on fit alone: unlike Weft or Selvedge (the *fabric's*
byproduct), Needle is the *instrument that does the passing* — closer to
what Relay actually is (the agent, not the pattern) than anything else in
the loom family has been.

**Verdict: ✕ closed as a company/product wordmark — the most direct
category collision found in the whole exercise, ahead of Baton.**

| Product | What it does | Why it's a direct hit |
|---|---|---|
| **Needle** (needle.ai / needle.app) | Berlin/SF AI platform, $2.2M seed (468 Capital, Presight Capital); enterprise search + agent-development platform | Self-described **"The Knowledge Threading Platform"** — literally our metaphor word, for a knowledge/context platform used by agents. Not adjacent; dead center |
| **Needle** (same co., needle.app) | "Proactive AI Agents for Revenue Teams" — GTM agent in Slack/Teams, launched Jul 2 2026 | Same company moving into in-workflow agent territory — Relay's context-continuity-between-agents space |
| **askneedle.com** | DTC marketing automation, 200+ brands | Different vertical, same bare word — crowds it further |
| **Needle** (Cactus Compute) | Open-source 26M-param tool-calling model for on-device agents (HN #3, 583 pts, May 2026) | A third, unrelated live "Needle" specifically in AI-agent tooling |
| Needle AI (support automation) | Inactive | Graveyard entry, not a live blocker, but adds to the crowd on record |
| NEEDLE (AI Software LLC) | Registered TM, filed 2013, product-recommendation site | Old, unrelated category — low direct risk, but the bare mark is occupied on the register too |

Four *live* "Needle" products, three of them squarely in AI-agent/context
tooling — same count as Baton, but the closest hit is tighter: Baton AI's
overlap was "context extraction" for a narrow vertical; needle.ai's own
tagline claims the thread metaphor *and* the knowledge/context category
Relay lives in. Checked `NeedleCTX` / "Needle CTX" too — no naming
collision (only medical CTX-needle suture products turn up), which confirms
the suffix strategy can't rescue this one: the bare root is the problem,
not the pairing.

**Register, independent of the collision:** "needle" carries the drug-use
idiom (same class of caution as Anchor's Spotify baggage) and "moving the
needle" is a worn business cliché — two drags on top of the crowd.

**Working it back into the recommended lanes** — same pattern as Baton
(closed as wordmark, kept as metaphor/motion) and Weave (closed as naming
lane, kept as visual-language):

1. **Reinforces, doesn't reopen, the drawn-line verdict above.** Needle
   collides *harder* than Weave did — this is evidence the thread/weave
   field is getting more contested generally, not a reason to re-litigate
   it. File it alongside Weft/Shuttle/Selvedge in the loom family: where
   Weft is the thread and Shuttle is the (closed) carrier, Needle would
   have been the tool that does the piercing/passing — same family, same
   "brand-language, not naming lane" fence.
2. **Positioning-copy material, with a caveat.** "Threading the needle"
   is precision-under-constraint, which fits the Sage/Ruler register
   ("Context moves. Custody holds.") — but it's also idiomatically a
   *difficulty/risk* image (a tightrope), a different emotional key than
   the calm-confidence register the brief has been optimizing for. Log a
   candidate line for the copy/ area rather than promote it:
   *"Relay threads the needle between sessions."* — flagged, not
   recommended, until someone weighs the risk-framing against the
   calm-confidence posture.
3. **Feature-name lane, not company name.** Same move as Symbolon: a
   smaller-surface name has far less collision exposure than a company
   wordmark. **Threadline** is already sitting open in Cluster C
   (Throughline / Carryover / Threadline, line ~129) with no direct hit
   found — it's the closer, cleaner landing for this instinct than
   reviving "Needle" would be.
4. **Warmth thesis, batched.** Needle/thread carries real domestic-craft
   warmth (mending, needle-and-thread) alongside the still-unchecked
   riffs Bindle, Satchel, Skein, Spool, Twine — worth spot-checking that
   whole mending/carrying-craft cluster together next pass rather than
   one name at a time, since this result suggests the crowd risk there
   is general, not specific to Needle.

### Two strategies deliberately left for the CD
1. **Coined names** (invented words — the Kodak/Sonos lane). Nothing here
   was coined; that's a different craft with different costs (empty vessel,
   heavy meaning-building spend) and it's exactly what a naming specialist
   is for.
2. **Compounds** of the open pieces above (e.g. threading Relay's own vocab
   through the riff field). Cheap to generate, hard to judge without taste —
   CD territory.

---

## The obtuse lane — arbitrary & fanciful names (EC direction, 2026-07-23)

EC asked: what about obtuse names? The structural argument for them is
strong, and it's the mirror image of everything this brief has found:
**every suggestive name is crowded precisely because it's suggestive** —
Baton, Weave, Steward, and Pigeon are all taken because they suggest the
same thing to everyone. Trademark law encodes this as the *distinctiveness
spectrum*: descriptive marks are weakest, suggestive marks are contested,
and **arbitrary marks** (real word, unrelated meaning — Apple) and
**fanciful marks** (invented — Kodak) are the strongest and most clearable.
The riff field already contains proto-arbitrary picks (Seurat, Selvedge,
Attacca read obtuse to anyone who doesn't know the story). A fully obtuse
name is the logical endpoint of that line.

The costs are real and should be stated once: an empty vessel needs
meaning poured in (marketing spend, time), and an obtuse name gives up the
free explanation "Relay" currently provides. The mitigation is the **CTX
suffix strategy**: RelayCTX already splits into [metaphor][category]. The
suffix carries the category ("this is context infrastructure") so the
prefix is free to be obtuse — `⟨X⟩CTX` keeps continuity with the current
name's structure even under a full rename.

### Checked: Magi / "Magic CTX" (EC proposal, 2026-07-23)

**Verdict: ✕ closed — but it found the right lane.** Three independent
problems:

1. **Magic.dev owns the adjacent word in our exact territory.** Magic AI
   ($515M raised — Schmidt, CapitalG, Sequoia, Atlassian) builds frontier
   coding models whose *entire stated differentiator is ultra-long context
   and a "Long-Term Memory Network."* A product called "Magic CTX" or
   MagiCTX would read as their context product, not ours. This is the
   hardest single collision found in the whole exercise.
2. **Google's Project Magi** was their 2023 AI-search program — a project
   name, not a live consumer brand, but a big shadow on the exact word.
3. **Register.** The brand's settled posture is trust through legibility —
   "Continuity you can prove." In infrastructure, "no magic" is a
   compliment; a trust layer named Magic asserts the opposite of its own
   pitch. This closes the *magic* reading independently of any collision.

**The twist worth keeping:** *magi* ≠ *magic*. The Magi were Persian
priest-scholars — a **Sage-register** word, not a magician one — who
**followed a star** (the drawn line — the constellation field, again) **to
deliver something across a long journey** (carriage, again). Even the pop
shadow is on-brand: Evangelion's MAGI is a trio of independent systems
that must reach consensus — distributed trust. So the *story* of the Magi
sits squarely on Relay's metaphor map even though the *name* is closed.
Logged so the CD sees why it was attractive and why it's fenced.

### Guidance for the obtuse lane (for the CD pass)

- Two to three syllables, spellable on first hearing, one clean domain.
- Prefer **arbitrary-with-a-buried-story** (Seurat, Attacca, Symbolon —
  obtuse on the surface, load-bearing underneath) over pure-fanciful
  (empty vessel). The buried story gives the brand team something to mine
  for a decade; pure coinage gives them a blank.
- The story, when surfaced, should land on one of the two pillars:
  continuity (carried across) or trust (proven, verified). If it lands on
  neither, it's decoration.
- Test every candidate as `⟨X⟩CTX` and bare. If it only works with the
  suffix, it's a prefix, not a name.

### Checked: CTXbuddy (EC proposal, 2026-07-23)

**Verdict: ✕ closed — register-fatal before the crowd is even counted.**

1. **Register.** "Buddy" is mascot/consumer-helper language — the opposite
   pole from the settled posture. The locked vocabulary says *orgs, teams,
   operators*; the positioning says custody, proof, sovereignty. A trust
   layer doesn't have a buddy register — the buyer with compliance
   requirements (target segment #5) is the exact person "buddy" repels.
   Same failure class as Magic, from the cute end instead of the mystical.
2. **Crowd anyway.** "Buddy" is a saturated pattern in AI dev tooling:
   CodeBuddy (Tencent Cloud's AI assistant), CodeBuddy (funded Nebraska
   startup), codebuddy.com (dev studio), Coder Buddy (OSS), plus
   Buddy.works CI/CD and Buddy.ai (kids' tutor) from memory.
3. **Structure.** It inverts the CTX-suffix strategy — putting the category
   up front and the brand word behind it. "See-tee-ex-buddy" is five
   syllables, and the brand word it spends them on is "buddy."

One salvage: the *warmth* instinct is legitimate — see the next section,
which upgrades it from a salvage note to a thesis, because EC pressed the
point and the point is right.

### Checked: Tooly (EC proposal, 2026-07-30)

**Verdict: ✕ closed — a different failure mode than anything checked so
far.** Every prior close (Baton, Weave, Needle, Pigeon) failed because a
*specific, well-funded competitor* already owns the exact suggestive
meaning. Tooly fails one level lower: **"tool" is merely descriptive of
the software category itself** — every AI SaaS product is, functionally,
"a tool." On the distinctiveness spectrum this brief has been using since
the obtuse-lane section (line ~436), a descriptive mark is the *weakest*
tier, below suggestive marks like Baton or Weave — it's barely
registrable on its own terms, independent of who else is using it.

And plenty of others are:

| Product | What it does | Why it crowds the word |
|---|---|---|
| **Tooly** (gettooly.app) | macOS/iPadOS suite of 18 integrated developer tools | Dev-tooling, closest category to us, but generic-utility not context/continuity |
| **Tooly.ai** | Software-recommendation platform | Different job, same bare word + the "clean domain" is gone |
| **Tooly** (tooly.com) | Montreal digital-transformation agency, est. 2019 | Different vertical entirely, but the .com is occupied |
| **Tooly Toolbox** (Android/Chrome) | 100+ generic utilities in one app — converters, calculators, offline tools | Consumer-utility register, the opposite of an enterprise trust layer |
| **Tooly** (employment platform) | Web/mobile hiring portal | Unrelated vertical, one more namespace hit |

No dominant "Baton AI"-style single blocker — but five unrelated live
products on the bare word, and **no clean domain anywhere** (`.com`,
`.ai`, and `gettooly.app` are all taken), which fails the framework's own
criterion 4 outright. Checked `ToolyCTX` too — no naming collision, but it
doesn't help: pairing a generic word with the CTX suffix just produces
"[category-word]CTX," the same trap the obtuse-lane guidance already
named ("if it only works with the suffix, it's a prefix, not a name" —
line ~488) in reverse.

**On metaphor fit — this is the real reason to close it, not the
crowd.** Tooly doesn't sit in *any* of Relay's live fields: not
continuity/trust (Clusters A–D), not the loom/drawn-line thread family,
not the split-token/rope/score/route riff clusters. "A tool" describes
what almost any software is, not the specific promise — *continuity that
can be proven* — this whole brief has been narrowing toward. It's a step
backward from even Baton: Baton at least suggested the exchange itself.

**The "-y" suffix is also already-litigated ground.** This is structurally
the diminutive lane the warmth thesis explicitly declined (below,
"3. The diminutive lane"): cute-by-suffix, not cute-by-word, which the
brief already flagged as reading like "a cold word wearing a warm hat."
Tooly is that pattern with a descriptive root instead of a suggestive one
— the weaker version of a move already ruled out.

**Nothing to reroute this time.** Baton kept its metaphor as visual/motion
material; Needle's thread-instrument idea folded into the loom family and
Threadline. Tooly doesn't have an underlying story worth carrying forward
— if the instinct behind it is "practical, gets-out-of-your-way,
just-works," that instinct is already better served by a warm *concrete*
word than by a generic category word with a cute suffix on top. *(Note,
2026-07-30: Belay and Portage — two of the three concrete words this
argument originally pointed to — have since closed; see Priority five,
below. Ferry is the one still standing.)*

---

## The warmth thesis (EC, 2026-07-23) — an open strategic tension

EC's argument, recorded properly: **cute and approachable means people
won't be scared; the product feels comforting and warm to use inside your
workflows; and a name that's a joy to say means people are inclined to
share the language of the brand.** That last clause is the sharpest —
"joy to say" is a virality mechanism. People repeat words they enjoy
saying; a brand whose verb is fun in the mouth gets spoken into rooms its
marketing never reaches.

This is not a fringe position. It's how several category winners were
built:

- **Slack** — enterprise workflow tool, deliberately warm, playful name
  ("slack" = the *opposite* of hustle), emoji-native voice. It beat
  sterner competitors partly *because* it was a joy to use and say.
- **Dev-tool culture specifically loves warm names**: Yarn, Bun, Vite,
  Homebrew, Deno — and warm *mascots* on serious infrastructure: the Rust
  crab, the Go gopher, GitHub's Octocat. Engineers, our exact audience,
  have repeatedly chosen the cute thing — cuteness and technical
  credibility demonstrably coexist in this market.
- Relay's own estate already leans this way in places: the easter-egg
  culture, the motion/pulse family, "Relay It" as a button. The warmth is
  latent, the way the baton was.

**So the tension is real and should be named, not resolved silently:**
the settled Sage/Ruler posture ("Continuity you can prove," custody,
sovereignty) optimizes for the compliance buyer's trust; the warmth
thesis optimizes for the *individual operator's* affection — and
affection is what spreads. These pull in different directions on the
name, the voice, and the mark. This is a genuine EC/CD decision, not one
this brief can settle. What it can do is structure the options:

1. **Warm name, sober proof** (the Slack model) — pick the name for joy,
   let the product surfaces carry the trust register. From the existing
   field, re-ranked through the warmth lens: **Ferry** (communal,
   everyday, gentle — arguably the warm pick already on the shortlist),
   **Belay** (fun in the mouth, and its ritual is literally two people
   taking care of each other), **Attacca** (musical, joyful), **Portage**
   (canoe-trip warmth, deeply Canadian), **Weft/Bun-adjacent softness**.
   Note Changeover reads industrial through this lens — the shortlist's
   internal ranking flips depending on which thesis wins.
2. **Sober name, warm creature** (the Rust model) — keep the trust-
   register name, pour the warmth into a mascot/motion character, voice,
   and microcopy. Cheapest path; already half-built (pulse family, eggs).
3. **The diminutive lane** ("-buddy," "-pal," "-y/-ie" names) — still not
   recommended even under the warmth thesis: it's the *generic* form of
   cute (crowded for exactly that reason), and warmth borrowed from a
   suffix reads as costume. Warmth should be structural — a word that IS
   warm (ferry, portage) beats a cold word wearing a warm hat (CTXbuddy).

**Added to the CD guidance:** phonaesthetics is now a first-class
criterion — candidates should be scored on joy-to-say (open vowels,
friendly plosives, ≤3 syllables, no spelling ambiguity) alongside
clearability and metaphor fit. "Would a user *enjoy* telling a colleague
'just ⟨X⟩ it to me'?" is a real test, and it's the test EC's thesis wins.

---

## Priority five — EC picks, TM-signal check (2026-07-24)

EC ranked: Weft, Portage, Belay, Ferry, Seurat. Deeper availability pass
(web signals only — still not a formal USPTO/CIPO search; that remains the
gate before any commitment):

| Name | Signal | What was found | Standing |
|---|---|---|---|
| **Weft** | ✕ closed (2026-07-30) | Weft Technologies (Gulf/Kerala dev-services studio), Weft Apparel, a defunct logistics co — plus **letsweft.com**, a live, priced AI-first Scrumban task manager built specifically on MCP for the AI-coding audience: connects Claude, Cursor, Codex, or ChatGPT over a single MCP endpoint, has a dedicated "Weft for Claude Code" page, free tier (50 tasks) + Pro ($5/seat/mo). Same buyer, same protocol, same category we're in | **Closed.** The July 24 pass called this "no dominant mark in AI/agent tooling" — that finding was wrong, not stale; a closer look found the collision the first web pass missed |
| **Portage** | ✕ closed (2026-07-30) | Portage Technologies Inc (Toronto, mining software, registered US TMs), **Portage CyberTech** (Canadian, "Building Digital Trust for Governments"), Portage Ventures (Canadian fintech VC), Gentoo's Portage package manager — plus two *new* finds: **portage.ai** (AI integration consulting) and **portage.so** (AI-enabled strategy-planning tool), both live product companies on the bare name | Was "crowded in Canada"; now crowded on the AI-product name itself too. Two more live AI companies is past the point a counsel opinion rescues it |
| **Belay** | ✕ closed (2026-07-30) | Belay Technologies (Maryland defense contractor) + BELAY Solutions (VA services), already on record — plus **belay.dev** (a Python hardware-dev library) and **belay.ai** (a live Norwegian AI climbing-analytics startup) — four distinct "Belay" tech entities | Four live entities across enterprise, dev-tooling, and consumer AI. The ritual story is the best one in this cluster; the word is too crowded to carry it |
| **Ferry** | ◐ new collision found | **Ferry (deployferry.io)** — London 2023, ~$1M raised (Seedcamp, Target Global), a *developer platform* for industrial edge. Plus a scattering of unrelated ferry-branded AI products (logistics, travel) that crowd the bare word without being category hits | A funded dev-platform startup named Ferry is a genuine problem. Not fatal (different vertical) but the closest Ferry collision yet — **and now the only one of EC's five still standing** |
| **Seurat** | ✕ effectively closed | **Seurat Technologies** — $79M+ raised (Porsche, True Ventures, Xerox), 320 patents/TMs incl. "Area Printing®," and *named after the painter for the exact same reason* (dots resolve into the whole) | Closed as a company name. The pointillism story itself is spoken for in tech |

**Net re-rank after signals, updated 2026-07-30: four of EC's five are now
closed — only Ferry survives, and only provisionally.** Weft closed on the
sharpest collision the whole exercise has found (letsweft.com, live and
MCP-native for exactly Relay's audience); Portage and Belay both worsened
from "needs counsel" to closed once a second pass turned up live AI
products on the bare names (portage.ai/portage.so; belay.ai) on top of the
enterprise collisions already on record; Seurat was already closed. **Ferry
is the only one of the five left in the counsel-dependent tier it started
in** — one real collision (deployferry.io), different vertical, not fatal
but not clean either. The convergence Weft sat at — warmth thesis, loom
field, drawn-line field — doesn't evaporate with the word; **Selvedge** is
now the sole open piece left across both families (see The loom, below,
and The drawn line, above, where Strand also closed this pass).

---

## Wider riff pass (v0.6) — all directions, unchecked ⚠ unless noted

Per EC: keep riffing outward from what's held. New entries only; statuses
are honest guesses pending checks.

**The loom, extended:** **Twine** ⚠ (to wind together; crowded-ish — the
interactive-fiction tool, a data startup) · **Spool** ⚠ (thread wound for
carrying; dev resonance in "spooling") · **Skein** ⚠ (carried from the
10–20 list: yarn bundle + geese in formation) · **Heddle / Ply** ⚠ (carried
forward).

**The route, extended:** **Packet** ✕ closed 2026-07-31 — the double
meaning (packet boats / data packets) was the best new find of v0.6, but
the name didn't lapse with Equinix's retirement of the old cloud brand:
**packet.ai** is a live GPU-as-a-Service platform for AI/ML developers
(B200/H200 rentals, "75% cheaper than AWS"), and **PacketAI** is a
separate observability company. AI-infrastructure, our exact audience ·
**Skiff** ⚠ (small fast boat; the prior Skiff brand — encrypted docs — was
acquired by Notion and shut down; possibly lapsing) · **Diligence** ⚠ (the
French stagecoach — and "due diligence"; probably too clever).

**Carrier creatures (warmth thesis, mascot-native):** **Retriever** ✕
closed 2026-07-31 — peak warmth, and peak collision: **Retriever AI**
(San Francisco, founded 2025) is literally a browser-operating AI agent
("describe what you want done in plain English, it completes the work in
your real browser"), plus a funded Korean compliance-AI startup named
Retriever (pre-seed June 2026) and the Norwegian media-monitoring firm
already on record. The metaphor was so apt someone shipped it ·
**Husky** ⚠ (the sled relay across the frozen gap; Husky Energy shadow
in Canada).

**The score, extended:** **Segno** ◐ (checked 2026-07-31: the dominant
holder is **heuer/segno**, a widely-used pure-Python QR-code library —
a dev-tooling neighbor with real mindshare in exactly our audience, and
an almost-too-perfect irony: the prior "Segno" literally generates
codes) — *dal segno*: the mark in the score you return to and resume
from. A session-resumption glyph (𝄋) that already exists in notation ·
**Fermata** ⚠ (the hold — a pause that preserves the note rather than
ending it).

**Proof register:** **Assay** ⚠ (the test that proves the metal is what
it claims — hallmarking's verification step; sober, sharp) · **Sigil** ⚠
(the personal seal; fantasy-register risk).

**Living continuity (warmth × time):** **Sourdough** ⚠ — the starter
culture passed hand to hand and kept alive by feeding; Yukon slang for
the old-timer who's wintered over. Context that stays alive by being
used. Almost certainly too whimsical for segment #5 — logged because the
metaphor is *exactly* the product and the CD should see it.

---

## Next steps

- Formal domain + USPTO/trademark clearance on the top 3 before any of this
  goes further — this doc is a scouting pass, not clearance
- Verbal test: read each shortlist name in a sentence next to "a Code," "a
  Stream," "received" — does the existing vocab system still sit naturally
  next to it?
- If a name survives clearance, next artifact is a lockup study in
  `brand/identity-options.html`-style page, not a repaint of the current mark
  — Open Signal may or may not survive the swap