DM_ Advisory · the method · how we got here

The naming
method

Naming is not a creative guess. It is a strategic problem with a creative output, run against an established canon. This is the method we used, the sources behind it, and how each one shaped the decision.

01 — THE PRINCIPLE

Strategy before words

Founders who jump straight to brainstorming land on tame, descriptive, or copycat names, because they optimise for what the product does rather than how the brand should feel against its competitors. The discipline is the opposite: write the brief, map the competitive field, pick the strategic lane, then generate at volume, then filter hard, then clear. We ran that sequence end to end. The sources below are the spine of it.

02 — THE CANON

The sources behind the method

Four bodies of work plus the DM_ category context. Each contributed a specific tool, and each tool left a mark on the result.

The Lexicon school
David Placek · Lexicon Branding
Source 01

Three criteria, and map the ocean

Placek named Pentium, BlackBerry, Swiffer, Sonos. His test: a name must win attention, be processed easily, and carry surprise. His discipline: map the whole ocean before you dive, generating thousands of candidates across distinct territories, and use small independent teams, not group brainstorms.

How we applied it. Two independent engines worked the brief blind to each other, the modern form of his two-team method. We mapped fourteen territories and generated 693 candidates before judging any of them.
Hello, My Name Is Awesome
Alexandra Watkins · Eat My Words
Source 02

The SMILE and SCRATCH test

The sharpest quality filter in the industry. A great name makes people SMILE: suggestive, meaningful, imagery, legs, emotional. A name to reject makes you SCRATCH your head: spelling-challenged, copycat, restrictive, annoying, tame, curse of knowledge, hard to pronounce.

How we applied it. Every one of the 693 candidates carried a SMILE read and a SCRATCH flag in the scoring matrix. SCRATCH violations were the first cut.
The naming taxonomy
Igor International
Source 03

Four categories, and the whitespace

Every name sits in one of four lanes: functional, invented, experiential, evocative. Plot the competitors and the open lane appears. Most industries cluster on functional and experiential, leaving the evocative lane open, which is where the breakout brands live.

How we applied it. We classified the field and the competitive set. The evocative lane is unclaimed in industrial AI agents, which is the lane we recommend ODIN own.
Sound symbolism
Sapir · Klink · Yorkston & Menon
Source 04

The sounds carry meaning first

A century of replicated research shows the phonemes in a name shift perception before the meaning does. Front vowels and plosives read precise and sharp. Nasal onsets and open vowels read warm. The effect is automatic and measurable.

How we applied it. We wrote a sound specification per agent: precision for Quality and Production, warmth for People. The shortlist was matched to it.
DM_ Category Design
DM_ Advisory · for industrials
Source 05

Bind the name to the category

The naming sat downstream of the category work. ODIN's category is Enforced Assembly, and the discipline is that every name binds to an operational metric the agent moves.

How we applied it. Each recommended name maps to its slot's work: the witness to the record, the partner to the operator, the conductor to the line.
The clearance discipline
trademark · linguistic · cultural
Source 06

Pressure-test before commitment

A name no one can own is worthless. We ran an adversarial screen for trademark crowding, linguistic safety across six markets, and cultural risk.

How we applied it. See the short note in the recommendation. The screen informs the choice; it does not override brand, and the bar is held no higher than ODIN itself.
03 — THE SEARCH TERRITORIES · THE STEP BEFORE THE NAMES

Map the ocean, then dive

Lexicon's discipline, from Ship of Gold in the Deep Blue Sea: map the whole ocean floor before you dive, rather than rushing the first promising wreck. Before a single name was generated we defined fourteen distinct themes, the territories to mine. Norse is one of fourteen, not the default. Each engine searched every territory. This is the step that sits between the brief and the 693 candidates.

01Norse mythology, deep
The Edda beyond the headline figures. Extends the ODIN parent brand. Seeded the recommended Witness set.
02World mythologies
Greek, Roman, Egyptian, Celtic, Hindu, Japanese, Slavic. Global reach for ODIN's markets. Mitra surfaced here.
03Authority & jurisprudence
Standing, warrant, charter, standard. Earned authority, the heart of Enforced Assembly.
04Metrology & measurement
Datum, calibre, gauge, true, tolerance. Proving it is built right. High engineering credibility.
05Craft & the master-builder
Wright, mason, keystone, forge. The build-it-right maker register.
06Perception & memory
Sentinel, cortex, recall, trace, witness. Observe, remember, judge.
07Materials & elements
Obsidian, basalt, flint, cobalt. Hardness and permanence. The Pentium method.
08Aviation & safety
Datum, bearing, blackbox, checklist. The highest-stakes process compliance.
09Control systems & signals
Servo, gain, setpoint, loop, latch. Native engineering credibility.
10Sound-symbolism first
Built from phonetics, not meaning. Plosive-led, two-syllable, authoritative on a floor.
11Latin & Greek roots
Verus, probus, struct, forma. Premium, intelligent feel. Seeded the Proof set.
12Invented portmanteaus
Fragments of proof, build, true, loop. Maximum trademark flexibility.
13Nature & intelligent animals
Falcon, owl, magpie, kestrel, lynx. Vigilance, collecting, intelligence.
14Light, signal, observation
Beacon, lumen, halo, signal. Premium and observational, useful for tone balance.
See the candidates each territory produced →
04 — THE FRAMEWORKS, APPLIED

The tools on ODIN's problem

Igor taxonomy: where ODIN should play

Four lanes. The evocative lane is open in industrial AI agents, so it is the lane to own.
Functional
Describes the job. Zero explanation, zero feeling. Siemens Industrial Copilot, C3.
Invented
Coined, ownable, needs filling. Cognite, Augury.
Experiential
A felt experience. They cluster and blur.
Evocative
Evokes the positioning, not the function. The rarest and strongest.
ODIN'S LANE · UNCLAIMED IN INDUSTRIAL AI

The SMILE and SCRATCH filter

The quality test run on the full field. Make people lean in, or make them wince.
SMILE · keep
  • Suggestive of the brand experience
  • Meaningful, the customer gets it
  • Imagery that paints a picture
  • Legs, a theme that extends
  • Emotional
SCRATCH · cut
  • Spelling-challenged   Copycat
  • Restrictive   Annoying
  • Tame   Curse of knowledge
  • Hard to pronounce

Sound specification per agent

The phoneme profile each slot wants, and the names matched to it.
Quality · precision
Front vowels, plosive onset. Sharp, exact, forensic.
Mimir
People · warmth
Nasal onset, open trailing vowel. The strongest warmth signal in the literature.
Mitra
Production · potency
Voiced onset, clean ending. Decisive, forward, controlled.
Vidar

Two engines, one collision

Placek's two-team method, run with two model families working blind. The agreement between them is the confidence signal.
Engine A · Claude fleet
14 territory generators, research scans, per-slot scorers. 453 scored.
+
Engine B · Codex
Independent blind divergence. 240 candidates.
The collision
44 names surfaced in both engines independently. Mitra was each engine's first pick for People.

The funnel: 693 to 4

Quantity leads to quality. The breadth of the search, then the discipline of the cut.
generated
693 candidates · 2 engines
693
five sets
~18 named slots
18
survivors
after the screen
~6
recommended
the four
4
See all 693 names and the full funnel →
05 — THE SEQUENCE

How the run executed

01

Brief and taxonomy

Wrote the naming brief against Enforced Assembly. Mapped the competitive field on the Igor grid. Picked the evocative lane.

02

Diverge

Two engines, fourteen territories, 693 candidates. Permission to generate weak names is the condition for finding the strong ones.

03

Score and cut

SMILE, SCRATCH, sound fit, fluency across six languages. Five evocative sets rose to the top.

04

Clear and decide

An adversarial trademark, linguistic, and cultural screen, then the brand-first decision, held to the parent's bar.

06 — WHY THIS GIVES CONFIDENCE

The conclusion is not a preference. It is the output of a sequenced method run at volume, cross-checked by two independent engines, filtered through three published frameworks, matched to the sound-symbolism evidence, and pressure-tested for ownership. Where the two engines agreed, the confidence is highest. That is why the recommendation leads on the names it does.

The recommendation →