Notes on Naming Startups

May 1, 2026

1. 1500s-1700s — chartered, mercantile names

The name was the legal instrument. It documented who had been authorized to trade what, where, and under whose royal charter.

Real examples:

• The Muscovy Company (1555) — trade region plus corporate form, chartered for commerce with Russia.
• The East India Company (1600) — formally Governor and Company of Merchants of London Trading into the East-Indies until 1708; the long form was the legal license.
• The Royal African Company (1660) — royal descriptor plus region, signaling crown-backed monopoly.
• Hudson’s Bay Company (1670) — geographic watershed plus corporate form, still trading under that name more than 350 years later.

Common formulations:

• [Trade or region] + Company
• Royal/state descriptor + [Trade or region] + Company
• Governor and Company of Merchants of…

The naming convention was to document the charter: who is authorized, for what, and where.

2. Mid-to-late 1800s — infrastructural, territorial names

Once railroads, canals, and utilities defined the economy, names started reading like network maps. Legitimacy came from the territory you covered.

Real examples:

• Atchison, Topeka and Santa Fe Railway (chartered 1859) — a three-city route became the obvious corporate name.
• Union Pacific Railroad (1862) — cardinal-direction descriptor announcing continental scope.
• Great Northern Railway (1889) — superlative plus direction, asserting geographic reach.
• Pacific Gas and Electric (1905) — region plus utility category, the template for regulated local monopolies.

Common formulations:

• [Region or direction] + Pacific/Northern/Central + Railroad
• [City A] and [City B] Railroad
• [Place] + Gas/Electric/Telephone + Company

The naming idea was: describe the network you run and the territory you reach.

3. 1880s-1930s — branded consumer goods

Mass advertising and distribution created the modern brand. Names had to be ownable, memorable, and repeatable at national scale.

Kodak is the textbook case. George Eastman wanted a word that was short, couldn’t be mispronounced, and resembled nothing else. He liked the letter K — “a strong, incisive sort of letter” — and worked through anagrams with his mother until they landed on a word that started and ended with one.

That move would have made no sense in 1720, when a name was supposed to tell you who had been granted the right to trade in Russia.

Real examples:

• Singer (1851) — bare founder surname, ported across sewing-machine product lines.
• Campbell’s (1869) — possessive founder surname, later anchored by the red-and-white can.
• Ivory Soap (1879) — evocative compound pairing purity claim with product.
• Kodak (trademarked 1888) — coined mark, meaningless by design.
• Ford (1903) — bare founder surname applied at industrial scale.

Common formulations:

• Founder surname, sometimes possessive
• Coined product mark
• Evocative compound

The naming idea was: create a mark memorable enough to withstand a busy magazine spread, a shelf, or a radio jingle.

4. Late 1800s-1960s — corporate scale and administration

As companies got bigger and more diversified, the dominant naming style got broader, more managerial, more impersonal. The companies themselves often predate this period, but the formula crystallized as their size and reach grew. Many went by initialisms in everyday use, and the initialism form became dominant over the 20th century as broadcast and stock-ticker culture rewarded short references.

Real examples:

• Standard Oil (1870) — category adjective plus commodity, implying stability and scale.
• American Telephone and Telegraph (1885) — national descriptor plus two product categories, universally rendered as AT&T.
• General Electric (1892) — broad qualifier plus industry, signaling portfolio breadth.
• National Biscuit Company (1898) — national plus category plus corporate form, later compressed to Nabisco.
• International Business Machines (renamed 1924 from Computing-Tabulating-Recording) — descriptive long form, universally rendered as IBM.

Common formulations:

• General/National/United/Standard/American + category
• [Place] + [Category] + Corporation
• Initialism of the long form

The naming idea was: project broad, durable, administrative scale.

5. Late 1990s-2010s — software compression

Naming became a global trademark and domain-availability problem. The constraints of the dot-com era pushed toward novelty signaling: a name that sounded weird sounded new. The app era rounded the edges into soft dictionary words with friendly logos. The advantage was flexibility but the cost was sameness — every name optimized against the same constraints by the same kind of founder using the same trademark databases.

Real examples:

• Yahoo (1994) — backronym from “Yet Another Hierarchically Organized Oracle.”
• PayPal (1998) — “Pay” + “Pal,” the whole-word compound that came to define a whole subgenre of tech naming (HubSpot, DropBox, StubHub, SalesForce, Snapchat)
• Google (1998) — misspelling of “googol,” the mathematical term for 10^100.
• Flickr (2004) — dropped-vowel real word, the signature Web 2.0 move.
• Stripe (2010) — single dictionary word reapplied to payments.
• Instagram (2010) — portmanteau of “instant” and “telegram.”

Common formulations:

• Short abstract word
• Misspelled real word
• Single dictionary word used out of context
• Compound of two whole words
• Soft portmanteau

The naming idea was: own a short, trademarkable word with an available domain and a friendly app icon.

6. 2000s-2020s — counter-currents

A countertrend started forming in the 2000s, on two fronts.

The most distinctive companies began picking names that carried more weight, either by invoking a fictional canon or by behaving like an institution.

The mythic version starts with Palantir (2003), named by Peter Thiel after the seeing-stones in Lord of the Rings. Palmer Luckey named his defense company Anduril (2017) after Aragorn’s reforged sword. The Tolkien wave works because Middle-earth is a canonical world: distinctive vocabulary, weighty associations, and none of the accumulated tech-cute baggage of “Snapchat” or “Tinder.”

By the 2020s the Tolkien canon was itself getting picked over. Middle-earth Enterprises polices the IP, the best-known names are taken, and the Amazon adaptation has pulled the canon further into the mainstream. The saturation has become a running joke — Samuel Arbesman, among others, maintains a growing catalog of tech-companies-named-after-Tolkien at arbesman.net/lotrtech, and “which Tolkien names are still free” is a recurring tech-Twitter bit.

The natural response is a deeper cut. Farcaster (2020) takes its name from Dan Simmons’s 1989 novel Hyperion, where farcasters are wormhole portals that connect distant worlds instantly — a near-perfect fit for a network protocol, and a reference you can’t make without having read the book. Gene Wolfe, Ursula Le Guin, and Iain M. Banks are all sitting in similar territory. The signal shifts from “I’m ambitious” to become a cultural signifier for early investors or founding employees.

The institutional version arrived later. The Browser Company of New York (2019) sounds like a firm of architects; their product is called Arc. Software Applications Incorporated (2023, acquired by OpenAI in 2025) takes the aesthetic to its limit — so aggressively generic it loops back to distinctive. In a field of clever startup names, being maximally boring is a move. Their product is called Sky.

TBCNY/Arc and SAI/Sky are the same move twice: institutional corporate wrapper, soft modern product inside. The corporate name handles seriousness; the product name handles the app icon. Two approaches, two audiences, one team.

Real examples:

• Palantir (2003) — Lord of the Rings seeing-stones, applied to intelligence software.
• Anduril (2017) — Aragorn’s reforged sword, applied to a defense company.
• The Browser Company of New York (2019) — institutional long-form corporate name; product called Arc.
• Farcaster (2020) — wormhole portals from Hyperion, applied to a decentralized social protocol.
• Software Applications Incorporated (2023) — maximally generic corporate form, refusing to try; product called Sky.
• Applied Compute (2024) — bureaucratic noun phrase that could be the title of a 1968 Bell Labs report.
• Thinking Machines Lab (2025) — revival of a beloved defunct computing-era name, claiming inheritance.

Common formulations:

• The [Category] Company of [Place]
• [Institute / Bureau / House] of [Concept]
• Long, plain-English, deliberately formal names
• [Word from established mythological or literary canon]

The naming idea was: escape the soft-syllable app aesthetic by reaching for weight — the weight of canon or the weight of institution.

7. 2020s onward — the AI era (TBD)

Too early to call, but some trends are forming. The current crop of AI-native companies divides between names that reach for intellectual weight (Anthropic, Hippocratic) and action verbs that claim agency (Cursor, Adept, Magic, Imbue). That’s the foundation-model layer.

The much more interesting question is what naming looks like one layer down, where the new corporate forms live: OPCs, human-plus-agent teams, and eventually fully agent-run firms.

There’s a supply-side change to factor in first. Part of why so many recent names are short abstract words is artificial scarcity. Clearing trademarks, hunting for overlooked domains, and negotiating for premium ones costs more time than most founders can spare, so the inventory of “available good names” gets compressed into whatever’s left after the obvious sweep.

Agents dissolve most of that. They can scan the trademark register continuously, surface lapsed registrations and dormant marks in adjacent classes, identify good names that nobody bothered to register in less obvious TLDs, and run acquisition outreach at scale. Names that today read as “taken” will turn out to be reachable; names no human had bandwidth to find will surface. The pool widens toward more specific names than scarcity has been pushing founders toward.

Inside that wider pool, the new corporate forms each pull in different directions.

OPCs (one-person companies) at meaningful scale are still hypothetical. Solopreneurs and indie hackers have been around for decades — Pieter Levels, Tyler Tringas, the Indie Hackers community — but they’ve operated at lifestyle-business scale, not at the nine or ten-figure ARR that AI agents now make plausible for a single operator. Not many widely recognized OPCs of that size exist yet. But the structural shift is worth thinking through, because a name’s job is determined by its audience, and OPCs collapse the audience. A traditional company name has to survive multiple filters at once: customer trust, talent recruiting, investor scrutiny, partner integration. An OPC removes most of those. No employees means no recruiting brand. No outside capital, for AI-native operators who don’t raise, means no deck-ready name. What remains is the founder and the customer.

A few second-order effects fall out:

• Names get more idiosyncratic. Without having to survive a Series A conversation or an all-hands, founders can pick what they actually want — inside jokes, hobbyist references, slang terms that would feel unprofessional in a corporate context. Polsia, an AI platform run by a solo founder, is “AI slop” spelled backwards — a name its founder chose after he stopped caring about branding. A corporate naming committee would never sign off on that, which was the point.
• Names become more disposable. If the cost of launching approaches zero, agonizing over a “forever name” makes less sense. Names start behaving like usernames: stable while the project is active, swapped when the focus shifts.
• Names split sharply by who the customer is. For OPCs serving AI agents — agentic commerce rails, API-priced services, automated integrations — the name is a routing string and little else. For OPCs serving humans who specifically want hand-made or hand-curated work, the name matters more than ever. When most competitors are AI-generated, the name becomes the primary signal that a person is behind the thing.
• The founder becomes the brand. An OPC plus agents is one person in a way that hasn’t been common since the 19th-century surname era. Names may skip the company-name layer entirely and collapse into the founder’s own name or online handle.

Human-plus-agent companies will probably bring back “& Co.” — except the “Co.” is now implicit and synthetic. “Smith & Co.” worked in 1890 because everyone knew there were clerks or offspring employed with the founder. It may work again in 2030 because everyone will know there are agents behind the founder.

Or, maybe the form goes explicit, and we see things like Smith & Daemons. (Daemon, of course, is both the Unix term for an autonomous background process and the classical Greek word for an attendant spirit — two thousand years apart, both accurate.)

Fully agent-run companies are where it will get really interesting. Names picked by the agent itself will be hyper-optimized for SEO, reach, availability, and trademark in real-time. This might give rise to names that read more like product SKUs or pseudonymous handles, the way crypto founders use them, but for the org itself.

As agents become ubiquitous, the countertrend will be a move toward the old-fashioned and warm (Hartwell & Sons, Pemberton Goods), because many customers will long for nostalgia.

None of this has cohered yet. These are structural possibilities, not established patterns — what actually emerges may look nothing like any of the above.

Real examples (so far):

• Anthropic (2021) — Greek-rooted term, posture of intellectual seriousness.
• Cursor (2022) — interface verb, action-coded.
• Hippocratic AI (2023) — classical medical oath, posture of ethical seriousness.

Common formulations (speculative):

• [Person] & Co. / & Daemons
• The [Function] Bureau / Studio / Workshop
• [Handle] — pseudonymous, synthetic
• Heritage-coded human names

The naming idea (so far): a name for an OPC plus agents only has to work for the founder and the customer — much smaller surface area, much more room to get weird.

The meta-game

A name’s job is to transmit whatever form of legitimacy the era recognizes. What counts as legitimate changes over time – from charters to descriptive services to coined terms to internet-native inventions, and back.

Early eras rewarded conformity to the dominant form, because the dominant form was the legitimacy. Later eras, as conventions saturated and competition intensified, started rewarding the opposite — standing outside a tired convention became the credential.

If you’re naming a startup, knowing which kind of era you’re in is the meta-game.

What this means if you’re naming a company

When you’re staring at a blank page, the temptation is to ask what a “good” name looks like in your category. You collect competitor names, find their common features, and try to produce something that fits while being slightly different, and available.

This is how you end up with a name that’s correct and forgettable.

The better question is the inverse:

• What is the dominant naming convention in my space?
• What would it mean to ignore it?

Most categories have a regime. Crypto names skew cosmic and cryptic. AI tools skew toward soft action verbs. DTC brands skew toward lowercase friendly nouns. Enterprise SaaS skews toward acronyms and mythology. By the time you can identify the regime, it’s saturated. The interesting moves are outside of it.

That doesn’t mean every company should sound like a 19th-century railroad. The Browser Company works because the people building it are serious about software craft, and the institutional name is consistent with that posture. A name has to fit the work.

But you’re allowed to consider the full toolkit: chartered formality, infrastructural geography, founder surnames, descriptive compounds, initialisms, mythic canon, fabulous new constructs that define the agentic era. They still work. They just don’t work in the cases where they’re already what everyone else is doing.

Every company name being chosen this week is an artifact of an era. The only question, before you commit, is whether you want it to be an artifact of the convention or an artifact of your decision to step out of it.