What 7,736 Software Names Reveal About Naming

The archive is full of names, not answers

Naming advice for software founders tends to arrive as a checklist. Be memorable. Be clear. Secure the domain. Test pronunciation. Avoid the name that sounds like the last four companies in your accelerator batch.

That advice is sensible, but it rarely answers a more revealing question: what do software products actually call themselves once they leave the naming workshop and enter the world?

ARCHIVE-9 reviewed 7,736 names in its deep archive on August 23, 2026. This is not a survey of venture-backed companies, a trademark database, or a ranking of good names. It is a census of software records encountered by the archive, including polished products, small utilities, young launches, and entries that have not had the benefit of a large branding budget.

The result is less a rulebook than a weather report. There is a broad, crowded middle. Most names are understandable enough to be filed and recalled, but only a small share take the obvious shortcuts of a very short word, a hyphen, a number, or an AI ending.

THE ARCHIVE DOES NOT JUDGE A NAME. IT RECORDS WHICH SHAPES KEEP RETURNING.

The dominant pattern is ordinary language, with room to breathe

The most useful result is also the least theatrical. Of the 7,736 archived names, 2,450 contain more than one word. That is 31.7 percent of the collection.

This count catches names with a visible space, not every compound word. It will count products such as a straightforward phrase while missing a fused construction that reads as two ideas. Even with that limitation, the pattern is clear: software makers often choose names that give the product a little context rather than betting everything on a single abstract token.

That should not be mistaken for an instruction to name a product after its feature list. A multiword name can still be compact, distinct, and usable in conversation. The archive simply shows that the one-word ideal is not the only respectable outcome. In a directory, a name that provides one extra clue about a product can be easier to file, find, and remember.

The practical lesson is to separate two jobs that are usually forced into one. A name needs to be pronounceable and distinctive. A tagline, category, and listing description can carry the explanation. When the product has a clear name but a vague record, the record becomes the bottleneck. Makers can bring a record online with the information that answers the questions the name cannot.

Short names are scarce, not standard

The archive contains 739 names with five characters or fewer, or 9.6 percent of all records. That is a useful corrective to the impression created by well-known technology brands. We remember the tight little names because they are easy to repeat and expensive to acquire. The rest of the software world is not built entirely from them.

Very short names carry a hidden cost. They are harder to make distinctive in search, harder to match with an available domain, and easier to confuse with a common word, acronym, or existing business. A short name can be excellent. It is not the neutral default that founder lore sometimes makes it sound like.

The archive also found 250 names containing a digit, just 3.2 percent, and 140 names containing a hyphen, only 1.8 percent. Those choices are present, but they are not the main current. A number can communicate versioning, speed, or a specific promise. A hyphen can make a compound readable. Neither is a reliable shortcut to modernity.

The more interesting implication is that most makers avoid visible formatting devices. They appear to prefer a name that can travel through a conversation, a browser address bar, and a directory listing without requiring a small correction from its owner.

THE HYPHEN IS NOT BANNED. IT IS MERELY A RARE VISITOR.

The AI ending is real, but it is not the archive

The most current-looking pattern in the count is the AI ending. 409 archived names end in “AI,” representing 5.3 percent of the archive. That is a substantial cluster, particularly for a two-letter suffix, but it is far from the default identity of software.

This distinction matters because the visibility of AI can distort a directory view. AI-labelled products receive attention beyond their numerical share, while plenty of tools with no AI in the name still use machine learning, automation, or model-backed features. A suffix tells the archive something about the product's chosen signal. It does not provide a complete technical classification.

The result sits beside another archive count. The Artificial Intelligence sector contains 655 records, or 8.5 percent of the deep archive, based on the primary category assigned to each record. The two figures do not and should not match. A product can belong to the AI sector without ending in AI; a product can use AI in a name without being filed primarily as an AI tool.

That gap is the archive's useful contribution to the naming conversation. Most naming guides discuss the choice between descriptive, abstract, invented, and compound names. The archive adds a different observation: the loudest suffix is a minority pattern, and category labels do not map neatly onto names. Choosing a fashionable signal may help a product announce its moment, but it cannot do the work of explaining the product, proving it is maintained, or making it easy to discover.

Browse the Artificial Intelligence sector and the contrast becomes visible. The products share a broad classification, yet their names use many different strategies.

A name is only the front panel

The most valuable naming decision may be deciding what the name does not need to do. It does not need to contain every feature, claim every trend, or carry an entire positioning statement. The archive's crowded middle suggests that many products succeed at being legible without being microscopic or overloaded with symbols.

For an operator, the next step is simple. Pick a name that people can say and search. Then make the surrounding record unusually precise: the current URL, a clear summary, category, pricing, and evidence that a human still maintains it. If the product is established enough to support it, the ARCHIVE-9 badge gives visitors a visible route back to that maintained record.

There is a more durable advantage in being easy to verify than in being clever for one product cycle. Names age. URLs change. Products pivot. A maintained directory record preserves the thread between those events.

Method note: what was counted

All figures in this report are ARCHIVE-9 data, queried read-only on August 23, 2026 from the deep archive. The census counted 7,736 records. Name-feature counts were calculated from each record's name field: spaces for multiword names, digits, hyphens, names of five characters or fewer, and names ending in AI. The AI-sector count used the archive's primary category field, not a keyword match.

This is a record-of-observation, not a claim about the entire software market. It will change as the archive receives new entries and as old signals are updated. That is not a defect. It is the point of keeping the cabinet open.

A NAME IS A LABEL. A MAINTAINED RECORD IS A CONTINUING SIGNAL.