About
Tool No. 51 of Pikkola, 21 November 2026
An identifier is a small promise: this value has not been used before, and it will not be used again. Every kind below keeps that promise in a different way, and the difference matters when the value becomes a database key, a URL or a log line.
What it does
- Makes UUID v4, UUID v7, ULID and NanoID values, one to one hundred per run, and lets several kinds run together.
- Writes the result as plain lines, a URN, a JSON array, a SQL INSERT or a TypeScript union, with hyphens and letter case under control.
- Copies the text in one press and downloads it as a .txt or a .json file.
- Reads any identifier pasted back: version, variant, NIL and MAX, and the timestamp inside v1, v6, v7 and ULID.
- Draws the bit layout field by field, colours each nibble, and shows the same moment in UTC, local time and Unix milliseconds.
- Computes collision odds on the birthday bound for each kind and for a count you choose.
- English, Korean and Japanese, a light and a dark theme, and a phone layout that does not scroll sideways.
UUID v4
Sixteen random bytes, with the version nibble set to 4 and the two variant bits set to 10. That leaves 122 bits of randomness and nothing else: no clock, no node, no counter. Two v4 values made a microsecond apart in the same process are unrelated in every way except their fixed bits.
The randomness comes from crypto.getRandomValues, which is the platform's cryptographic generator. The page has no fallback to Math.random, because a fallback is exactly where a weak value would come from without anyone noticing.
UUID v7
RFC 9562 defines v7 as a Unix millisecond timestamp in the first 48 bits, then the version, then 12 bits of rand_a, then the variant, then 62 bits of rand_b. Because the clock comes first and is written most significant byte first, a list of v7 values sorts by creation time as plain text. That is the whole point: a B-tree index on such a key appends instead of splitting pages in every direction.
This page uses the 12 bits of rand_a as a counter. Inside one millisecond each new value adds one, so a batch made in a tight loop is still strictly increasing. The counter starts at a random value each millisecond and steps the clock forward if it ever runs past 4095. The 62 random bits keep the value unguessable even when an attacker knows exactly when it was made.
ULID
A ULID is 128 bits written in Crockford base32: ten characters of 48-bit millisecond time and sixteen characters of 80 random bits, 26 characters in all. The alphabet leaves out I, L, O and U so that a person reading a value aloud does not confuse them with 1 and 0, and decoding is case-insensitive. The first character can only be 0 to 7, because ten base32 characters hold 50 bits and the timestamp is 48; a larger first character would overflow the field, and the parser rejects it.
NanoID
NanoID is not a standard. It is a small library and a convention: draw a fixed number of characters from an alphabet you choose. The default alphabet is 64 URL-safe symbols, and the usual length is 21, which is about 126 random bits. This page samples with rejection: a random byte is masked to the next power of two above the alphabet size and thrown away when it lands outside the alphabet, so every character is equally likely. Taking a modulo instead would bias the first characters of the alphabet, a small flaw that gets larger as the alphabet gets smaller.
You can set the length from 8 to 64 and paste any alphabet. Repeats are removed, because a repeated character would make one symbol twice as likely as another, and the note under the field shows the resulting bits.
What the parser can tell
A UUID carries its version in the high nibble of the 13th hexadecimal digit and its variant in the top bits of the 17th. Version 1 and 6 hold a 60-bit count of 100-nanosecond steps since 15 October 1582, split differently; v7 holds the Unix millisecond directly. ULID holds the same kind of millisecond clock in its first 48 bits. All of those can be turned back into a moment, and the page shows it in UTC, in your timezone and as a number.
What it cannot tell is whether the value was generated well, who made it, or what it names. A NanoID has no version, no checksum and no fixed shape, so the parser can only say that a string has the right alphabet and length. It says so in those words rather than pretending to know more.
What it deliberately does not do
- It does not use
Math.randomanywhere, and it does not fall back to it. - It does not v1 or v6 generate, because both want a node identifier and a clock sequence that a browser cannot supply honestly.
- It does not guess at a NanoID. A value that looks like one is labelled a shape, not a kind.
- It does not check values against a registry, a database or a server. There is no server.
- It does not promise that an identifier is a secret. See the third rule on the workbench.
Privacy, in one paragraph
Every value is made in the tab and stays there. The content policy sets connect-src 'none', so the page could not send anything even if a later change tried to. Only your language, theme and control settings are kept in local storage. Generated values and anything you paste into the parser are never written to it, and the page has no account, no cookie and no analytics.
Contact
A value the parser gets wrong, a format that is missing, a Korean or Japanese sentence that lands badly: hello@pikkola.com. If a value breaks the page, send the value itself; it is the whole reproduction.