Dating and Witnessing Invention Records

Every inventor dates their notes. Very few do it in a way that would survive a challenge, and the reason is that dating feels like a clerical act rather than an evidential one. It is the second. A date written by the only person who benefits from it establishes that somebody wrote a date; a date confirmed by a second, disinterested person who understood what they were reading establishes when the described thing existed. The distance between those two positions is the whole subject.

The practice matters less than it once did for one specific purpose and more than ever for several others. Priority contests between rival inventors, in which the earlier conceiver could displace the earlier filer, have largely disappeared from major systems: the applicant who reaches the granting office first generally wins the date. That change removed the classic reason for keeping a witnessed book. It removed none of the remaining ones.

Purpose

What a Witness Is Actually For

Four live questions still turn on corroborated dates. Inventorship, first — who conceived the subject matter that ended up in the granted claims, which determines who must be named and who is entitled to a share. Naming the wrong set of people is a defect in the grant, correctable in most systems but only on evidence, and the evidence is dated contribution notes.

Second, derivation. Where an applicant obtained the invention from somebody else rather than devising it, the true inventor can challenge the filing — but the challenge is won on documents showing prior possession and, crucially, showing communication of the idea to the person who filed. Third, ownership under employment or collaboration terms, where the operative question is often exactly when a conception occurred relative to a contract's start or end. Fourth, defence against later allegations of copying, where a dated record of independent development is the cleanest answer available.

None of those four are exotic. Collaboration is normal, employment overlaps are normal, and any invention that reaches a market will at some point be described by somebody else as obvious, derivative or already theirs. Witnessing invention records is insurance against ordinary events, priced at a couple of minutes per working session.

Method

How the Signing Should Be Done

The physical protocol has been stable for a century because it works. A bound book with pre-printed sequential page numbers, so that removal or insertion is visible. Entries in permanent ink, written on the day of the work described. No blank space left mid-page; unused space struck through with a diagonal line so nothing can be added later. Corrections made by a single line through the original text, leaving it readable, with the replacement, the initials and the date beside it — never by erasure or correction fluid, both of which destroy the credibility of the page they tidy.

Then the signature. At the foot of the entry: the inventor's signature and date, and beneath it a second line reading, in substance, read and understood by, followed by the witness's signature and date. The wording carries weight. A witness who signs only that they saw the page attests to a document's existence. A witness who signs that they read and understood it attests to the content, which is the thing in dispute.

• The witness should be technically competent to understand the entry — a colleague in an adjacent field, not a family member who nodded politely.
• The witness must have no ownership interest in the invention. A co-inventor cannot corroborate their own conception.
• The witness should be under a confidentiality obligation, in writing, signed before they read anything. This is the step most often skipped and the one with the sharpest consequences.
• Sign contemporaneously. A batch of pages witnessed in one sitting six months later corroborates the existence of pages, not their dates.
• Keep the witness reachable. A signature from somebody who cannot later be identified or contacted is worth considerably less than one who can confirm the circumstances.

Where an invention involves a physical build, photograph each stage as a separate file rather than overwriting a single image, and reference the image filenames in the written entry so that picture and text are locked together. The value of the practice becomes obvious the first time a claim has to be traced back to the week a particular tolerance was chosen — the wider logic of a working archive is set out in the inventor's evidence file, of which the witnessed book is one component rather than the whole.

Digital

Timestamps, Hashes and the Limits of Metadata

Most invention work now happens in files, and file metadata is close to worthless as evidence. Creation and modification timestamps are set by the operating system, are trivially editable, and change when files are copied, synced or restored from backup. A folder of dated documents on a personal machine demonstrates, at best, that the machine's clock said something.

Three digital mechanisms actually do the job. A version-controlled repository, where each commit contains a hash of the entire preceding history, so altering an earlier entry invalidates every hash after it. A trusted timestamping service, which signs a cryptographic hash of a document with an authoritative time — the document itself never leaves the inventor's hands, and the service can later confirm that a file with that exact hash existed at that moment. And a disclosure-document deposit scheme, where some granting offices accept a dated filing of a description as evidence of the date without examining it or conferring rights.

Note what none of them do. A timestamp proves that a document existed on a date. It says nothing about whether the description is enabling, whether the inventor was the true deviser, or whether the content is accurate — and it confers no rights at all. Digital dating replaces the date function of a witnessed book. It does not replace the comprehension function, which is why serious practice still pairs a hash with a human who read the thing.

Failure Modes

The Four Ways Good Records Still Fail

Records fail predictably. The first failure is the interested witness: a co-founder, a spouse, an investor. All three have a stake, and a stake is precisely what disqualifies the signature. The second is the incomprehending witness — somebody who signed a page of equations they could not follow, whose attestation therefore covers nothing that is actually contested.

The third is the unprotected disclosure. Handing a technical description to a witness who has signed no confidentiality agreement is a disclosure to a person under no obligation of secrecy, which in many territories is exactly the event that starts destroying novelty. The witnessing step, done carelessly, can cost the right it was meant to protect. Sign the agreement first, in writing, dated.

The fourth is retrospective tidying. An inventor preparing for a meeting rewrites a scruffy year of notes into a clean narrative, and in doing so replaces contemporaneous evidence with a reconstruction. Once discovered — and it is discovered, because the paper, the ink and the handwriting are all uniform — the credibility of every genuine entry in the file goes with it. Keep the scruffy version. Annotate it if necessary, dated as an annotation.

There is a cultural pull working against all of this. Invention is narrated publicly as a moment of insight, and that framing is everywhere: the screen fascination with the instant of discovery leaves very little room for a numbered page and a countersignature. Real development looks nothing like it. Accounts of how consumer products actually reach a shelf — the long documented iteration behind a single household mixing device, or the equally unglamorous development history behind the same invention — read as sequences of dated revisions, because that is what they are.

Which is the argument for the ritual, in the end. Witnessing invention records imposes a rhythm: work, write, explain to somebody competent, sign, date. The evidential benefit is real but occasional. The immediate benefit is that explaining an entry to a person who will have to understand it forces the entry to be clear, and clear entries are what a specification gets drafted from.

Two signatures, one date, ninety seconds. The cheapest line item in the whole process.