Primitive Discovery in History

Primitive Discovery in History

This document gives readers external anchors for the KFD Foundation. It is explanatory, not normative. The numbered texts in decisions/KFD-N.md remain authoritative.

Each case has two layers:

• Historical record states what the cited sources support.
• KFD replay asks what becomes visible when that record is examined through declared perspective, reality pressure, candidate genesis, and qualification.

The replay is an interpretation, not a claim that KFD caused the history or that a successful product proves every KFD proposition.

This document contains retrospective historical anchors. Provisional candidates whose genesis and qualification are still unfolding live under cases/. The first live case records how applying KFD to its own propagation problem first exposed a fused Proof-Carrying Work Object hypothesis and then split it into provisional Pursuit and Warrant candidate tracks. The preserved change is evidence of a reviewable discovery process, not evidence that KFD has already discovered either Primitive.

What a primitive changes

Most work improves an answer within an existing vocabulary. A primitive changes the vocabulary itself.

The code implementing these objects matters. The deeper change is that people and machines can now point to the same object, assign responsibility to it, compose it with other objects, and ask questions that were previously expensive or unnatural.

Case 1: Git after years of kernel pressure

The scene

The Linux kernel project exchanged changes as patches and archived files from 1991 to 2002. It then used the proprietary distributed version-control system BitKeeper. In 2005, the relationship with BitKeeper’s developer broke down and free use was withdrawn. The kernel community needed another system.

Torvalds later described an unusually short construction interval. Git became self-hosting after roughly one day, and after roughly ten days he used it for his first kernel commit. He emphasized that the hard part was not the amount of code. It was deciding how to organize the data. He had already been thinking about the problem, studying failures in other systems, and working inside a kernel workflow that could require tens of merges per day.

Sources:

• Git, A Short History of Git
• Linux Foundation, 10 Years of Git: An Interview with Git Creator Linus Torvalds

What changed

Git did not merely automate patch copying. Its object model made content, history, ancestry, integrity, branching, and distributed authority composable. Many technical ingredients and neighboring systems predated Git. The primitive was the operational compression that made them fit the Linux workflow and then travel beyond it.

years of integrated pressure -> about ten days of construction

This asymmetry is consistent with a discovery bottleneck. It is not proof by itself. Prior systems, prior art, implementation ability, and adoption all remain part of the causal history.

KFD replay

The historical events were not literally confined to one person’s nervous system. Patches, repositories, discussions, and failures existed elsewhere. What was scarce was the integrated observer position from which all those pressures formed one actionable object problem.

Case 2: VisiCalc and the object hidden inside recalculation

The scene

In 1978, Harvard Business School student Dan Bricklin watched a professor sketch a complex business model. Changing one parameter required the later entries to be recalculated. Bricklin saw that a computer could carry that burden. The idea became VisiCalc, created with Bob Frankston and released in 1979.

Source:

• Harvard Gazette, A vision of computing’s future

What changed

The missing object was not arithmetic. Computers could already calculate. The change was to make a model visible as a grid of addressable cells whose values and dependencies could be edited and recalculated interactively.

The spreadsheet therefore changed the unit of knowledge work. A business model was no longer only a static document plus a person’s reconstruction procedure. It became a manipulable dependency space.

KFD replay

No dramatic forcing event is required in every case. Here, a perspective shift made an ordinary burden suddenly legible as an object problem. That variation matters: KFD-5 permits direct situated judgment as genesis and does not require failure accumulation before a candidate may be proposed.

Case 3: The log becomes architecture

The scene

In 2013, Jay Kreps described lessons from LinkedIn’s transition from one centralized database to many specialized distributed systems. Logs were not new. Kreps noted that they had existed almost as long as computers and appeared at least as early as IBM System R. They were commonly hidden inside databases for recovery and replication.

LinkedIn’s data-integration pressure changed the useful boundary. Instead of treating the log only as an internal implementation mechanism, Kreps presented it as a standalone service: a persistent, replayable record of history that could connect many producers, consumers, and projections. He also traced this experience to the focus that produced Kafka.

Source:

• Jay Kreps, The Log: What every software engineer should know about real-time data’s unifying abstraction

What changed

This is a case of promotion rather than invention from nothing. The mechanism already existed, but its previous boundary kept it subordinate to another object. A new pressure field made it useful to name the mechanism, expose its authority, and organize systems around it.

database implementation detail -> standalone integration primitive

KFD replay

The case warns against equating primitive discovery with inventing a new data structure. Sometimes reality is asking for an old mechanism to receive a new identity, boundary, and responsibility.

Everyday vignette: the trace with no declared observer

A cross-machine trace screen often looks like one coherent history. The user sees a single ordered list and naturally reads it as what happened.

But the screen may silently combine process clocks, collection delays, missing events, retries, sampling, inferred parentage, and projection rules. It is not necessarily false. It is a view whose observer and evidence boundary have been hidden by presentation.

From an implementation perspective, the local answer may be a better clock correction or query. From an operator perspective, the missing object may be a bounded action whose consequence must be reconstructed across machines. From a user perspective, the real question may be whether delegated work finished and who can safely act next.

KFD-4 does not require one of these perspectives to become the universal truth. It requires each view to remain declared and makes replay or contrast available when another perspective is needed. KFD-5 begins when any declared genesis method proposes a candidate object that may deserve its own durable boundary.

What these cases cannot rank

The cases are compatible with several genesis explanations at once:

A retrospective replay cannot identify which mechanism was necessary, which was sufficient, or which would have found the candidate first under equal evidence and resource budgets. The same success can often be narrated through several methods after the primitive is known. KFD therefore treats these cases as evidence that the methods are plausible, not evidence that perspective transformation dominates anomaly, reconstruction, causality, or compression.

The recurring structure

The common structure is not “a brilliant person had an idea.” It is:

reality pressure
  -> a declared perspective and current ontology
  -> one or more genesis methods propose a candidate
  -> qualification through use, consequences, and transfer

The historical participants still matter. Judgment, taste, implementation, authority, and timing cannot be removed from the record. The KFD claim is narrower: more of the path can be preserved and examined than biography alone normally preserves.

From historical accident to inspectable process

These cases do not prove that primitive discovery can be industrialized. They show why the missing method matters.

History usually gives later readers the promoted object and fragments of its rationale. KFD proposes a stronger path:

1. KFD-1 keeps pressure, evidence, and candidate contracts from drifting.
2. KFD-2 keeps confidence bound to those facts and exposes residual risk.
3. KFD-3 lets participants contribute trusted value without coercion.
4. KFD-4 makes observation perspectives declared and makes transformation durable, transferable, and comparable.
5. KFD-5 keeps genesis method-plural and separates it from fact-bound qualification.
6. KFD-6 asks whether causal experience can eventually sustain and compare that loop with less direct human translation.

This does not turn discovery into a deterministic formula. It changes the default from an unrepeatable convergence inside one life to an inspectable process that later participants can replay, challenge, and improve.

Limits and falsifiers

Famous successes create survivorship bias. A pressure field can produce no new primitive. A candidate may be only a useful local abstraction. A later system may subsume it. Adoption can reflect distribution power rather than object quality.

A KFD replay should therefore remain open to at least these alternatives:

• the existing object set was adequate and only implementation quality was missing;
• the candidate reflected one perspective artifact rather than a durable object;
• the candidate reflected an anomaly threshold, model class, causal confounder, or compression language rather than a durable object;
• a narrower abstraction explains the same compression;
• the object transferred poorly outside its original pressure field;
• historical evidence is too incomplete to support the proposed replay.

The cases are valuable when readers can test the lens against their own experience and reject an interpretation that does not fit. They are not origin myths and should not become one.