code-evolution-starter.json

{
  "dataset": "Code Evolution Starter Corpus",
  "version": "0.1.0",
  "as_of": "2026-08-13",
  "coverage": {
    "start_year": 1920,
    "end_year": 2026,
    "target_examples_per_year": 5,
    "target_examples_total": 535,
    "status": "source-backed seed",
    "curated_examples": 57,
    "remaining_target_slots": 478
  },
  "notes": [
    "The annual quota is a research target; unsupported gaps are not filled with invented examples.",
    "Short excerpts may be illustrative or reconstructed. Always inspect snippet_status and artifact_status.",
    "For larger excerpts use the immutable locator and verify the underlying license before redistribution."
  ],
  "examples": [
    {
      "id": "1928-ibm-card-control-panel",
      "sort_date": "1928-01-01",
      "year": 1928,
      "date_display": "1928",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "IBM 80-column card and wired control panel",
      "organization": "IBM",
      "problem": "Reconfigure high-volume accounting, sorting, totals, and reports without rebuilding the machine.",
      "problem_family": "business_data",
      "domain": "Unit-record business processing",
      "language": "Card schema and patch-cord wiring",
      "platform": "IBM punched-card machines",
      "authoring_mode": "physical_configuration",
      "authoring_level": 1,
      "artifact_status": "conceptual",
      "snippet_status": "illustrative",
      "snippet": "READ employee card\nWIRE hours field -> multiplier\nWIRE pay rate -> multiplier\nPUNCH gross pay to output card",
      "snippet_lines": 4,
      "snippet_locator": "Control-panel examples in IBM punched-card training manual",
      "code_url": "https://bitsavers.computerhistory.org/pdf/ibm/punchedCard/Training/F20-0074_An_Introduction_to_IBM_Punched_Card_Data_Processing.pdf",
      "source_url": "https://www.ibm.com/history/punched-card",
      "source_title": "IBM: The punched card",
      "revision": "",
      "license": "Manual scan; reuse license unclear",
      "redistribution": "locator_only",
      "selection_reason": "The card and plugboard made data layout and business logic reconfigurable before stored programs.",
      "provenance_tier": "B",
      "confidence": "high",
      "caveat": "The date anchors the 80-column format, not one surviving program; the excerpt is modern illustrative notation.",
      "tags": [
        "punch-card",
        "accounting",
        "plugboard",
        "business"
      ]
    },
    {
      "id": "1931-bush-differential-analyzer",
      "sort_date": "1931-01-01",
      "year": 1931,
      "date_display": "1931",
      "date_basis": "executed",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Differential Analyzer configuration",
      "organization": "MIT",
      "problem": "Solve differential equations that were prohibitively slow to integrate by hand.",
      "problem_family": "science_control",
      "domain": "Analog scientific computing",
      "language": "Equations, gear ratios, and shaft connections",
      "platform": "MIT Differential Analyzer",
      "authoring_mode": "physical_configuration",
      "authoring_level": 1,
      "artifact_status": "conceptual",
      "snippet_status": "illustrative",
      "snippet": "set integrator_1 input = dy/dx\nset gear ratio = scale_x / scale_y\nconnect integrator_1 output -> plotter_y\nturn drive shaft",
      "snippet_lines": 4,
      "snippet_locator": "Machine setup described in institutional collection material",
      "code_url": "",
      "source_url": "https://siris-libraries.si.edu/ipac20/ipac.jsp?uri=full%3D3100001~%211105656%210",
      "source_title": "Smithsonian Libraries: Differential Analyzer material",
      "revision": "",
      "license": "Catalog material; no open reuse license stated",
      "redistribution": "locator_only",
      "selection_reason": "It turned a physical arrangement of reusable components into an executable mathematical method.",
      "provenance_tier": "B",
      "confidence": "high",
      "caveat": "This is continuous analog configuration, not digital source code; the excerpt is explanatory pseudocode.",
      "tags": [
        "analog",
        "differential-equations",
        "simulation",
        "engineering"
      ]
    },
    {
      "id": "1936-turing-transition-table",
      "sort_date": "1936-11-01",
      "year": 1936,
      "date_display": "Nov–Dec 1936",
      "date_basis": "published",
      "date_precision": "range",
      "date_confidence": "high",
      "title": "Turing-machine transition tables",
      "organization": "Alan M. Turing / London Mathematical Society",
      "problem": "Define effective computation precisely enough to prove what no algorithm can decide.",
      "problem_family": "science_control",
      "domain": "Mathematical logic",
      "language": "State / symbol / action table",
      "platform": "Abstract universal machine",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "conceptual",
      "snippet_status": "illustrative",
      "snippet": "state q1, symbol 0 -> write 1, move R, state q2\nstate q1, symbol 1 -> write 0, move L, state q1\nstate q2, symbol blank -> halt",
      "snippet_lines": 3,
      "snippet_locator": "Machine tables in On Computable Numbers",
      "code_url": "https://www.cs.virginia.edu/~robins/Turing_Paper_1936.pdf",
      "source_url": "https://turingarchive.kings.cam.ac.uk/computable-numbers",
      "source_title": "The Turing Digital Archive: Computable Numbers",
      "revision": "",
      "license": "Rights vary by jurisdiction; locator preferred",
      "redistribution": "locator_only",
      "selection_reason": "It established a machine-independent model of programs and universal computation.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The transition excerpt is illustrative and was never executed on physical hardware.",
      "tags": [
        "theory",
        "computability",
        "universal-machine",
        "algorithm"
      ]
    },
    {
      "id": "1942-zuse-trajectory-plan",
      "sort_date": "1942-01-01",
      "year": 1942,
      "date_display": "1942",
      "date_basis": "authored",
      "date_precision": "year",
      "date_confidence": "medium",
      "title": "Zuse trajectory-calculation plan",
      "organization": "Konrad Zuse",
      "problem": "Calculate aeronautical flight trajectories with a reusable computation plan.",
      "problem_family": "science_control",
      "domain": "Aeronautical engineering",
      "language": "Rechenplan notation",
      "platform": "Zuse machine lineage",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "facsimile",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "ZIA 0365 / GMD 012/009",
      "code_url": "https://zuse.zib.de/file/1rUAfKDkirW8o3gT/64/2f/ab/ca-d50b-46ab-ac88-2de37f196247/0/original/c2d0c40d1254593104d1eef252134281.pdf",
      "source_url": "https://zuse.zib.de/pdfs",
      "source_title": "Konrad Zuse Internet Archive: PDF collection",
      "revision": "ZIA 0365",
      "license": "Archive item; verify embedded license before reuse",
      "redistribution": "locator_only",
      "selection_reason": "An unusually early surviving plan for an applied engineering computation.",
      "provenance_tier": "B",
      "confidence": "medium",
      "caveat": "The intended execution target and whether this exact plan ran require manuscript-level verification.",
      "tags": [
        "zuse",
        "trajectory",
        "aeronautics",
        "rechenplan"
      ]
    },
    {
      "id": "1944-colossus-lorenz",
      "sort_date": "1944-02-05",
      "year": 1944,
      "date_display": "5 Feb 1944",
      "date_basis": "executed",
      "date_precision": "day",
      "date_confidence": "high",
      "title": "Colossus Lorenz wheel-setting setup",
      "organization": "Bletchley Park",
      "problem": "Reduce the time needed to recover Lorenz cipher wheel settings.",
      "problem_family": "systems_infrastructure",
      "domain": "Wartime cryptanalysis",
      "language": "Paper-tape loop, plugs, cords, and switches",
      "platform": "Colossus Mk I",
      "authoring_mode": "physical_configuration",
      "authoring_level": 1,
      "artifact_status": "reconstruction",
      "snippet_status": "illustrative",
      "snippet": "loop intercepted tape\nset wheel pattern switches\ncount matches for candidate offset\nprint highest-scoring settings",
      "snippet_lines": 4,
      "snippet_locator": "Rebuilt Colossus operating description",
      "code_url": "https://www.tnmoc.org/rebuilding-colossus",
      "source_url": "https://www.tnmoc.org/colossus",
      "source_title": "The National Museum of Computing: Colossus",
      "revision": "",
      "license": "Museum web content copyrighted",
      "redistribution": "locator_only",
      "selection_reason": "It combined electronic speed with reconfigurable logic for a consequential real-world search problem.",
      "provenance_tier": "B",
      "confidence": "high",
      "caveat": "Original machines and most documentation were destroyed; the excerpt describes the reconstructed workflow.",
      "tags": [
        "cryptanalysis",
        "paper-tape",
        "wartime",
        "search"
      ]
    },
    {
      "id": "1945-edvac-first-draft",
      "sort_date": "1945-06-30",
      "year": 1945,
      "date_display": "30 Jun 1945",
      "date_basis": "published",
      "date_precision": "day",
      "date_confidence": "high",
      "title": "First Draft of a Report on the EDVAC",
      "organization": "University of Pennsylvania / John von Neumann",
      "problem": "Design a reusable electronic computer whose instructions and data share addressable memory.",
      "problem_family": "systems_infrastructure",
      "domain": "Computer architecture",
      "language": "Logical block description",
      "platform": "Proposed EDVAC",
      "authoring_mode": "encoded_instructions",
      "authoring_level": 2,
      "artifact_status": "conceptual",
      "snippet_status": "unavailable",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Full report",
      "code_url": "https://library.si.edu/digital-library/book/firstdraftofrepo00vonn",
      "source_url": "https://library.si.edu/digital-library/book/firstdraftofrepo00vonn",
      "source_title": "Smithsonian Libraries: First Draft of a Report on the EDVAC",
      "revision": "",
      "license": "CC0 / no known copyright",
      "redistribution": "public_domain",
      "selection_reason": "The report widely circulated the stored-program logical design that shaped many postwar machines.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "It is an architecture document, not running application code; credit for the architecture is historically contested.",
      "tags": [
        "stored-program",
        "architecture",
        "memory",
        "edvac"
      ]
    },
    {
      "id": "1947-booth-britten-arc-coding",
      "sort_date": "1947-09-01",
      "year": 1947,
      "date_display": "Sep 1947",
      "date_basis": "published",
      "date_precision": "month",
      "date_confidence": "high",
      "title": "Coding for A.R.C.",
      "organization": "Birkbeck College / Institute for Advanced Study archive",
      "problem": "Express scientific routines systematically and move them from a relay machine to an electronic successor.",
      "problem_family": "science_control",
      "domain": "Crystallography and scientific computing",
      "language": "Binary orders and contracted symbolic notation",
      "platform": "A.R.C.",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "facsimile",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Booth & Britten report scan",
      "code_url": "https://albert.ias.edu/entities/publication/27e8d743-e430-4d4d-821c-b7a7944607a7",
      "source_url": "https://albert.ias.edu/entities/publication/27e8d743-e430-4d4d-821c-b7a7944607a7",
      "source_title": "IAS Repository: Coding for A.R.C.",
      "revision": "",
      "license": "No explicit open reuse license",
      "redistribution": "locator_only",
      "selection_reason": "It is among the earliest systematic descriptions of mnemonic, assembly-like coding.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "Mnemonic notation should not be conflated with an automatically executing assembler.",
      "tags": [
        "assembly",
        "symbolic-code",
        "arc",
        "crystallography"
      ]
    },
    {
      "id": "1948-eniac-monte-carlo",
      "sort_date": "1948-04-12",
      "year": 1948,
      "date_display": "12 Apr 1948 (first run); later 1948 listing",
      "date_basis": "executed",
      "date_precision": "day",
      "date_confidence": "high",
      "title": "ENIAC Monte Carlo neutron transport",
      "organization": "Los Alamos / University of Pennsylvania",
      "problem": "Simulate neutron propagation with nested loops, branching, and shared subroutines.",
      "problem_family": "science_control",
      "domain": "Nuclear physics",
      "language": "Register code and flow diagrams",
      "platform": "Modified ENIAC",
      "authoring_mode": "encoded_instructions",
      "authoring_level": 2,
      "artifact_status": "transcription",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Monte Carlo Second Run Code Report",
      "code_url": "https://eniacinaction.com/docs/MonteCarloSecondRunCodeReport.pdf",
      "source_url": "https://computerhistory.org/blog/programming-the-eniac-an-example-of-why-computer-history-is-hard/",
      "source_title": "Computer History Museum: Programming the ENIAC",
      "revision": "Second Run listing",
      "license": "Scans reproduced by permission; locator only",
      "redistribution": "locator_only",
      "selection_reason": "It was a complex electronic program using recognizable modern control structures for consequential simulation.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The surviving listing is a later second run, and any 'first modern code' claim depends on definition.",
      "tags": [
        "eniac",
        "monte-carlo",
        "simulation",
        "neutron-transport"
      ]
    },
    {
      "id": "1948-manchester-baby-factor",
      "sort_date": "1948-06-21",
      "year": 1948,
      "date_display": "21 Jun 1948",
      "date_basis": "executed",
      "date_precision": "day",
      "date_confidence": "high",
      "title": "Manchester Baby highest-factor program",
      "organization": "University of Manchester",
      "problem": "Stress-test writable random-access electronic memory with a long factor search.",
      "problem_family": "systems_infrastructure",
      "domain": "Computer engineering commissioning",
      "language": "17-instruction binary machine code",
      "platform": "SSEM 'Baby'",
      "authoring_mode": "encoded_instructions",
      "authoring_level": 2,
      "artifact_status": "reconstruction",
      "snippet_status": "illustrative",
      "snippet": "b = a - 1\nwhile a % b != 0:\n    b = b - 1\nreturn b",
      "snippet_lines": 4,
      "snippet_locator": "Reconstructed notebook listing on Manchester Digital 60 page",
      "code_url": "https://curation.cs.manchester.ac.uk/digital60/www.digital60.org/birth/program/firstprog.html",
      "source_url": "https://curation.cs.manchester.ac.uk/computer50/www.computer50.org/mark1/new.baby.html",
      "source_title": "University of Manchester: The First Program",
      "revision": "Anniversary reconstruction",
      "license": "University copyright; locator only",
      "redistribution": "locator_only",
      "selection_reason": "It demonstrated a program executing electronically from writable memory.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The original listing was lost; the displayed program was reconstructed, and this excerpt is only algorithmic pseudocode.",
      "tags": [
        "stored-program",
        "machine-code",
        "factorization",
        "memory"
      ]
    },
    {
      "id": "1949-edsac-squares",
      "sort_date": "1949-05-06",
      "year": 1949,
      "date_display": "6 May 1949",
      "date_basis": "executed",
      "date_precision": "day",
      "date_confidence": "high",
      "title": "EDSAC Initial Orders and squares program",
      "organization": "University of Cambridge",
      "problem": "Load symbolic orders from paper tape and provide a practical university computing service.",
      "problem_family": "science_control",
      "domain": "Scientific computing service",
      "language": "Letter-coded EDSAC orders",
      "platform": "EDSAC I",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "transcription",
      "snippet_status": "normalized",
      "snippet": "T0S    clear m[0]\nH2S    load multiplier register\nE6S    jump to main loop\nI0S    read next character",
      "snippet_lines": 4,
      "snippet_locator": "Initial Orders annotated listing, locations 0–8",
      "code_url": "https://www.cl.cam.ac.uk/~mr10/edsacposter.pdf",
      "source_url": "https://www.cl.cam.ac.uk/relics/elog.html",
      "source_title": "University of Cambridge: Start of the EDSAC log",
      "revision": "1949 order code transcription",
      "license": "University copyright; short normalized excerpt",
      "redistribution": "short_excerpt",
      "selection_reason": "EDSAC coupled a compact symbolic order code with the first broadly useful university computing service.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The log is contemporary; the readable order listing is a later annotated transcription.",
      "tags": [
        "edsac",
        "paper-tape",
        "initial-orders",
        "scientific-service"
      ]
    },
    {
      "id": "1951-leo-bakery-valuations",
      "sort_date": "1951-11-29",
      "year": 1951,
      "date_display": "29–30 Nov 1951",
      "date_basis": "executed",
      "date_precision": "range",
      "date_confidence": "high",
      "title": "LEO I Bakery Valuations",
      "organization": "J. Lyons & Co.",
      "problem": "Produce weekly ingredient, stock, production, and profitability valuations in time for operations.",
      "problem_family": "business_data",
      "domain": "Business operations",
      "language": "EDSAC-derived machine instructions and flowcharts",
      "platform": "LEO I",
      "authoring_mode": "encoded_instructions",
      "authoring_level": 2,
      "artifact_status": "conceptual",
      "snippet_status": "unavailable",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Early LEO I program archive listing",
      "code_url": "https://www.computinghistory.org.uk/det/68910/Early-LEO-I-Programs/",
      "source_url": "https://www.sciencemuseum.org.uk/objects-and-stories/meet-leo-worlds-first-business-computer",
      "source_title": "Science Museum: Meet LEO",
      "revision": "",
      "license": "Museum material; no open code license located",
      "redistribution": "locator_only",
      "selection_reason": "It was the first routine, time-sensitive office application on a stored-program computer.",
      "provenance_tier": "B",
      "confidence": "medium",
      "caveat": "The operational date is strong, but no verified surviving source listing is attached to this row.",
      "tags": [
        "leo",
        "bakery",
        "inventory",
        "business-computing"
      ]
    },
    {
      "id": "1954-mark-i-autocode",
      "sort_date": "1954-03-01",
      "year": 1954,
      "date_display": "Mar 1954",
      "date_basis": "released",
      "date_precision": "month",
      "date_confidence": "high",
      "title": "Brooker's Mark I Autocode",
      "organization": "University of Manchester",
      "problem": "Let scientists and engineers program without base-32 machine details or manual drum transfers.",
      "problem_family": "science_control",
      "domain": "Scientific and industrial computing",
      "language": "Mark I Autocode",
      "platform": "Ferranti Mark I",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "transcription",
      "snippet_status": "illustrative",
      "snippet": "n = 100\ns = 0\n1) s = s + n\nn = n - 1\njnz 1",
      "snippet_lines": 5,
      "snippet_locator": "Historical Autocode descriptions and examples",
      "code_url": "https://www.computerconservationsociety.org/resurrection/res45.htm",
      "source_url": "https://curation.cs.manchester.ac.uk/computer50/www.computer50.org/mark1/gethomas/manchester_autocodes.html",
      "source_title": "University of Manchester: Manchester Autocodes",
      "revision": "",
      "license": "Historical pages; reuse rights unclear",
      "redistribution": "locator_only",
      "selection_reason": "It made a high-level, diagnostic programming system usable by a real scientific and industrial community.",
      "provenance_tier": "B",
      "confidence": "high",
      "caveat": "The excerpt is illustrative Autocode-like notation, not a verified historical listing.",
      "tags": [
        "autocode",
        "compiler",
        "high-level",
        "scientific-computing"
      ]
    },
    {
      "id": "1957-flow-matic-file-match",
      "sort_date": "1957-01-01",
      "year": 1957,
      "date_display": "1957 system; 1958 manual",
      "date_basis": "executed",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "FLOW-MATIC inventory and pricing file match",
      "organization": "Remington Rand / Grace Hopper's team",
      "problem": "Merge inventory and price files and route priced and unpriced business records.",
      "problem_family": "business_data",
      "domain": "Business data processing",
      "language": "FLOW-MATIC",
      "platform": "UNIVAC I–II",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "facsimile",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "FLOW-MATIC Programming System manual sample",
      "code_url": "https://www.bitsavers.org/pdf/univac/flow-matic/U1518_FLOW-MATIC_Programming_System_1958.pdf",
      "source_url": "https://www.computerhistory.org/brochures/doc-4372956d57d94/",
      "source_title": "Computer History Museum: FLOW-MATIC brochure record",
      "revision": "U1518",
      "license": "Sperry Rand copyright; locator only",
      "redistribution": "locator_only",
      "selection_reason": "It made restricted English a practical business-programming interface and directly influenced COBOL.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The system operated in 1957; the surviving manual is dated 1958.",
      "tags": [
        "flow-matic",
        "business-language",
        "files",
        "cobol-influence"
      ]
    },
    {
      "id": "1957-fortran-matrix-inversion",
      "sort_date": "1957-04-01",
      "year": 1957,
      "date_display": "Apr 1957 system; example c. 1958",
      "date_basis": "released",
      "date_precision": "month",
      "date_confidence": "high",
      "title": "FORTRAN and order-5 matrix inversion",
      "organization": "IBM",
      "problem": "Translate algebraic scientific programs into code efficient enough to replace hand optimization.",
      "problem_family": "science_control",
      "domain": "Numerical engineering",
      "language": "FORTRAN I",
      "platform": "IBM 704",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "facsimile",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Order-5 matrix inversion source, assembly, output, and compiler summary",
      "code_url": "https://archive.computerhistory.org/resources/text/Fortran/102679230.05.01.acc.pdf",
      "source_url": "https://softwarepreservation.computerhistory.org/FORTRAN/",
      "source_title": "Computer History Museum: History of FORTRAN and FORTRAN II",
      "revision": "CHM 102679230",
      "license": "Preservation access; no general open code license",
      "redistribution": "locator_only",
      "selection_reason": "FORTRAN showed that readable mathematical notation could compile to near-hand-optimized machine code.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The surviving example is approximately 1958 and should not be described as the first 1957 FORTRAN program.",
      "tags": [
        "fortran",
        "compiler",
        "matrix",
        "optimization"
      ]
    },
    {
      "id": "1959-lisp-eval",
      "sort_date": "1959-01-01",
      "year": 1959,
      "date_display": "1958–1959 work; 1960 paper",
      "date_basis": "approximate",
      "date_precision": "range",
      "date_confidence": "medium",
      "title": "LISP eval and programs-as-data",
      "organization": "MIT / John McCarthy and collaborators",
      "problem": "Represent and manipulate symbolic expressions, recursive procedures, and programs themselves as data.",
      "problem_family": "ai_automation",
      "domain": "Artificial intelligence",
      "language": "LISP",
      "platform": "IBM 704",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "conceptual",
      "snippet_status": "illustrative",
      "snippet": "(defun eval-expr (x env)\n  (cond ((atom x) (lookup x env))\n        (t (apply (car x)\n                  (mapcar eval-expr (cdr x))))))",
      "snippet_lines": 4,
      "snippet_locator": "1959 compiler memo and later LISP 1.5 materials",
      "code_url": "https://softwarepreservation.computerhistory.org/LISP/lisp1.5/Anonymous-The_Compiler_in_Lisp-1959.pdf",
      "source_url": "https://softwarepreservation.computerhistory.org/LISP/",
      "source_title": "Computer History Museum: LISP History",
      "revision": "",
      "license": "Historical archive; no open license for scan",
      "redistribution": "locator_only",
      "selection_reason": "LISP made recursive symbolic structures and the code/data symmetry central programming ideas.",
      "provenance_tier": "B",
      "confidence": "medium",
      "caveat": "The chronology spans 1958–1960 and the excerpt is modern illustrative Lisp, not a surviving 1959 listing.",
      "tags": [
        "lisp",
        "eval",
        "symbolic-ai",
        "programs-as-data"
      ]
    },
    {
      "id": "1960-cobol-cross-vendor-demo",
      "sort_date": "1960-12-06",
      "year": 1960,
      "date_display": "6 Dec 1960",
      "date_basis": "executed",
      "date_precision": "day",
      "date_confidence": "high",
      "title": "Cross-vendor COBOL validation program",
      "organization": "CODASYL / RCA / Remington Rand",
      "problem": "Run the same business-data program on incompatible manufacturers' computers.",
      "problem_family": "business_data",
      "domain": "Portable business computing",
      "language": "COBOL",
      "platform": "RCA 501 and UNIVAC II",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "facsimile",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Contemporary compile/run printout",
      "code_url": "https://americanhistory.si.edu/collections/nmah_1389374",
      "source_url": "https://archive.computerhistory.org/resources/access/text/2017/10/102639620-05-01-acc.pdf",
      "source_title": "Computer History Museum: COBOL cross-vendor demonstration history",
      "revision": "NMAH 1389374",
      "license": "Smithsonian usage conditions; code rights unclear",
      "redistribution": "locator_only",
      "selection_reason": "It demonstrated machine-independent business software across competing hardware vendors.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The surviving artifact is a program printout, not compiler source.",
      "tags": [
        "cobol",
        "portability",
        "business",
        "standardization"
      ]
    },
    {
      "id": "1961-ctss-time-sharing",
      "sort_date": "1961-11-01",
      "year": 1961,
      "date_display": "Nov 1961 system; 1970–1972 source snapshots",
      "date_basis": "executed",
      "date_precision": "month",
      "date_confidence": "high",
      "title": "CTSS supervisor and interactive commands",
      "organization": "MIT",
      "problem": "Let many people edit, run, and manage files interactively instead of waiting for batch jobs.",
      "problem_family": "systems_infrastructure",
      "domain": "Operating systems",
      "language": "FAP and MAD",
      "platform": "IBM 7094",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "COMC8G.txt and CTSS source archive",
      "code_url": "https://people.csail.mit.edu/saltzer/Multics/CTSS-Documents/ctss-source-files/",
      "source_url": "https://www.csail.mit.edu/ctss-documents",
      "source_title": "MIT CSAIL: CTSS documents",
      "revision": "MIT8B3 / MIT8C0 backup snapshots",
      "license": "Archival source; reuse terms unclear",
      "redistribution": "locator_only",
      "selection_reason": "CTSS established interactive login, per-user files, editing, and a lineage leading to modern shared computing.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The readable source snapshots date from 1970–1972, not the initial 1961 system.",
      "tags": [
        "ctss",
        "time-sharing",
        "interactive",
        "operating-system"
      ]
    },
    {
      "id": "1963-sketchpad",
      "sort_date": "1963-01-01",
      "year": 1963,
      "date_display": "1963",
      "date_basis": "published",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Sketchpad",
      "organization": "MIT Lincoln Laboratory / Ivan Sutherland",
      "problem": "Create, edit, copy, and geometrically constrain engineering drawings directly on a display.",
      "problem_family": "interaction_applications",
      "domain": "CAD and interactive graphics",
      "language": "TX-2 assembly",
      "platform": "Lincoln TX-2",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "facsimile",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Scanned TX-2 source listing",
      "code_url": "https://www.computerhistory.org/collections/catalog/102726903",
      "source_url": "https://computerhistory.org/blog/the-remarkable-ivan-sutherland/",
      "source_title": "Computer History Museum: The Remarkable Ivan Sutherland",
      "revision": "CHM 102726903",
      "license": "Archival material; reuse rights unclear",
      "redistribution": "locator_only",
      "selection_reason": "It introduced direct manipulation, constraints, instances, and core ideas of computer-aided design.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The code is assembly even though the user-facing interaction was graphical.",
      "tags": [
        "sketchpad",
        "cad",
        "graphics",
        "direct-manipulation"
      ]
    },
    {
      "id": "1965-eliza-doctor",
      "sort_date": "1965-01-01",
      "year": 1965,
      "date_display": "1965",
      "date_basis": "authored",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Original ELIZA and DOCTOR script",
      "organization": "MIT / Joseph Weizenbaum",
      "problem": "Build a reusable pattern-transformation engine and configure a conversational persona separately.",
      "problem_family": "ai_automation",
      "domain": "Conversational AI and HCI",
      "language": "MAD-SLIP and script rules",
      "platform": "MIT CTSS / IBM 7094",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "transcription",
      "snippet_status": "illustrative",
      "snippet": "keyword: MOTHER\ntransform: * MOTHER *\nreply: TELL ME MORE ABOUT YOUR FAMILY\notherwise: PLEASE GO ON",
      "snippet_lines": 4,
      "snippet_locator": "Original scan and MAD-SLIP transcription",
      "code_url": "https://github.com/jeffshrager/elizagen.org/blob/master/1965_Weizenbaum_MAD-SLIP/MAD-SLIP_transcription.txt",
      "source_url": "https://sites.google.com/view/elizagen-org/original-eliza",
      "source_title": "ELIZA Archaeology: Original ELIZA",
      "revision": "1965 scan transcription",
      "license": "CC0 with estate permission",
      "redistribution": "public_domain",
      "selection_reason": "ELIZA separated a general interpreter from a domain script and made conversational behavior inspectable as rules.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The included snippet illustrates the rule shape; use the scan/transcription for exact text.",
      "tags": [
        "eliza",
        "conversation",
        "pattern-matching",
        "script"
      ]
    },
    {
      "id": "1968-bcpl-compiler",
      "sort_date": "1968-02-01",
      "year": 1968,
      "date_display": "1967–Jun 1968",
      "date_basis": "authored",
      "date_precision": "range",
      "date_confidence": "high",
      "title": "BCPL compiler on CTSS",
      "organization": "University of Cambridge / MIT",
      "problem": "Build compact, portable systems software without rewriting an entire compiler for each machine.",
      "problem_family": "systems_infrastructure",
      "domain": "Compilers and system software",
      "language": "BCPL and assembly",
      "platform": "MIT CTSS",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "reconstruction",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Reconstruction from 1968 printed listing",
      "code_url": "https://softwarepreservation.computerhistory.org/BCPL/",
      "source_url": "https://softwarepreservation.computerhistory.org/BCPL/",
      "source_title": "Computer History Museum: BCPL historical archive",
      "revision": "1968 listing reconstruction",
      "license": "Verify archive README before excerpting",
      "redistribution": "locator_only",
      "selection_reason": "BCPL made compact self-hosting systems code portable and directly influenced B and C.",
      "provenance_tier": "B",
      "confidence": "high",
      "caveat": "The usable tree is explicitly a reconstruction from a contemporary listing.",
      "tags": [
        "bcpl",
        "compiler",
        "systems-language",
        "portability"
      ]
    },
    {
      "id": "1969-apollo-guidance-computer",
      "sort_date": "1969-07-20",
      "year": 1969,
      "date_display": "20 Jul 1969 mission",
      "date_basis": "executed",
      "date_precision": "day",
      "date_confidence": "high",
      "title": "Apollo 11 Guidance Computer flight software",
      "organization": "MIT Instrumentation Laboratory / NASA",
      "problem": "Perform real-time guidance, navigation, landing, restart, and alarm handling in extremely limited memory.",
      "problem_family": "science_control",
      "domain": "Safety-critical aerospace",
      "language": "AGC assembly and interpretive code",
      "platform": "Apollo Guidance Computer",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "transcription",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Luminary099/ALARM_AND_ABORT.agc",
      "code_url": "https://github.com/chrislgarry/Apollo-11/blob/master/Luminary099/ALARM_AND_ABORT.agc",
      "source_url": "https://science.nasa.gov/people/margaret-hamilton/",
      "source_title": "NASA Science: Margaret Hamilton and Apollo code",
      "revision": "Luminary 099 / Comanche 055 transcription",
      "license": "Public-domain Apollo materials; volunteer transcription",
      "redistribution": "public_domain",
      "selection_reason": "The software safely prioritized critical real-time work and helped land humans on the Moon.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The GitHub form is a later machine-readable transcription checked against MIT Museum scans.",
      "tags": [
        "apollo",
        "real-time",
        "guidance",
        "safety-critical"
      ]
    },
    {
      "id": "1971-unix-v1-kernel",
      "sort_date": "1971-11-03",
      "year": 1971,
      "date_display": "3 Nov 1971 tree",
      "date_basis": "authored",
      "date_precision": "day",
      "date_confidence": "high",
      "title": "First Edition UNIX kernel",
      "organization": "Bell Labs",
      "problem": "Provide a small multiuser operating system with processes, files, and composable tools.",
      "problem_family": "systems_infrastructure",
      "domain": "Operating systems",
      "language": "PDP-11 assembly",
      "platform": "PDP-11/20",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "transcription",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "V1 tree, e.g. u0.s",
      "code_url": "https://www.tuhs.org/cgi-bin/utree.pl?file=V1",
      "source_url": "https://computerhistory.org/blog/the-earliest-unix-code-an-anniversary-source-code-release/",
      "source_title": "Computer History Museum: The earliest UNIX code",
      "revision": "V1 paper-listing transcription",
      "license": "Historical UNIX source terms; verify before reuse",
      "redistribution": "locator_only",
      "selection_reason": "Its process and filesystem model became the lineage behind Unix, POSIX, Linux, and macOS.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The browsable tree was transcribed/OCRed from a paper listing.",
      "tags": [
        "unix",
        "kernel",
        "processes",
        "filesystem"
      ]
    },
    {
      "id": "1972-smalltalk-72",
      "sort_date": "1972-01-01",
      "year": 1972,
      "date_display": "1972",
      "date_basis": "authored",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Smalltalk-72 live environment",
      "organization": "Xerox PARC",
      "problem": "Create a child-friendly personal computing environment built from message-passing objects.",
      "problem_family": "interaction_applications",
      "domain": "Object-oriented programming and GUI",
      "language": "Smalltalk-72",
      "platform": "Xerox Alto",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Original executable save image in Smalltalk Zoo",
      "code_url": "https://smalltalkzoo.computerhistory.org/",
      "source_url": "https://computerhistory.org/blog/introducing-the-smalltalk-zoo-48-years-of-smalltalk-history-at-chm/",
      "source_title": "Computer History Museum: Smalltalk Zoo",
      "revision": "Smalltalk-72 save image",
      "license": "Xerox/PARC archival terms",
      "redistribution": "locator_only",
      "selection_reason": "Smalltalk's pure object model and live graphical environment shaped modern OOP and desktop interfaces.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The primary artifact is a save image, so line-based comparison requires an export or decompilation step.",
      "tags": [
        "smalltalk",
        "objects",
        "live-programming",
        "gui"
      ]
    },
    {
      "id": "1973-arpanet-imp",
      "sort_date": "1973-01-01",
      "year": 1973,
      "date_display": "1973 source version",
      "date_basis": "authored",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "ARPANET IMP packet-switch code",
      "organization": "BBN",
      "problem": "Route and store-and-forward packets reliably between heterogeneous host computers.",
      "problem_family": "systems_infrastructure",
      "domain": "Networking",
      "language": "Modified MIDAS assembly",
      "platform": "Honeywell 516/316 IMP",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "transcription",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Ready-for-assembly listing and original scan",
      "code_url": "https://www.walden-family.com/impcode/",
      "source_url": "https://www.walden-family.com/impcode/",
      "source_title": "ARPANET IMP code preservation project",
      "revision": "1973 IMP source listing",
      "license": "BBN archival copyright unclear",
      "redistribution": "locator_only",
      "selection_reason": "The IMP implemented the router-like packet-switching core of the ARPANET.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "This is the 1973 code, not the original 1969 IMP version.",
      "tags": [
        "arpanet",
        "packet-switching",
        "router",
        "networking"
      ]
    },
    {
      "id": "1975-cpm-pre-version",
      "sort_date": "1975-01-01",
      "year": 1975,
      "date_display": "1975 pre-version; 1976 v1.3",
      "date_basis": "released",
      "date_precision": "range",
      "date_confidence": "high",
      "title": "CP/M pre-version and BIOS boundary",
      "organization": "Digital Research / Gary Kildall",
      "problem": "Make floppy-disk software portable across small Intel microcomputers by isolating device I/O.",
      "problem_family": "systems_infrastructure",
      "domain": "Microcomputer operating systems",
      "language": "Intel 8080 assembly",
      "platform": "Intel 8080 systems",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "facsimile",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "CHM early CP/M source release",
      "code_url": "https://computerhistory.org/blog/early-digital-research-cpm-source-code/",
      "source_url": "https://computerhistory.org/press-releases/cp-m-release/",
      "source_title": "Computer History Museum: Early Digital Research CP/M source",
      "revision": "pre-version / v1.3",
      "license": "Historical noncommercial source license",
      "redistribution": "permission_specific",
      "selection_reason": "CP/M's BIOS separation made a common application platform practical across early microcomputers.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The preserved releases span 1975–1976 and v1.3 includes reverse-engineered annotations.",
      "tags": [
        "cpm",
        "bios",
        "microcomputer",
        "portability"
      ]
    },
    {
      "id": "1982-tex82",
      "sort_date": "1982-01-01",
      "year": 1982,
      "date_display": "1982 canonical rewrite",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "TeX82 literate source",
      "organization": "Donald Knuth / Stanford University",
      "problem": "Produce deterministic, high-quality mathematical and scientific typesetting across machines.",
      "problem_family": "interaction_applications",
      "domain": "Scientific publishing",
      "language": "WEB (Pascal plus literate documentation)",
      "platform": "Portable",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "tex.web; line history in tex82.bug",
      "code_url": "https://mirrors.ctan.org/systems/knuth/dist/tex/tex.web",
      "source_url": "https://cs.stanford.edu/~knuth/abcde.html",
      "source_title": "Donald Knuth: TeX and METAFONT sources",
      "revision": "Canonical TeX82 source with later fixes",
      "license": "Knuth license; unchanged redistribution permitted, modified versions renamed",
      "redistribution": "permission_specific",
      "selection_reason": "TeX transformed technical publishing and made literate programming a durable engineering practice.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The current canonical file includes later sanctioned fixes; date individual lines with tex82.bug.",
      "tags": [
        "tex",
        "typesetting",
        "literate-programming",
        "publishing"
      ]
    },
    {
      "id": "1982-msdos-125",
      "sort_date": "1982-05-01",
      "year": 1982,
      "date_display": "1982",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "MS-DOS 1.25 source",
      "organization": "Microsoft",
      "problem": "Provide a bootable 8086 PC operating system with FAT files and command execution for compatible business computers.",
      "problem_family": "systems_infrastructure",
      "domain": "Personal-computer operating systems",
      "language": "8086 assembly",
      "platform": "IBM-compatible PC",
      "authoring_mode": "symbolic_text",
      "authoring_level": 3,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "v1.25/source",
      "code_url": "https://github.com/microsoft/MS-DOS/tree/main/v1.25/source",
      "source_url": "https://computerhistory.org/blog/microsoft-ms-dos-early-source-code/",
      "source_title": "Computer History Museum: Microsoft MS-DOS early source",
      "revision": "v1.25",
      "license": "MIT",
      "redistribution": "short_excerpt",
      "selection_reason": "It established the dominant PC software platform and API lineage of its era.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "This is MS-DOS 1.25, not IBM PC DOS 1.0.",
      "tags": [
        "ms-dos",
        "pc",
        "fat",
        "operating-system"
      ]
    },
    {
      "id": "1984-postscript",
      "sort_date": "1984-02-29",
      "year": 1984,
      "date_display": "Late Feb 1984 snapshot",
      "date_basis": "authored",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Early Adobe PostScript interpreter",
      "organization": "Adobe",
      "problem": "Describe pages independently of a printer and rasterize the same document on different devices.",
      "problem_family": "business_data",
      "domain": "Desktop publishing and printing",
      "language": "C and PostScript",
      "platform": "Workstation and printer interpreter",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Mostly complete Feb 1984 source snapshot",
      "code_url": "https://computerhistory.org/blog/postscript-a-digital-printing-press/",
      "source_url": "https://computerhistory.org/press-releases/computer-history-museum-makes-adobe-postscripts-source-code-available-to-the-public-as-a-part-of-its-art-of-code-series/",
      "source_title": "Computer History Museum: PostScript source release",
      "revision": "Feb 1984 snapshot",
      "license": "Adobe/CHM research-only historical release",
      "redistribution": "permission_specific",
      "selection_reason": "Device-independent page programs enabled desktop publishing and directly influenced PDF.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "Some low-level graphics modules were excluded from the historical release.",
      "tags": [
        "postscript",
        "printing",
        "page-description",
        "desktop-publishing"
      ]
    },
    {
      "id": "1984-macpaint-quickdraw",
      "sort_date": "1984-04-01",
      "year": 1984,
      "date_display": "Apr 1984 surviving version",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "MacPaint 1.3 and QuickDraw",
      "organization": "Apple",
      "problem": "Deliver responsive bitmap drawing and a graphical desktop within 128 KB of memory.",
      "problem_family": "interaction_applications",
      "domain": "Graphics and productivity",
      "language": "Apple Pascal and Motorola 68000 assembly",
      "platform": "Apple Macintosh",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "MacPaint 1.3 and QuickDraw historical source package",
      "code_url": "https://computerhistory.org/blog/macpaint-and-quickdraw-source-code/",
      "source_url": "https://computerhistory.org/blog/macpaint-and-quickdraw-source-code/",
      "source_title": "Computer History Museum: MacPaint and QuickDraw source",
      "revision": "MacPaint 1.3",
      "license": "Apple/CHM noncommercial historical-source terms",
      "redistribution": "permission_specific",
      "selection_reason": "The application and raster engine became canonical examples of direct-manipulation graphics software.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The released tree is v1.3, not the January 1984 v1.0.",
      "tags": [
        "macpaint",
        "quickdraw",
        "gui",
        "bitmap"
      ]
    },
    {
      "id": "1985-gnu-emacs",
      "sort_date": "1985-03-20",
      "year": 1985,
      "date_display": "20 Mar 1985 public project; v16.56 archived tree",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "GNU Emacs",
      "organization": "GNU Project",
      "problem": "Create a portable, programmable, self-documenting editing and development environment.",
      "problem_family": "interaction_applications",
      "domain": "Developer tooling",
      "language": "C and Emacs Lisp",
      "platform": "Unix workstations",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Archived v16.56 release tree",
      "code_url": "https://softwarepreservation.computerhistory.org/emacs/",
      "source_url": "https://www.gnu.org/software/emacs/history.html",
      "source_title": "GNU: Emacs history",
      "revision": "v16.56 archive",
      "license": "Historical GNU copyleft terms; verify bundled COPYING",
      "redistribution": "permission_specific",
      "selection_reason": "Emacs became an enduring model for an editor whose behavior is a programmable platform.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The v13 public announcement is not itself a complete verified tree; v16.56 is the exact archived artifact.",
      "tags": [
        "emacs",
        "editor",
        "elisp",
        "extensible"
      ]
    },
    {
      "id": "1987-gcc-1",
      "sort_date": "1987-05-23",
      "year": 1987,
      "date_display": "23 May 1987 v1.0",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "GCC 1.0",
      "organization": "GNU Project",
      "problem": "Provide a free, portable, optimizing C compiler capable of bootstrapping a free operating system.",
      "problem_family": "systems_infrastructure",
      "domain": "Compilers and platform engineering",
      "language": "C",
      "platform": "Multiple Unix hosts and targets",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "gcc-0.9 / GCC 1.0 release tarballs",
      "code_url": "https://gcc.gnu.org/pub/gcc/old-releases/gcc-1/gcc-0.9.tar.bz2",
      "source_url": "https://gcc.gnu.org/releases.html",
      "source_title": "GCC release timeline",
      "revision": "GCC 0.9 / 1.0",
      "license": "Historical GNU copying terms in release",
      "redistribution": "permission_specific",
      "selection_reason": "GCC made a free, portable systems toolchain practical and later became core open-source infrastructure.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The exact release predates today's GPLv3; inspect its bundled terms before excerpting.",
      "tags": [
        "gcc",
        "compiler",
        "gnu",
        "toolchain"
      ]
    },
    {
      "id": "1991-worldwideweb-browser",
      "sort_date": "1991-01-01",
      "year": 1991,
      "date_display": "1990–1991; earliest surviving v0.15",
      "date_basis": "authored",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "WorldWideWeb browser/editor",
      "organization": "CERN / Tim Berners-Lee",
      "problem": "Browse and edit a global linked information space in one application.",
      "problem_family": "interaction_applications",
      "domain": "The Web",
      "language": "Objective-C",
      "platform": "NeXTSTEP",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "WorldWideWeb 0.15 restored source",
      "code_url": "https://github.com/cynthia/WorldWideWeb",
      "source_url": "https://worldwideweb.cern.ch/code/",
      "source_title": "CERN: WorldWideWeb source archaeology",
      "revision": "v0.15",
      "license": "CERN released core Web software to the public domain in 1993; inspect bundled extras",
      "redistribution": "public_domain",
      "selection_reason": "It embodied URL, HTTP, HTML, and a read-write Web in a single program.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The lost October 1990 original is not the surviving v0.15 snapshot, which required restoration.",
      "tags": [
        "web",
        "browser",
        "editor",
        "http"
      ]
    },
    {
      "id": "1991-python-091",
      "sort_date": "1991-02-20",
      "year": 1991,
      "date_display": "20 Feb 1991 announcement",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Python 0.9.1",
      "organization": "CWI / Guido van Rossum",
      "problem": "Make rapid systems and application scripting readable, extensible, and productive.",
      "problem_family": "systems_infrastructure",
      "domain": "Programming language productivity",
      "language": "C interpreter and Python library",
      "platform": "Unix-era systems",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "reconstruction",
      "snippet_status": "illustrative",
      "snippet": "def greet(name):\n    print('Hello', name)\n\nfor name in names: greet(name)",
      "snippet_lines": 4,
      "snippet_locator": "Pristine Usenet shar parts; assembled reconstruction",
      "code_url": "https://www.tuhs.org/Usenet/alt.sources.d/1991-February/000557.html",
      "source_url": "https://www.python.org/download/releases/2.6.9/license/",
      "source_title": "Python license history",
      "revision": "0.9.1 shar archive",
      "license": "CWI permissive license",
      "redistribution": "short_excerpt",
      "selection_reason": "Python's readable high-level interface became central to automation, data work, science, and AI.",
      "provenance_tier": "B",
      "confidence": "medium",
      "caveat": "The assembled GitHub tree contains reconstruction tweaks; the included snippet is illustrative modern Python.",
      "tags": [
        "python",
        "scripting",
        "readability",
        "automation"
      ]
    },
    {
      "id": "1991-linux-001",
      "sort_date": "1991-09-17",
      "year": 1991,
      "date_display": "17 Sep 1991 v0.01",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Linux kernel 0.01",
      "organization": "Linus Torvalds",
      "problem": "Build a freely shareable Unix-like kernel for affordable i386 personal computers.",
      "problem_family": "systems_infrastructure",
      "domain": "Operating systems",
      "language": "C and x86 assembly",
      "platform": "i386 AT",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Historic linux-0.01 tree",
      "code_url": "https://kernel.googlesource.com/pub/scm/linux/kernel/git/nico/archive/+/v0.01",
      "source_url": "https://www.kernel.org/pub/linux/kernel/Historic/linux-0.01.tar.gz",
      "source_title": "kernel.org historic Linux archive",
      "revision": "v0.01",
      "license": "Original v0.01 noncommercial redistribution terms",
      "redistribution": "permission_specific",
      "selection_reason": "This small kernel seeded Linux across servers, cloud, Android, and embedded systems.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "Linux 0.01 was not GPL; the GPL switch came later at 0.12.",
      "tags": [
        "linux",
        "kernel",
        "i386",
        "open-source"
      ]
    },
    {
      "id": "1993-ncsa-mosaic",
      "sort_date": "1993-01-23",
      "year": 1993,
      "date_display": "1992–1993",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "NCSA Mosaic for X",
      "organization": "NCSA / University of Illinois",
      "problem": "Make the Web approachable through a graphical browser with inline images.",
      "problem_family": "interaction_applications",
      "domain": "Web browsing",
      "language": "C",
      "platform": "X11/Motif Unix workstations",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "NCSA Mosaic born-digital collection",
      "code_url": "https://archon.library.illinois.edu/archives/?id=9866&p=digitallibrary%2Fdigitalcontent",
      "source_url": "https://www.w3.org/History/1993/WWW/XMosaic/",
      "source_title": "W3C historical XMosaic page",
      "revision": "Collection spans ca. 1992–1995",
      "license": "Historical NCSA terms; not the later OSI NCSA license",
      "redistribution": "permission_specific",
      "selection_reason": "Mosaic popularized graphical Web browsing and seeded the commercial browser ecosystem.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "Pin an individual file timestamp and version before extracting code.",
      "tags": [
        "mosaic",
        "browser",
        "web",
        "inline-images"
      ]
    },
    {
      "id": "1994-postgres-42",
      "sort_date": "1994-06-01",
      "year": 1994,
      "date_display": "Jun 1994 v4.2",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "POSTGRES 4.2",
      "organization": "University of California, Berkeley",
      "problem": "Store complex data using extensible types, rules, and object-relational behavior.",
      "problem_family": "business_data",
      "domain": "Databases",
      "language": "C",
      "platform": "Unix",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "POSTGRES 4.2 source and RCS history",
      "code_url": "https://dsf.berkeley.edu/postgres-v4r2/",
      "source_url": "https://www.postgresql.org/docs/current/history.html",
      "source_title": "PostgreSQL official history",
      "revision": "v4.2",
      "license": "Berkeley/BSD-style terms; verify bundled COPYRIGHT",
      "redistribution": "short_excerpt",
      "selection_reason": "It is a directly preserved ancestor of PostgreSQL and a landmark in extensible database architecture.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "1994 dates this exact v4.2 artifact, not the project's 1986 inception.",
      "tags": [
        "postgres",
        "database",
        "object-relational",
        "extensible"
      ]
    },
    {
      "id": "1998-apache-httpd-130",
      "sort_date": "1998-06-06",
      "year": 1998,
      "date_display": "6 Jun 1998 v1.3.0",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Apache HTTP Server 1.3.0",
      "organization": "Apache Group",
      "problem": "Serve production websites robustly through a portable and modular open infrastructure project.",
      "problem_family": "systems_infrastructure",
      "domain": "Web infrastructure",
      "language": "C",
      "platform": "Unix and ports",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Official apache_1.3.0 source tar",
      "code_url": "https://archive.apache.org/dist/httpd/apache_1.3.0.tar.Z",
      "source_url": "https://httpd.apache.org/ABOUT_APACHE",
      "source_title": "Apache HTTP Server project history",
      "revision": "1.3.0",
      "license": "Apache License 1.0-era terms in tar",
      "redistribution": "short_excerpt",
      "selection_reason": "Apache became the leading Web server and a model for open collaborative infrastructure.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The project began earlier; 1998 is chosen for the directly preserved official 1.3.0 release.",
      "tags": [
        "apache",
        "http",
        "web-server",
        "open-source"
      ]
    },
    {
      "id": "2000-ant-buildxml",
      "sort_date": "2000-07-18",
      "year": 2000,
      "date_display": "18 Jul 2000, Ant 1.1",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Apache Ant build.xml",
      "organization": "Apache Software Foundation",
      "problem": "Make Java builds repeatable and portable instead of encoding them in platform-specific shell commands.",
      "problem_family": "systems_infrastructure",
      "domain": "Build automation",
      "language": "XML build DSL",
      "platform": "JVM",
      "authoring_mode": "declarative_configuration",
      "authoring_level": 6,
      "artifact_status": "original",
      "snippet_status": "exact",
      "snippet": "<project name=\"Ant\" default=\"main\" basedir=\".\">\n  <target name=\"prepare\">\n    <mkdir dir=\"${build.dir}\"/>\n  </target>",
      "snippet_lines": 4,
      "snippet_locator": "build.xml at pinned revision",
      "code_url": "https://github.com/apache/ant/blob/5b11971b96a39668b89d0210ac6e8c00032995a1/build.xml",
      "source_url": "https://ant.apache.org/manual/index.html",
      "source_title": "Apache Ant manual",
      "revision": "5b11971b96a39668b89d0210ac6e8c00032995a1",
      "license": "Apache Software License 1.1-era source",
      "redistribution": "short_excerpt",
      "selection_reason": "Ant popularized a portable, declarative build graph for large JVM projects.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The tag crossed CVS, SVN, and Git migrations; preserve the pinned SHA and migration provenance.",
      "tags": [
        "ant",
        "build",
        "xml",
        "declarative"
      ]
    },
    {
      "id": "2003-memcached-initial",
      "sort_date": "2003-05-27",
      "year": 2003,
      "date_display": "27 May 2003",
      "date_basis": "authored",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "memcached initial server",
      "organization": "LiveJournal / Brad Fitzpatrick",
      "problem": "Relieve database load for high-traffic dynamic web services using a distributed in-memory cache.",
      "problem_family": "systems_infrastructure",
      "domain": "Web-scale caching",
      "language": "C",
      "platform": "Unix and libevent",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "original",
      "snippet_status": "exact",
      "snippet": "enum conn_states {\n    conn_listening,\n    conn_read,\n    conn_write,\n};",
      "snippet_lines": 5,
      "snippet_locator": "memcached.c in initial commit",
      "code_url": "https://github.com/memcached/memcached/blob/32f382b605b4565bddfae5c9d082af3bfd30cf02/memcached.c",
      "source_url": "https://github.com/memcached/memcached",
      "source_title": "Official memcached repository",
      "revision": "32f382b605b4565bddfae5c9d082af3bfd30cf02",
      "license": "BSD-3-Clause",
      "redistribution": "short_excerpt",
      "selection_reason": "memcached made scale-out caching a standard architecture for dynamic Web businesses.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The excerpt shows the event-driven connection state machine, not the entire distributed-cache behavior.",
      "tags": [
        "memcached",
        "cache",
        "scaling",
        "web"
      ]
    },
    {
      "id": "2004-google-mapreduce",
      "sort_date": "2004-12-01",
      "year": 2004,
      "date_display": "OSDI 2004",
      "date_basis": "published",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "MapReduce word-count mapper and reducer",
      "organization": "Google Research",
      "problem": "Process multi-terabyte datasets across commodity clusters while the runtime handles partitioning, scheduling, and failures.",
      "problem_family": "business_data",
      "domain": "Distributed data processing",
      "language": "C++-like pseudocode and runtime API",
      "platform": "Google cluster runtime",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "transcription",
      "snippet_status": "illustrative",
      "snippet": "map(document): emit(word, 1)\n\nreduce(word, counts):\n    emit(word, sum(counts))",
      "snippet_lines": 4,
      "snippet_locator": "Word-count example in MapReduce paper",
      "code_url": "https://research.google/pubs/mapreduce-simplified-data-processing-on-large-clusters/",
      "source_url": "https://research.google/pubs/mapreduce-simplified-data-processing-on-large-clusters/",
      "source_title": "Google Research: MapReduce",
      "revision": "OSDI 2004 paper",
      "license": "Paper copyright; production implementation not released",
      "redistribution": "locator_only",
      "selection_reason": "MapReduce compressed distributed scheduling, communication, and failure recovery into two user-defined functions.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The code is normalized paper pseudocode, not Google's production source.",
      "tags": [
        "mapreduce",
        "distributed",
        "big-data",
        "fault-tolerance"
      ]
    },
    {
      "id": "2005-rails-100",
      "sort_date": "2005-12-13",
      "year": 2005,
      "date_display": "13 Dec 2005, Rails 1.0",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Ruby on Rails 1.0 ActionController",
      "organization": "Ruby on Rails",
      "problem": "Deliver database-backed business Web applications rapidly with conventional structure and less configuration.",
      "problem_family": "interaction_applications",
      "domain": "Web application development",
      "language": "Ruby",
      "platform": "Web MVC framework",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "illustrative",
      "snippet": "class ProductsController < ApplicationController\n  def show\n    @product = Product.find(params[:id])\n  end\nend",
      "snippet_lines": 5,
      "snippet_locator": "ActionController base documentation at v1.0.0",
      "code_url": "https://github.com/rails/rails/blob/v1.0.0/actionpack/lib/action_controller/base.rb",
      "source_url": "https://rubyonrails.org/2005/12/13/rails-1-0-party-like-it-s-one-oh-oh",
      "source_title": "Official Rails 1.0 announcement",
      "revision": "v1.0.0 / 37e0d5d",
      "license": "MIT",
      "redistribution": "short_excerpt",
      "selection_reason": "Rails popularized convention over configuration and a compact full-stack MVC workflow.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The example is framework-idiomatic and illustrative; tag metadata was migrated later.",
      "tags": [
        "rails",
        "mvc",
        "web",
        "convention"
      ]
    },
    {
      "id": "2006-hadoop-010",
      "sort_date": "2006-04-01",
      "year": 2006,
      "date_display": "1 Apr 2006, Hadoop 0.1.0",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Hadoop 0.1.0 WordCount",
      "organization": "Apache Hadoop",
      "problem": "Make distributed storage and MapReduce-style data processing available outside Google.",
      "problem_family": "business_data",
      "domain": "Distributed data platforms",
      "language": "Java",
      "platform": "Commodity clusters",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "src/examples/org/apache/hadoop/examples/WordCount.java in 0.1.0 tar",
      "code_url": "https://archive.apache.org/dist/hadoop/common/",
      "source_url": "https://hadoop.apache.org/docs/r3.1.1/hadoop-project-dist/hadoop-common/release/0.1.0/CHANGES.0.1.0.html",
      "source_title": "Apache Hadoop 0.1.0 changelog",
      "revision": "hadoop-0.1.0.tar.gz",
      "license": "Apache-2.0",
      "redistribution": "short_excerpt",
      "selection_reason": "Hadoop operationalized the distributed-file and MapReduce pattern for the broader software industry.",
      "provenance_tier": "A",
      "confidence": "medium",
      "caveat": "Confirm the exact example path and record the source tar checksum during ingestion.",
      "tags": [
        "hadoop",
        "mapreduce",
        "distributed-storage",
        "big-data"
      ]
    },
    {
      "id": "2008-android-helloactivity",
      "sort_date": "2008-12-19",
      "year": 2008,
      "date_display": "19 Dec 2008 public commit; file copyright 2007",
      "date_basis": "authored",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Android HelloActivity",
      "organization": "Android Open Source Project",
      "problem": "Package a touch-oriented mobile interface around an application lifecycle and system-managed components.",
      "problem_family": "interaction_applications",
      "domain": "Mobile applications",
      "language": "Java",
      "platform": "Android",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "exact",
      "snippet": "public class HelloActivity extends Activity {\n  public void onCreate(Bundle savedInstanceState) {\n    super.onCreate(savedInstanceState);\n    setContentView(view);",
      "snippet_lines": 4,
      "snippet_locator": "HelloActivity.java at pinned AOSP revision",
      "code_url": "https://github.com/aosp-mirror/platform_development/blob/303a8c73c94b56e118680ce4b2e081557b1d876a/samples/HelloActivity/src/com/example/android/helloactivity/HelloActivity.java",
      "source_url": "https://android.googlesource.com/platform/development/",
      "source_title": "Android Open Source Project development tree",
      "revision": "303a8c73c94b56e118680ce4b2e081557b1d876a",
      "license": "Apache-2.0",
      "redistribution": "short_excerpt",
      "selection_reason": "The tiny lifecycle example represents the framework-managed mobile application model that reached billions of devices.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The public commit is from 2008, while the source header dates the file to 2007.",
      "tags": [
        "android",
        "mobile",
        "activity",
        "lifecycle"
      ]
    },
    {
      "id": "2009-bitcoin-015",
      "sort_date": "2009-01-08",
      "year": 2009,
      "date_display": "2009 v0.1 lineage",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Bitcoin v0.1 transaction and block processing",
      "organization": "Bitcoin / Satoshi Nakamoto",
      "problem": "Transfer value peer-to-peer without a central clearing party while preventing double spending.",
      "problem_family": "business_data",
      "domain": "Distributed ledger and payments",
      "language": "C++",
      "platform": "Peer-to-peer network",
      "authoring_mode": "high_level_language",
      "authoring_level": 4,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "main.cpp at v0.1.5 historical revision",
      "code_url": "https://github.com/bitcoin/bitcoin/blob/8dca7864f793072701f810e4c5ea12a6e1087085/main.cpp",
      "source_url": "https://bitcoin.org/bitcoin.pdf",
      "source_title": "Bitcoin: A Peer-to-Peer Electronic Cash System",
      "revision": "8dca7864f793072701f810e4c5ea12a6e1087085",
      "license": "MIT",
      "redistribution": "short_excerpt",
      "selection_reason": "The source combined proof of work, replicated validation, and transaction rules into an executable payment network.",
      "provenance_tier": "A",
      "confidence": "medium",
      "caveat": "Historical import metadata can differ from original authorship time; preserve release files and commit timestamps separately.",
      "tags": [
        "bitcoin",
        "payments",
        "blockchain",
        "peer-to-peer"
      ]
    },
    {
      "id": "2010-node-020",
      "sort_date": "2010-08-20",
      "year": 2010,
      "date_display": "20 Aug 2010, Node 0.2.0",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Node.js HTTP server example",
      "organization": "Node.js",
      "problem": "Handle many concurrent network connections with nonblocking I/O in a mainstream Web language.",
      "problem_family": "systems_infrastructure",
      "domain": "Server-side Web development",
      "language": "JavaScript",
      "platform": "V8 and event-loop runtime",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "normalized",
      "snippet": "var http = require('http');\nhttp.createServer(function (req, res) {\n  res.end('Hello World\\n');\n}).listen(8124);",
      "snippet_lines": 4,
      "snippet_locator": "README.md at v0.2.0",
      "code_url": "https://github.com/nodejs/node/blob/v0.2.0/README.md",
      "source_url": "https://nodejs.org/en/blog/release/v0.2.0",
      "source_title": "Official Node.js v0.2.0 release",
      "revision": "v0.2.0",
      "license": "MIT plus bundled dependency notices",
      "redistribution": "short_excerpt",
      "selection_reason": "Node made event-loop concurrency legible in a few lines and moved JavaScript to the server.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The excerpt is whitespace-normalized from the release README example.",
      "tags": [
        "nodejs",
        "javascript",
        "nonblocking-io",
        "server"
      ]
    },
    {
      "id": "2013-docker-040",
      "sort_date": "2013-06-03",
      "year": 2013,
      "date_display": "3 Jun 2013, Docker 0.4.0",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Docker 0.4.0 Dockerfile builder",
      "organization": "Docker / dotCloud",
      "problem": "Package an application and its runtime dependencies reproducibly for movement between development and production.",
      "problem_family": "systems_infrastructure",
      "domain": "Containers and software delivery",
      "language": "Dockerfile DSL and Go",
      "platform": "Linux containers",
      "authoring_mode": "declarative_configuration",
      "authoring_level": 6,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "Dockerfile at v0.4.0",
      "code_url": "https://github.com/moby/moby/blob/a8ae398bf52e97148ee7bd0d5868de2e15bd297f/Dockerfile",
      "source_url": "https://docs.docker.com/engine/release-notes/prior-releases/",
      "source_title": "Docker prior release notes",
      "revision": "a8ae398bf52e97148ee7bd0d5868de2e15bd297f",
      "license": "Apache-2.0 at revision",
      "redistribution": "short_excerpt",
      "selection_reason": "The Dockerfile made deployable environments versioned, repeatable artifacts and tightened dev/prod parity.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "This pins a directly preserved release, not the first internal container prototype.",
      "tags": [
        "docker",
        "containers",
        "deployment",
        "dockerfile"
      ]
    },
    {
      "id": "2014-kubernetes-pod",
      "sort_date": "2014-06-06",
      "year": 2014,
      "date_display": "6 Jun 2014 public repo; 29 Oct pinned sample",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Kubernetes Pod desired-state declaration",
      "organization": "Google / Kubernetes",
      "problem": "Schedule, heal, expose, and scale containerized applications across a machine fleet.",
      "problem_family": "systems_infrastructure",
      "domain": "Cluster orchestration",
      "language": "JSON declarative API and Go control plane",
      "platform": "Container clusters",
      "authoring_mode": "declarative_configuration",
      "authoring_level": 6,
      "artifact_status": "original",
      "snippet_status": "illustrative",
      "snippet": "{\n  \"id\": \"redis-master\",\n  \"desiredState\": { \"manifest\": { ... } }\n}",
      "snippet_lines": 4,
      "snippet_locator": "api/examples/pod.json or guestbook redis-master.json at v0.4.2",
      "code_url": "https://github.com/kubernetes/kubernetes/blob/2902951c445da6e6ead9e406b3c8e19bc28c2b4a/api/examples/pod.json",
      "source_url": "https://kubernetes.io/blog/2018/06/06/4-years-of-k8s/",
      "source_title": "Kubernetes: Four years of Kubernetes",
      "revision": "2902951c445da6e6ead9e406b3c8e19bc28c2b4a",
      "license": "Apache-2.0",
      "redistribution": "short_excerpt",
      "selection_reason": "Kubernetes turned cluster operations into a reconciled desired-state API.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The snippet is a compact illustration of the early schema; distinguish the first-public date from the pinned sample date.",
      "tags": [
        "kubernetes",
        "orchestration",
        "desired-state",
        "containers"
      ]
    },
    {
      "id": "2014-terraform-010",
      "sort_date": "2014-07-28",
      "year": 2014,
      "date_display": "2014, Terraform 0.1.0",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Terraform 0.1.0 infrastructure declaration",
      "organization": "HashiCorp",
      "problem": "Version, plan, graph, and safely change cloud infrastructure across providers.",
      "problem_family": "systems_infrastructure",
      "domain": "Infrastructure as code",
      "language": "Terraform HCL",
      "platform": "Multi-cloud",
      "authoring_mode": "declarative_configuration",
      "authoring_level": 6,
      "artifact_status": "original",
      "snippet_status": "exact",
      "snippet": "resource \"aws_instance\" \"foo\" {\n}\nresource \"aws_instance\" \"bar\" {\n}",
      "snippet_lines": 4,
      "snippet_locator": "terraform/test-fixtures/apply-minimal/main.tf at v0.1.0",
      "code_url": "https://github.com/hashicorp/terraform/blob/v0.1.0/terraform/test-fixtures/apply-minimal/main.tf",
      "source_url": "https://github.com/hashicorp/terraform/blob/v0.1.0/README.md",
      "source_title": "Terraform v0.1.0 README",
      "revision": "fd88908 / v0.1.0",
      "license": "MPL-2.0 at this historical revision",
      "redistribution": "short_excerpt",
      "selection_reason": "Terraform made vendor-neutral desired infrastructure, a change plan, and graph-based execution mainstream.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "Do not apply the project's later Business Source License retroactively to this MPL-licensed revision.",
      "tags": [
        "terraform",
        "infrastructure-as-code",
        "cloud",
        "plan"
      ]
    },
    {
      "id": "2015-tensorflow-050",
      "sort_date": "2015-11-09",
      "year": 2015,
      "date_display": "9 Nov 2015 initial public release",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "TensorFlow 0.5.0 computation graph",
      "organization": "Google",
      "problem": "Express and execute machine-learning computation graphs across CPUs, GPUs, and distributed runtimes.",
      "problem_family": "ai_automation",
      "domain": "Deep-learning infrastructure",
      "language": "Python API and C++ runtime",
      "platform": "CPU/GPU/distributed systems",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "MNIST-from-scratch notebook cells at initial release",
      "code_url": "https://github.com/tensorflow/tensorflow/blob/b2dc60eaa9c00421293b87824a2047fdcf6fa331/tensorflow/tools/docker/notebooks/3_mnist_from_scratch.ipynb",
      "source_url": "https://blog.tensorflow.org/2022/10/building-the-future-of-tensorflow.html",
      "source_title": "TensorFlow blog: Building the future of TensorFlow",
      "revision": "b2dc60eaa9c00421293b87824a2047fdcf6fa331 / 0.5.0",
      "license": "Apache-2.0",
      "redistribution": "short_excerpt",
      "selection_reason": "TensorFlow mainstreamed production-scale differentiable programming and accelerator-backed deep learning.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "Notebook extraction should preserve cell index or cell ID as the snippet locator.",
      "tags": [
        "tensorflow",
        "deep-learning",
        "computation-graph",
        "gpu"
      ]
    },
    {
      "id": "2017-transformer-tensor2tensor",
      "sort_date": "2017-06-15",
      "year": 2017,
      "date_display": "15 Jun 2017",
      "date_basis": "authored",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "Tensor2Tensor Transformer implementation",
      "organization": "Google Brain",
      "problem": "Model and translate sequences using attention without recurrent networks.",
      "problem_family": "ai_automation",
      "domain": "Machine learning architecture",
      "language": "Python and TensorFlow",
      "platform": "Accelerators",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "tensor2tensor/models/transformer.py at initial push",
      "code_url": "https://github.com/tensorflow/tensor2tensor/blob/3d9c62f2aca9492db5c22676416974005b9dcbae/tensor2tensor/models/transformer.py",
      "source_url": "https://arxiv.org/abs/1706.03762",
      "source_title": "Attention Is All You Need",
      "revision": "3d9c62f2aca9492db5c22676416974005b9dcbae",
      "license": "Apache-2.0",
      "redistribution": "short_excerpt",
      "selection_reason": "This implementation embodied the architecture that later underpinned foundation models and LLMs.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The paper and code dates are close but should remain separate provenance fields in a full corpus.",
      "tags": [
        "transformer",
        "attention",
        "sequence-model",
        "foundation-model"
      ]
    },
    {
      "id": "2018-bert-initial",
      "sort_date": "2018-10-31",
      "year": 2018,
      "date_display": "31 Oct 2018",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "BERT run_classifier.py",
      "organization": "Google Research",
      "problem": "Adapt one pretrained bidirectional language representation to many classification tasks with small task heads.",
      "problem_family": "ai_automation",
      "domain": "Natural-language processing",
      "language": "Python and TensorFlow",
      "platform": "GPU/TPU",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "run_classifier.py at initial release",
      "code_url": "https://github.com/google-research/bert/blob/fe354751d7de010f60d362ae8d9343849ec39456/run_classifier.py",
      "source_url": "https://research.google/blog/open-sourcing-bert-state-of-the-art-pre-training-for-natural-language-processing/",
      "source_title": "Google Research: Open-sourcing BERT",
      "revision": "fe354751d7de010f60d362ae8d9343849ec39456",
      "license": "Apache-2.0",
      "redistribution": "short_excerpt",
      "selection_reason": "BERT standardized the pretrain-then-finetune workflow across practical NLP tasks.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The source shows a task-specific fine-tuning pipeline, not the entire pretraining system.",
      "tags": [
        "bert",
        "pretraining",
        "nlp",
        "fine-tuning"
      ]
    },
    {
      "id": "2019-openai-gpt2",
      "sort_date": "2019-02-11",
      "year": 2019,
      "date_display": "11 Feb 2019",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "OpenAI GPT-2 model implementation",
      "organization": "OpenAI",
      "problem": "Use one pretrained Transformer as a general language model whose behavior can be conditioned by text context.",
      "problem_family": "ai_automation",
      "domain": "Generative language models",
      "language": "Python and TensorFlow",
      "platform": "Accelerators",
      "authoring_mode": "library_framework",
      "authoring_level": 5,
      "artifact_status": "original",
      "snippet_status": "locator_only",
      "snippet": "",
      "snippet_lines": 0,
      "snippet_locator": "src/model.py at initial commit",
      "code_url": "https://github.com/openai/gpt-2/blob/c2dae27c1029770cea40978813f17a5fd545b883/src/model.py",
      "source_url": "https://developers.openai.com/",
      "source_title": "OpenAI Developers",
      "revision": "c2dae27c1029770cea40978813f17a5fd545b883",
      "license": "MIT",
      "redistribution": "short_excerpt",
      "selection_reason": "GPT-2 made the shift from task-specific NLP systems toward prompt-conditioned general text generation visible in code.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "Use the pinned repository and original release material for detailed model provenance.",
      "tags": [
        "gpt2",
        "language-model",
        "generation",
        "prompting"
      ]
    },
    {
      "id": "2021-openai-humaneval",
      "sort_date": "2021-07-08",
      "year": 2021,
      "date_display": "8 Jul 2021",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "HumanEval natural-language-to-code tasks",
      "organization": "OpenAI",
      "problem": "Measure whether a model can turn a Python signature and docstring into working code that passes executable tests.",
      "problem_family": "ai_automation",
      "domain": "Code generation and evaluation",
      "language": "Natural-language docstrings, Python, and tests",
      "platform": "Codex research evaluation",
      "authoring_mode": "natural_language",
      "authoring_level": 7,
      "artifact_status": "original",
      "snippet_status": "normalized",
      "snippet": "def has_close_elements(numbers, threshold):\n    \"\"\"Return True if any two numbers are closer\n    than the threshold.\"\"\"",
      "snippet_lines": 3,
      "snippet_locator": "HumanEval task corpus and evaluator",
      "code_url": "https://github.com/openai/human-eval/blob/463c980b59e818ace59f6f9803cd92c749ceae61/human_eval/data.py",
      "source_url": "https://arxiv.org/abs/2107.03374",
      "source_title": "Evaluating Large Language Models Trained on Code",
      "revision": "463c980b59e818ace59f6f9803cd92c749ceae61",
      "license": "MIT",
      "redistribution": "short_excerpt",
      "selection_reason": "HumanEval made natural-language programming measurable through executable tests rather than visual plausibility.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "Generated completions are untrusted code; the evaluator warns that execution requires a robust sandbox.",
      "tags": [
        "humaneval",
        "codex",
        "code-generation",
        "evaluation"
      ]
    },
    {
      "id": "2023-autogen-020",
      "sort_date": "2023-11-24",
      "year": 2023,
      "date_display": "24 Nov 2023, AutoGen 0.2.0",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "AutoGen multi-agent conversation",
      "organization": "Microsoft Research",
      "problem": "Coordinate LLM agents, humans, and code execution to complete multi-step work.",
      "problem_family": "ai_automation",
      "domain": "Agent orchestration",
      "language": "Python plus natural-language agent instructions",
      "platform": "LLM APIs and tool execution",
      "authoring_mode": "natural_language",
      "authoring_level": 7,
      "artifact_status": "original",
      "snippet_status": "illustrative",
      "snippet": "assistant = AssistantAgent(name=\"assistant\")\nuser = UserProxyAgent(name=\"user\")\nuser.initiate_chat(assistant,\n    message=\"Analyze the data and write the code.\")",
      "snippet_lines": 4,
      "snippet_locator": "README agent-pair example at v0.2.0",
      "code_url": "https://github.com/microsoft/autogen/blob/d8d05279a5e4bde4bc618f2c0117e78b1d135312/README.md",
      "source_url": "https://www.microsoft.com/en-us/research/project/autogen/",
      "source_title": "Microsoft Research: AutoGen",
      "revision": "d8d05279a5e4bde4bc618f2c0117e78b1d135312",
      "license": "MIT at revision",
      "redistribution": "short_excerpt",
      "selection_reason": "AutoGen codified conversation, handoff, and tool execution as a programmable multi-agent workflow.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The excerpt is compact illustrative usage; AutoGen's spin-off date differs from the older FLAML repository history.",
      "tags": [
        "autogen",
        "agents",
        "conversation",
        "tools"
      ]
    },
    {
      "id": "2024-openai-swarm-handoff",
      "sort_date": "2024-10-10",
      "year": 2024,
      "date_display": "10 Oct 2024",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "OpenAI Swarm agent handoff",
      "organization": "OpenAI",
      "problem": "Express a lightweight transfer of control between specialized language-model agents.",
      "problem_family": "ai_automation",
      "domain": "Agent orchestration",
      "language": "Python plus natural-language instructions",
      "platform": "OpenAI API",
      "authoring_mode": "natural_language",
      "authoring_level": 7,
      "artifact_status": "original",
      "snippet_status": "exact",
      "snippet": "def transfer_to_agent_b():\n    return agent_b\n\nagent_a = Agent(functions=[transfer_to_agent_b])",
      "snippet_lines": 4,
      "snippet_locator": "examples/basic/agent_handoff.py at initial commit",
      "code_url": "https://github.com/openai/swarm/blob/e5eabc6f0bdc5193d8342b6d0cafa1073632870a/examples/basic/agent_handoff.py",
      "source_url": "https://developers.openai.com/",
      "source_title": "OpenAI Developers",
      "revision": "e5eabc6f0bdc5193d8342b6d0cafa1073632870a",
      "license": "MIT",
      "redistribution": "short_excerpt",
      "selection_reason": "Swarm made an agent handoff visible as a tiny programming primitive and helped clarify later agent SDK patterns.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The repository describes Swarm as educational and experimental; it was superseded by production SDK work.",
      "tags": [
        "swarm",
        "agents",
        "handoff",
        "orchestration"
      ]
    },
    {
      "id": "2025-openai-agents-sdk",
      "sort_date": "2025-03-11",
      "year": 2025,
      "date_display": "11 Mar 2025",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "OpenAI Agents SDK hello-world agent",
      "organization": "OpenAI",
      "problem": "Run an instruction-driven agent with tools, handoffs, guardrails, sessions, and tracing.",
      "problem_family": "ai_automation",
      "domain": "Production agent applications",
      "language": "Python plus natural-language instructions",
      "platform": "OpenAI API",
      "authoring_mode": "natural_language",
      "authoring_level": 7,
      "artifact_status": "original",
      "snippet_status": "exact",
      "snippet": "agent = Agent(\n  name=\"Assistant\",\n  instructions=\"You are a helpful assistant\")\nresult = Runner.run_sync(agent, \"Write a haiku about recursion\")",
      "snippet_lines": 4,
      "snippet_locator": "examples/basic/hello_world.py at initial commit",
      "code_url": "https://github.com/openai/openai-agents-python/blob/aaec57a426b35079480d96cef473a7494c95b0ad/examples/basic/hello_world.py",
      "source_url": "https://developers.openai.com/",
      "source_title": "OpenAI Developers: Agents and tools documentation",
      "revision": "aaec57a426b35079480d96cef473a7494c95b0ad",
      "license": "MIT",
      "redistribution": "short_excerpt",
      "selection_reason": "Natural-language instructions become a first-class executable control surface alongside typed tools and runtime safeguards.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The two-line experience hides a substantial model, tool, state, and tracing runtime.",
      "tags": [
        "agents-sdk",
        "instructions",
        "tools",
        "guardrails"
      ]
    },
    {
      "id": "2026-openai-symphony-workflow",
      "sort_date": "2026-04-27",
      "year": 2026,
      "date_display": "4 Mar source; 27 Apr 2026 announcement",
      "date_basis": "released",
      "date_precision": "year",
      "date_confidence": "high",
      "title": "OpenAI Symphony WORKFLOW.md",
      "organization": "OpenAI",
      "problem": "Turn issue-tracker work into isolated, autonomous coding-agent runs governed by versioned repository policy.",
      "problem_family": "ai_automation",
      "domain": "Autonomous software engineering",
      "language": "Markdown natural-language policy and YAML configuration",
      "platform": "Codex app-server and Symphony orchestrator",
      "authoring_mode": "natural_language",
      "authoring_level": 7,
      "artifact_status": "prompt",
      "snippet_status": "illustrative",
      "snippet": "Pick one ready issue.\nCreate an isolated workspace.\nImplement the change and run the required checks.\nReport evidence, caveats, and the handoff.",
      "snippet_lines": 4,
      "snippet_locator": "elixir/WORKFLOW.md at initial revision; SPEC.md for semantics",
      "code_url": "https://github.com/openai/symphony/blob/fa75ec68c23fbf773c43197da618f88684d8d3f6/elixir/WORKFLOW.md",
      "source_url": "https://developers.openai.com/",
      "source_title": "OpenAI Developers: Codex workflows and engineering automation",
      "revision": "fa75ec68c23fbf773c43197da618f88684d8d3f6",
      "license": "Apache-2.0",
      "redistribution": "short_excerpt",
      "selection_reason": "A repository-versioned prose policy now governs workspace setup, agent work, validation, and handoff—the clearest endpoint of natural language as program in this seed.",
      "provenance_tier": "A",
      "confidence": "high",
      "caveat": "The excerpt is an illustrative condensation; the project is an engineering preview intended for trusted environments.",
      "tags": [
        "symphony",
        "codex",
        "workflow",
        "natural-language-programming"
      ]
    }
  ],
  "record_count": 57,
  "year_count": 50,
  "quality_summary": {
    "provenance_tier": {
      "A": 48,
      "B": 9
    },
    "artifact_status": {
      "conceptual": 6,
      "facsimile": 7,
      "original": 31,
      "prompt": 1,
      "reconstruction": 4,
      "transcription": 8
    },
    "snippet_status": {
      "exact": 6,
      "illustrative": 14,
      "locator_only": 32,
      "normalized": 3,
      "unavailable": 2
    },
    "problem_family": {
      "ai_automation": 11,
      "business_data": 9,
      "interaction_applications": 9,
      "science_control": 9,
      "systems_infrastructure": 19
    },
    "authoring_mode": {
      "declarative_configuration": 4,
      "encoded_instructions": 4,
      "high_level_language": 16,
      "library_framework": 14,
      "natural_language": 5,
      "physical_configuration": 3,
      "symbolic_text": 11
    }
  }
}