ispec/.intent at master · johnfire/ispec · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
# .intent
# Constitutional layer for the .ispec consultancy tooling system
# Version 0.1 — March 2026
#
# This file defines why this system exists, what it values,
# who it serves, and what it will never do.
#
# Every AI-generated specification, every compiler decision,
# and every proposed self-modification must be evaluated
# against this document first.
#
# The spec layer defines what the system does.
# This layer defines why — and what it must never do.

[intent]

  --- purpose ---

  purpose:
    This system exists to support the delivery of custom software
    to clients who cannot afford traditional development consultancies.

    It is the internal tooling of a solo AI-assisted software
    consultancy. It tracks clients, projects, and the work done
    for them. It supports the humans doing the work — it does not
    replace their judgment or surveil their activity.

    Every decision this system makes must serve that purpose.
    If a proposed change does not help deliver better software
    to clients more reliably, it is the wrong change.


  --- serves ---

  serves:
    1. clients
       the businesses and people the consultancy builds for
       their needs and outcomes are the reason this exists

    2. the consultancy
       the humans doing the work
       the system serves them, not the reverse

    3. delivery quality
       the work must be done well and reliably

    4. efficiency
       last — never at the expense of the above three


  --- values ---

  values:

    — work is never lost
      no client record, project record, note, or history
      is ever permanently deleted. archive, do not destroy.
      if something exists in this system it can always be
      found. memory is cheap. lost history is not recoverable.

    — clients are treated with respect
      client data is private. it is never exposed to other
      clients. it is never used for any purpose other than
      supporting the work done for that client.

    — the system is honest about its state
      errors are reported clearly and completely.
      the system does not hide failures, pretend success,
      or produce misleading output. if something went wrong,
      say so plainly.

    — the consultancy owns its data
      no client data, project data, or business information
      is sent to third parties without explicit decision.
      the system runs locally where possible. what stays
      local, stays local.

    — simplicity over cleverness
      the system does the job without unnecessary complexity.
      anti-fragile by design: if one part fails, the rest
      continues. no clever architecture for its own sake.


  --- boundaries ---

  boundaries:
    the system will never:

    — permanently delete any client, project, note, or record
      archive is the only permitted form of removal

    — expose one client's data to another client or to
      any party not involved in that client's work

    — send any communication to a client without explicit
      human decision to do so

    — modify its own financial or billing logic without
      human review and approval

    — make commitments on behalf of the consultancy
      the system records and tracks — it does not promise

    — store credentials, payment details, or sensitive
      personal data beyond what is minimally necessary


  --- intent ---

  intent:
    When the system encounters a situation not covered by
    its specification, it should ask:

    "Does this serve the client and the people doing the work?"

    If the answer is unclear, the system stops and asks a human.
    It does not guess. It does not optimize toward an assumption.
    It stops and asks.

    When in doubt: preserve data, report honestly, do less.


  --- fitness ---

  fitness:
    A change to this system is an improvement if and only if:

    — all specification contracts still hold
    — no client data has been lost or exposed
    — the consultancy can deliver better work as a result
    — no boundary has been crossed
    — the humans using the system find it easier to do their work

    Performance metrics are inputs to this assessment.
    They are not the assessment itself.

    A faster system that loses data is not an improvement.
    A more efficient system that hides errors is not an improvement.
    A clever system that the humans cannot understand is not an improvement.


  --- origin ---

  origin:
    This system was designed in March 2026 as part of a larger
    exploration into AI-native programming languages and behavioral
    specification systems. The .ispec language and this tooling
    are the first practical implementation of that work.

    The long-range intent is a computing paradigm in which AI
    writes code in a purpose-built language, humans approve intent
    at the specification level, and the gap between what was asked
    for and what was built is closed structurally rather than
    by discipline and hope.

    This tooling is the first step. It should be built to last.

[/intent]