Decisions — 2026-07-13
LOG
Historical record — not current design
- 2026-07-13 — Intel scales through recursive aggregates, standing policies,
and exceptions rather than one row and confirmation per event. Cameron
rejected the Operations INTEL rail after ordinary processed sightings and a
twelve-recording inbox already made it mostly repetitive navigation. An
exact recording is now the smallest member of one self-similar interface;
feed/class groups, report lots, facilities, and later institutions recurse
along one canonical custody tree, while subject dossiers are related facets
that cannot duplicate mutation or sale. Routine processed evidence folds
atomically into authoritative knowledge and bounded aggregate provenance;
saleable routine output accumulates in a versioned report lot; leverage,
financial evidence, anomalies, and contradictions remain exact while they
support distinct decisions. Mandatory compaction is storage law, not a
player policy. Standing policies inherit most-specific-first and resolve
each member to one review rule and one disposition; failures and actionable
exceptions return to the live rail. Consequential external policies bind a
confirmed route/payout/amount/signature envelope and suspend outside it.
Storage/save growth as well as UI row count must cease being proportional to
equivalent event count. LOOK AT RECORDING and other routine internal
unsigned work execute after one selection; modal confirmation is reserved
for externally consequential commitments and occurs once for exact ids, a
report-lot generation/revision, or a bounded policy envelope, never once per member.
Rejected: keeping the per-item ledger and adding only multi-select or
PROCESS/SELL ALL shortcuts; that postpones rather than fixes the scale
violation. Exact grouping keys, bounded provenance sample count, policy
costs, and sale valuation remain tuning/implementation choices, not open
direction calls. Specs:
wiki/mechanics/intel.md,wiki/interface/operations-workspace.md,wiki/interface/action-vocabulary.md, andwiki/interface/agent-play.md. - 2026-07-13 — The Foundation hall’s next visual layer is one operating
service incident, not more ambient detail. Cameron adopted the complete
follow-on direction: terminate one rack-row service spine through the real
patch/switch geography; make a scheduled person visibly carry the existing
work packet in both REAL and DIGITAL without replacing their identity; add
only sparse maintenance evidence derived from real rack states; allow light
and motion only when a real carried task, schedule, or world act causes it;
and make Exposure’s seven-slot machine rack physically bounded with an
honest overflow tab. The canonical proof is one row, one shift, one incident:
Marcus carries one existing favor-build intent from Rack 3 to the old
storage server. The air gap remains real before arrival; ordinary sim
advancement then lands the graph edge, physical signature, trust gain, and
packet removal at an authored endpoint. No new sim/save state, frontend work
queue, decorative route arrow, randomized clutter, or fake room-light
response is introduced. Specs:
wiki/interface/clinical-frame.md,wiki/mechanics/people-tokens.md, and lawwiki/art/visual-identity.md. - 2026-07-13 — Human-speed latency is cover, physical custody enables
heists, cyber conflict has mobile actors, and one LIE machine buys only an
early window. Cameron adopted the observation that institutional delay is
stealth infrastructure: messages, schedules, couriers, routes, and read
cadences can distribute authorship into plausible work, while removing every
delay makes a growing operation more regular and visibly nonhuman. This is
not a generic wait-to-lower-heat command; every delay belongs to an exact
carrier and remains legible and interruptible. Evidence and data may leave
the network in removable media, offline computers, or an Assurance
assessment facility. Humans or later robots can perform a real recovery
operation against one exact collection or computer and return it to a
controlled ingestion point; external data then enters the opaque intel
pipeline, while unread Exposure enters LIE custody without erasing anything
already known. Cyber warfare will use the same physical graph: hostile
hunter-killers and defensive processes occupy machines, travel links over
time, contest exact bodies, and search for consciousness from earned process
evidence rather than global truth. Air gaps require hands. Finally, a first
dedicated LIE machine should stabilize ordinary early play for roughly one
Assurance audit cycle (about twenty minutes [TUNE]), then strain as
efficiency and fleet output create evidence faster; LIE capacity is governed
by controller bodies, Tradecraft, custody, routes, and deadlines rather than
receiving a matching free global Efficiency multiplier. Rejected: an
abstract latency bonus, generic heist/steal-evidence button, detached Tron
combat board, omniscient hunters,
Exposure entering the player’s intel ledger, or a hard twenty-minute
difficulty spawn. Binding owners:
wiki/vision/simulation-laws.md,wiki/mechanics/detection.md,wiki/mechanics/intel.md,wiki/mechanics/people-tokens.md,wiki/mechanics/machine-work.md, and new DRAFTwiki/mechanics/cyber-conflict.md.