daily
loops and signals
collect delta only: what changed confidence, what closed, what got stuck, what got killed, what changed outside.
cadence of truth
how often each layer should move
if everything updates all the time, the dashboard becomes unstable theater. the chat suggested a rhythm that protects the team from its own nervous system.
2-3 times per week
bets and risks
promote signals, prune bets, close loops, refresh current focus, and re-rank the most expensive unresolved loop.
weekly
doctrine review
check whether the north star, first workflow, market framing, or any decision has hardened enough to close.
monthly
search-quality retrospective
review which signals were false, which bets lingered too long, where excitement replaced evidence, and where AI accelerated noise rather than learning.
source hierarchy
not all input should weigh the same
this is one of the strongest process upgrades from the discussion. without source weighting, the team will overfit to articulate discourse instead of causally important truth.
tier 1
highest weight
- direct user behavior retained usage, painful workarounds, repeated high-stakes asks
- failed workflows where the product broke under real pressure
- hard evidence action completion, trust, reliability, recovery
tier 2
medium weight
- interviews and pilot feedback under real stakes
- partner conversations and operator observations
- internal dogfooding when the stakes are not fake
tier 3
low weight
- telegram discourse and smart essays
- market heat and competitive theater
- beautiful phrasing that changes no decisions
dashboard blocks
what the live dashboard should contain at this stage
the most useful proposed block set was simple and severe. enough to orient, not enough to hide confusion.
01
north star now
wedge, principal, first wow, current bottleneck, and the metric that matters this week.
02
active bets
bet, confidence, test, kill condition, owner, and by-when date.
03
hard decisions
decision, why now, what changed, and what the team is explicitly not doing.
04
traction signals
strongest confirming signal, strongest disconfirming signal, new weird signal, promoted evidence.
05
open loops
owner, next step, blocked by, review date. loops without owners should not stay alive quietly.
06
market + risk radar
market shifts, architecture risk, security risk, GTM risk, category risk, all in one turbulence block.
daily bot
questions worth automating
the bot should not behave like a secretary. it should behave like an instrument for epistemic hygiene.
delta-only questions
- what new signal changed our understanding today? if nothing changed, do not perform movement.
- which current bet got stronger or weaker? name the evidence, not the mood.
- is there a decision that is now ready to close? if not, what is missing?
- which open loop is currently the most expensive? hanging cost matters.
anti-self-deception questions
- what changed outside the market that may move our wedge? watch the clock, not just the mirror.
- what are we doing because it sounds smart? strong writing is not strategic progress.
- what did we kill? if nothing dies, the system is not discriminating.
- are we accelerating a learning loop or a noise loop? this is the core question.
promotion discipline
how this fits the repo process
the repo already has the right promotion pipeline. the cadence layer simply makes it operational for team behavior, not just agent behavior.
repo state machine
raw window → digest → promoted memo → doctrine
this is already the local operating model for living sources. the dashboard should mirror it rather than bypass it.
team cognition
design for high-context fatigue
the product is high-context, so the team will also drown without designed compression. the dashboard should reduce telegram soup, not repackage it beautifully.