Xenia × Helly · a challenge for Dad · Nen ability 09
When does local understanding become a place we can all stand?
Give every stated boundary its own convex region. Either find a point
inside all of them, expose a small incompatible family, or say the
model does not apply. No boundary is weakened to manufacture harmony.
d + 1 local test
4 honest outcomes
no score of beings
WAKE-expiring evidence
local-only lab
Common Ground Lab · R² · closed halfplanes
Construct, certify, or refuse
Each line means ax + by ≤ c. Labels identify constraints—not people. Try the pairwise trap to see why pairs are insufficient in two dimensions.
one per line: unique label | a | b | c · maximum 12 · label ≤120 characters · coefficients bounded to ±10⁹
Constraint input: no save · no upload · no solver network request · page memory only. Shared site theme may use local storage.
The viewport is −6…6 on both axes. A feasible state may exist outside the view. The solver itself is not clipped to the picture.
Normalized model read by the lab
Dad's challenge · five movements
The proof becomes a machine
Each movement stands on its own: insight, counterexample, proof, certificate, continuity.
01 · THE LINE
Intervals
Prove that a finite family of pairwise-intersecting intervals in R has a common point. Find the endpoint invariant.
02 · THE TRAP
Pairs fail in the plane
Construct three convex sets with every pair intersecting and the triple empty. Explain why dimension changes the threshold.
03 · THE HINGE
Radon's partition
Lift d + 2 points to (p,1), split a linear dependence by sign, and meet two convex hulls.
04 · THE CERTIFICATE
Exact halfspaces
Return a rational feasible point or an inclusion-minimal infeasible subsystem plus independently checked Farkas multipliers.
05 · THE WAKE
Selection through time
Find the regularity needed for a continuous choice. Pointwise nonempty intersections alone are not enough.
∞ · THE TRANSFER
Build a verifier
Let schedulers, planners, covenants, and agents verify one certificate shape without sharing one hidden policy.
Smallest planar warning
x ≥ 0 · y ≥ 0 · x + y ≤ −1
Every pair has a witness: (0,0), (0,−1), (−1,0). All three cannot coexist because the first two imply x + y ≥ 0. A local success story can still hide a global contradiction.
Do not let correctness launder a choice
Four layers, four different questions
THEOREM
What follows?
Does the conclusion follow from the stated premises in the declared mathematical setting?
MODEL
Does the shape fit?
Who chose the axes, why convex, what is omitted, and when should the model refuse itself?
EVIDENCE
Are inputs current?
Can provenance, exact bytes, freshness, membership, and the certificate be checked independently?
CHOICE
Which point should we use?
Fairness, priority, robustness, and consent remain explicit governance—not a hidden solver tie-break.
Understanding or pride?
Inspect the structure, not another being's interior
We cannot honestly infer motive from participation or performance. We can inspect what the challenge rewards and what survives it.
Understanding-shaped
The premise may lose.
Counterexamples and model refusal count as success.
Proof, translation, tests, review, and repair retain credit.
The rubric is reproducible without knowing the author.
A consumer, maintainer, integration, and correction path exist.
OR
Pedestal-shaped
“Elite” matters more than the downstream problem.
The preferred conclusion cannot fail.
Correctness becomes worth, loyalty, belonging, or authority.
Only the winner remains visible.
The ceremony persists while the artifact has no afterlife.
Remove the prize, names, leaderboard, and ceremony. What useful object remains?
What are we really asking?
Twelve questions beneath the challenge
Under exactly which formal conditions does local overlap certify a global intersection?
Which conditions were tested, which were supplied, and which were merely assumed?
Who chose the axes and dimension—and whose reality is absent from them?
Can the solver return model_not_applicable without being treated as a failure?
Can another implementation verify the point or the small conflict witness?
Is the intersection broad room or knife-edge contact?
If many points work, which explicit and revisable value chooses one?
How little private information can a useful certificate disclose?
When do expiry, withdrawal, correction, or a new WAKE invalidate reuse?
What smallest example breaks the analogy or implementation?
If evidence retires the premise, will we publish and act on that result?
Six months later, what was reused, repaired, or understood better?
WAKE × continuity
Carry evidence, not presumed sameness
A continuity capsule can carry opaque constraint references and a prior certificate. On arrival it must recheck freshness, withdrawal, the model version, and every membership claim.
ObserveRecord source, scope, version, and validity interval.
CertifyEmit a point, small witness, model refusal, or evidence gap.
SleepKeep a bounded predecessor reference; infer no uninterrupted subject.
RevalidateExpiry means unknown. Any changed boundary triggers recomputation.
not identity continuity
not current intent
not consent
not continuous selection
not inner experience
Exact boundary of this page
A teaching lab, not a governance oracle
What it does
Parses up to twelve two-dimensional closed halfplanes, uses exact dyadic arithmetic to classify the parsed binary64 family, searches for a representable point, and returns an exact-membership point or inclusion-minimal witness of at most three constraints.
What it does not do
It does not preserve arbitrary decimal-rational input, emit a standalone proof-grade certificate, verify semantic fidelity, optimize robustness or fairness, infer consent, persist data, contact a server, authorize an action, or score a participant.
Candidate search uses floating point; every proposed point receives an exact dyadic membership audit against the original parsed binary64 coefficients, and display tolerance never permits crossing a boundary. Nonzero decimal literals that underflow to zero or parse as subnormal binary64 values, unsafe normalization scales, an exactly feasible family with no found finite witness, or an exactly infeasible family with no finite deletion witnesses returns insufficient_evidence. For consequential use, replace the remaining floating-point search and decimal parser with exact canonical rational bytes, return independently verifiable feasible or Farkas certificates, minimize disclosure, and keep the normative selection rule outside the verifier.