erdos-problems 0.3.5 → 0.3.6

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "erdos-problems",
-  "version": "0.3.5",
+  "version": "0.3.6",
   "description": "CLI atlas and workspace tools for Paul Erdos problems.",
   "type": "module",
   "bin": {

package/packs/number-theory/problems/848/BOUNDED_VERIFICATION_PLAN.md

@@ -38,8 +38,9 @@ What success would look like:
 - the remaining uncovered range shrinks monotonically
 
 Current progress:
-- the repo now has an exact small-`N` certificate for `1..3000`
-- a direct jump to `1..5000` no longer looks cheap enough to treat as the default next move
+- the repo now has an exact small-`N` certificate for `1..10000`
+- an incremental verifier replaced the misleading earlier cost wall and made larger exact
+  tranches practical again
 
 What this lane is not:
 - not brute force to `2.64 x 10^17`

package/packs/number-theory/problems/848/{EXACT_SMALL_N_1_3000_CERTIFICATE.md → EXACT_SMALL_N_1_10000_CERTIFICATE.md}

@@ -1,14 +1,15 @@
-# Problem 848 Exact Small-`N` Certificate: `1..3000`
+# Problem 848 Exact Small-`N` Certificate: `1..10000`
 
-This is the first exact bounded-verification certificate frozen in the repo for Problem `848`.
+This is the current exact bounded-verification base certificate frozen in the repo for
+Problem `848`.
 
 ## Claim
 
-For every `N` with `1 <= N <= 3000`, the maximum size of a set `A subseteq [N]` such that
+For every `N` with `1 <= N <= 10000`, the maximum size of a set `A subseteq [N]` such that
 `ab + 1` is never squarefree for all `a, b in A` is equal to the size of the `7 mod 25`
 or `18 mod 25` residue class in `[N]`.
 
-This certificate only covers the interval `1..3000`.
+This certificate only covers the interval `1..10000`.
 
 ## Method class
 
@@ -21,35 +22,37 @@ Command used:
 ```bash
 node packs/number-theory/problems/848/compute/problem848_small_n_exact_scan.mjs \
   --min 1 \
-  --max 3000 \
-  --json-output packs/number-theory/problems/848/EXACT_SMALL_N_1_3000_RESULTS.json
+  --max 10000 \
+  --json-output packs/number-theory/problems/848/EXACT_SMALL_N_1_10000_RESULTS.json
 ```
 
 The script:
 - filters vertices by the loop condition `a^2 + 1` not squarefree
-- builds the compatibility graph where `a` and `b` are adjacent exactly when `ab + 1` is not
-  squarefree
-- computes an exact maximum clique for each `N`
-- compares that maximum clique size against the `7 mod 25` class size
+- grows the compatibility graph incrementally as `N` increases
+- keeps the current maximum clique and only solves a new exact clique problem when the fresh
+  vertex can actually improve the bound
+- emits compact machine-readable rows by recording the residue-class witness directly whenever
+  the maximum clique is a full `7 mod 25` or `18 mod 25` class
 
 ## Outcome
 
-- interval: `1..3000`
+- interval: `1..10000`
 - result: verified
 - every scanned `N` satisfied:
   - `max_clique_size = max(|{n in [N] : n equiv 7 mod 25}|, |{n in [N] : n equiv 18 mod 25}|)`
 
 The machine-readable result packet is:
-- `EXACT_SMALL_N_1_3000_RESULTS.json`
+- `EXACT_SMALL_N_1_10000_RESULTS.json`
 
 ## Scope warning
 
-This does **not** certify anything above `3000`.
+This does **not** certify anything above `10000`.
 It is a base interval certificate only.
 
 ## Why this interval matters
 
 - it creates a real in-repo covered interval, not just a verification plan
-- it is large enough to be a meaningful finite-check foothold
+- it shows that exact coverage can reach substantially farther than the old pre-incremental
+  verifier suggested
 - it does not rely on disputed public computation
 - it gives the bounded-verification lane a concrete certificate format and reproduction path