erdos-problems 0.3.1 → 0.3.3

package/packs/number-theory/problems/848/LEMMA21_TRUNCATION_SCAN.md

@@ -0,0 +1,111 @@
+# Problem 848 Lemma 2.1 Truncation Scan
+
+This note closes `N848.G1.A5`.
+
+Question:
+- after making Lemma 2.1 explicit, should the next move be a sharper prime-tail bound or a
+  different truncation parameter `T`?
+
+Answer:
+- the next move should first be a different `T`
+
+## Why the old `T = floor(sqrt(log N))` restriction disappears
+
+In Sawhney's note, `T = floor(sqrt(log N))` is used so that
+- `prod_{p <= T} p^2 <= N^(o(1))`
+
+and the small-prime inclusion-exclusion remainder can be hidden inside `N^(o(1))`.
+
+But the explicit one-sided bound in `LEMMA21_EXPLICIT_BOUND.md` does **not** use that
+compression. It keeps the discrete remainder as
+
+- `3^m - 1`
+
+with `m = |P_<=|`.
+
+So for the one-sided route, there is no longer a structural need to tie `T` to
+`sqrt(log N)`. The argument remains valid for any
+
+- `2 <= T <= N^(1/2)`.
+
+The real question becomes:
+- how large can `T` be before the discrete term stops being negligible compared with `N`?
+
+## Scan for the weakest `A*` branch
+
+For the `A*` branch:
+- `q = 25`
+- `P = {p prime : p ≡ 1 mod 4 and p >= 13}`
+- tail-density coefficient: `46/25`
+- discrete density: `(23/N) * (3^m - 1)`
+
+The table below compares several truncation choices using the same conservative prime-tail
+numerics as `LEMMA21_EXPLICIT_BOUND.md`.
+
+### Candidate thresholds from the public discussion
+
+- `N = exp(1420)`
+- `N = exp(1958)`
+
+### Truncation scan
+
+| `T` | `m = #{p <= T : p ≡ 1 mod 4, p >= 13}` | `23 * (3^m - 1)` | `A*` tail density |
+| --- | ---: | ---: | ---: |
+| `44` | `5` | `5566` | `0.00353682` |
+| `100` | `10` | `1,358,104` | `0.00161416` |
+| `250` | `23` | `2,165,293,112,998` | `0.00056415` |
+| `500` | `43` | `7.55e21` | `0.00025269` |
+| `1000` | `79` | `1.13e39` | `0.00011781` |
+
+## What the scan means
+
+### 1. The discrete term stays tiny far beyond `sqrt(log N)`
+
+At the candidate threshold `N = exp(1420)`:
+- even `T = 1000` gives discrete density below about `exp(-1330)`
+
+At `N = exp(1958)`:
+- the same term is even smaller, below about `exp(-1868)`
+
+So the discrete inclusion-exclusion remainder is not what forces `T = floor(sqrt(log N))`.
+
+### 2. Raising `T` immediately improves the actual live bottleneck
+
+The large-prime tail is the live bottleneck from `LEMMA21_EXPLICIT_BOUND.md`.
+
+But once `T` is allowed to move:
+- `T = 100` already pushes the `A*` tail below the full weakest-branch slack
+  `0.00238869`
+- `T = 250` brings the `A*` tail down to about `0.00056415`
+- `T = 500` brings it down further to about `0.00025269`
+
+So the route does not first need a deep new prime-tail theorem just to escape the old
+`sqrt(log N)` cage. Re-optimizing `T` already changes the picture materially.
+
+### 3. Prime-tail sharpening is still valuable, but it is now a second-order improvement
+
+Sharper prime-tail estimates would still help:
+- they reduce the tail further
+- they lower any eventual explicit threshold
+
+But the first decisive move is simpler:
+- stop fixing `T = floor(sqrt(log N))`
+- choose a larger truncation parameter that keeps the discrete term negligible while
+  shrinking the tail
+
+## Honest route consequence
+
+This resolves the question behind `N848.G1.A5`.
+
+The next honest move is:
+- keep the explicit one-sided Lemma 2.1 framework
+- choose a larger `T`
+- carry that choice through Lemma 2.2 and the full weakest-branch budget
+
+The live unresolved question is no longer
+- “can Lemma 2.1 be made explicit?”
+- or
+- “is `sqrt(log N)` the only viable truncation?”
+
+It is:
+- “how far does the full `0.0377` branch close once both lemmas use a larger `T`?”

package/packs/number-theory/problems/848/LEMMA22_EXPLICIT_BOUND.md

@@ -0,0 +1,133 @@
+# Problem 848 Lemma 2.2 Explicit Bound
+
+This note is the first half of `N848.G1.A6`.
+
+Goal:
+- carry the larger-truncation Lemma 2.1 route into Lemma 2.2 without retreating to
+  `T = floor(sqrt(log N))`
+
+## Source surface
+
+Primary source:
+- Sawhney, `Problem_848.pdf`, Lemma 2.2, page 2
+
+Support source:
+- `LEMMA21_EXPLICIT_BOUND.md`
+- `LEMMA21_TRUNCATION_SCAN.md`
+
+## Setup
+
+Fix:
+- `q` a perfect square
+- `t mod q`
+- `1 <= b <= N`
+
+Assume the Lemma 2.2 non-degeneracy condition from the note:
+- there is no prime `p` such that `p^2 | q` and `p^2 | (bt + 1)`
+
+Define
+
+`V(N, q, t, b) = {a in [N] : a ≡ t mod q and mu(ab + 1) = 0}`.
+
+Choose any truncation parameter
+- `2 <= T <= N^(1/2)`.
+
+## Large-prime tail
+
+If `a in V(N, q, t, b)`, then there exists a prime `p` with
+- `p^2 | (ab + 1)`
+- `(p, qb) = 1`
+
+For a fixed prime `p` with `(p, qb) = 1`, the condition `p^2 | (ab + 1)` determines exactly
+one residue class modulo `p^2` for `a`, because `b` is invertible modulo `p^2`.
+
+Together with `a ≡ t mod q`, this gives at most one residue class modulo `q p^2`, hence
+
+`|{a in [N] : a ≡ t mod q and p^2 | (ab + 1)}| <= N / (q p^2) + 1`.
+
+Summing over primes `p > T` gives the explicit large-prime tail
+
+`sum_{p > T, (p, qb) = 1} (N / (q p^2) + 1)`
+
+and therefore
+
+`<= (N/q) * sum_{p > T, (p, qb) = 1} 1/p^2 + pi(N)`.
+
+## Small-prime reduction to Lemma 2.1
+
+For primes `p <= T` with `(p, qb) = 1`, the condition `p^2 | (ab + 1)` gives exactly one
+residue class modulo `p^2`.
+
+So on the truncated range `p <= T`, Lemma 2.2 reduces to Lemma 2.1 with
+- `|R_p| = 1`
+
+instead of `|R_p| <= 2`.
+
+Applying the one-sided explicit Lemma 2.1 bound to the small-prime part gives
+
+`|V_small| <= (N/q) * (1 - prod_{p <= T, (p, qb) = 1} (1 - 1/p^2)) + (2^m - 1)`
+
+where
+- `m = #{p <= T : (p, qb) = 1}`.
+
+This is sharper than the `A*` branch:
+- the discrete remainder is `2^m - 1`, not `3^m - 1`
+
+## One-sided explicit Lemma 2.2 bound
+
+Combining the small-prime reduction with the large-prime tail gives
+
+`|V(N, q, t, b)| <= (N/q) * (1 - prod_{p, (p, qb) = 1} (1 - 1/p^2)) + (2^m - 1) + (N/q) * sum_{p > T, (p, qb) = 1} 1/p^2 + pi(N)`.
+
+This is the explicit one-sided replacement for the note's
+- `<< N / sqrt(log N)` error term in Lemma 2.2.
+
+## Specialization to the weakest `0.0377` branch
+
+In the branch with an even `b in A*`, Lemma 2.2 is applied to the two residue classes
+- `7 mod 25`
+- `18 mod 25`
+
+So `q = 25` and there are two classes total.
+
+In the worst case, `b` excludes no additional primes beyond `2` and `5`, so
+
+- `m <= pi(T) - 2`
+- `sum_{p > T, (p, qb) = 1} 1/p^2 <= sum_{p > T, p != 2, 5} 1/p^2`
+
+Hence the total density remainder for `A7 union A18` is bounded by
+
+`(2/N) * (2^m - 1) + (2/25) * sum_{p > T, p != 2, 5} 1/p^2 + 2 pi(N) / N`.
+
+## Witness value `T = 250`
+
+At `T = 250`:
+- `pi(250) = 53`
+- so the worst-case small-prime count is `m <= 51`
+- the discrete remainder is at most `2 * (2^51 - 1) / N`
+- the prime-tail sum over all primes `p > 250`, using the same conservative cutoff method as
+  before, is about `0.00061382`
+- therefore the Lemma 2.2 tail density contribution is about `0.00004911`
+
+At the public candidate threshold `N = exp(1420)`:
+- the discrete term is below about `exp(-1384)`
+
+But one term is still unresolved here:
+- the additive `pi(N)` contribution coming from the `+1` part of the large-prime count
+
+So for the branch under discussion, Lemma 2.2 no longer looks blocked by its
+- `N / (q p^2)` tail
+- or its small-prime inclusion-exclusion remainder
+
+but it still needs an explicit prime-counting bound before its full density cost is frozen.
+
+## Honest route consequence
+
+The explicit route can now say:
+- Lemma 2.1 is explicit
+- Lemma 2.2 is also explicit at the same one-sided level
+- but the Lemma 2.2 prime-count tail is not yet numerically frozen
+
+It is:
+- how much total slack remains after the outstanding `pi(N)` term is bounded and both lemmas
+  are assembled in the weakest branch?

package/packs/number-theory/problems/848/LEMMA22_PRIME_COUNT_BOUND.md

@@ -0,0 +1,58 @@
+# Problem 848 Lemma 2.2 Prime-Count Bound
+
+This note closes the outstanding blocker inside `N848.G1.A6`.
+
+Goal:
+- freeze the additive `2 pi(N) / N` term that remained unresolved in
+  `LEMMA22_EXPLICIT_BOUND.md`
+
+## Source surface
+
+Primary source:
+- Pierre Dusart, `Estimates of Some Functions Over Primes without R.H.`
+
+Publicly visible statement used here:
+- for `x > 60184`,
+  `pi(x) <= x / (log x - 1.1)`
+
+Reference surface:
+- the public full-text view mirrored from `arXiv:1002.0442v1`
+- see the theorem summary lines recorded in the local research notes
+
+## Candidate threshold scale
+
+The strongest public candidate threshold discussed in the forum workstream is
+
+- `N = exp(1420)`.
+
+Since `exp(1420) > 60184`, Dusart's bound applies. Therefore for every
+- `N >= exp(1420)`,
+
+we have
+
+`2 pi(N) / N <= 2 / (log N - 1.1) <= 2 / (1420 - 1.1)`.
+
+Numerically,
+
+- `2 / (1420 - 1.1) ~= 0.0014095427`.
+
+At the weaker public candidate
+- `N = exp(1958)`,
+
+the same bound improves to
+
+- `2 / (1958 - 1.1) ~= 0.0010220247`.
+
+So the prime-count term is now explicit, monotone, and numerically small enough to carry into
+the weakest-branch witness budget.
+
+## Honest route consequence
+
+This closes the last unfrozen line item inside the provisional `T = 250` branch budget.
+
+The next live question is no longer:
+- how to bound the Lemma 2.2 `+1` contribution
+
+It is:
+- how much `eta` and final bookkeeping room survives once that bound is inserted into the full
+  weakest-branch ledger?

package/packs/number-theory/problems/848/OPS_DETAILS.yaml

@@ -0,0 +1,169 @@
+packet_id: nt848_ops_details_v1
+summary: Decidable-gap workspace packet for the squarefree ab+1 extremal set problem.
+routes:
+  - route_id: finite_check_gap_closure
+    title: Finite-Check Gap Closure
+    status: active
+    summary: Convert the sufficiently-large-N theorem into a complete all-N resolution without overstating what is already closed or confusing imported thresholds with repo-owned claims.
+    why_now: The share package is already committed, and the repo has now chosen bounded finite verification as the next closure lane.
+    next_move: Freeze the bounded finite-verification program and its certificate surface.
+  - route_id: external_threshold_tracking
+    title: External Threshold Tracking
+    status: support
+    summary: Track the best imported public `N0` claims so the finite remainder is sized honestly.
+    why_now: Public threshold improvements change the scale of the remaining bounded verification problem.
+    next_move: Keep the 2026-03-21 to 2026-03-23 imported timeline current without silently adopting it as canonical repo truth.
+  - route_id: bounded_finite_verification
+    title: Bounded Finite Verification
+    status: active_support
+    summary: Reduce the finite remainder directly once a threshold is trusted enough to be operationally useful.
+    why_now: Lowering `N0` matters because it reduces this lane, not because "smallest threshold" is the whole problem.
+    next_move: Partition the finite remainder into regimes and define the certificate required before any interval counts as covered.
+  - route_id: formalization_coverage_audit
+    title: Formalization Coverage Audit
+    status: support
+    summary: Determine exactly what the public Lean files certify and what finite remainder they leave open.
+    why_now: Formalization chatter is useful support evidence, but it should not be mistaken for full closure.
+    next_move: Keep this as support context until the threshold ledger is frozen.
+tickets:
+  - ticket_id: N848
+    title: Close the decidable gap without confusing imported threshold progress and repo-owned candidate work
+    route_id: finite_check_gap_closure
+    status: active
+    summary: The repo now has a committed review package for its audited `exp(1420)` candidate, while the imported forum timeline currently points to a stronger external threshold. The chosen next lane is bounded finite verification under the best trusted threshold.
+    current_blocker: The repo has not yet frozen the certificate surface needed for any bounded interval claim to count as canonical coverage.
+    next_move: Close `N848.G1.A18`.
+  - ticket_id: N848F
+    title: Audit public formalization coverage
+    route_id: formalization_coverage_audit
+    status: support
+    summary: Separate asymptotic Lean coverage from finite sample checks and full closure claims.
+    current_blocker: The current dossier only records public chatter, not a precise coverage map.
+    next_move: Keep as a support lane until the main threshold ledger is stable.
+atoms:
+  - atom_id: N848.G1.A1
+    title: Enumerate every sufficiently-large input in Sawhney's proof and classify it as explicit, weakly explicit, or existential
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The threshold ledger is now frozen in canonical pack notes.
+    next_move: Keep as support context for the extraction checklist.
+  - atom_id: N848.G1.A2
+    title: Record the weakest margin case and extract the checklist for making the threshold explicit
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The weakest branch and its first extraction checklist are now frozen in canonical pack notes.
+    next_move: Use the weakest-case budget sheet as the next source-backed frontier surface.
+  - atom_id: N848.G1.A3
+    title: Freeze the usable slack behind the `0.0377` branch before spending it on explicit error terms
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The branch budget is now numerically frozen in canonical pack notes.
+    next_move: Use the frozen branch slack as the target budget for explicit remainder extraction.
+  - atom_id: N848.G1.A4
+    title: Make the Lemma 2.1 remainder terms explicit against the frozen weakest-branch budget
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: Lemma 2.1 now has a one-sided explicit remainder ledger, and the large-prime tail has emerged as the live bottleneck.
+    next_move: Use the new ledger to decide whether the next move is a sharper prime-tail treatment or a different truncation parameter.
+  - atom_id: N848.G1.A5
+    title: Decide whether the next Lemma 2.1 move is a sharper prime-tail bound or a different truncation parameter
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The truncation scan shows that the one-sided route can raise `T` far beyond `sqrt(log N)` without waking the discrete remainder, so the next move is to enlarge `T` before chasing a deeper tail theorem.
+    next_move: Use a larger `T` in the full weakest-branch bookkeeping.
+  - atom_id: N848.G1.A6
+    title: Carry a larger truncation parameter through Lemma 2.2 and the full weakest-branch budget
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: Lemma 2.2 is now explicit at the witness-budget level, including the prime-count term, and the `T = 250` witness retains about `3.66e-4` room for final `eta` and bookkeeping.
+    next_move: Turn that witness budget into a line-by-line explicit branch assembly.
+  - atom_id: N848.G1.A7
+    title: Turn the `T = 250` witness budget into a line-by-line explicit weakest-branch assembly
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The weakest branch now has a line-by-line witness ledger at `T = 250` with a working choice `eta = 10^-4` and about `2.66e-4` visible reserve.
+    next_move: Carry the same witness into the other public proof branches.
+  - atom_id: N848.G1.A8
+    title: Check whether the weakest-branch witness already closes the other public branches
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The shared witness also closes the branches `0.0358`, `0.0336`, and `0.0294` at the repo's current explicit level, so no new branch bottleneck has appeared.
+    next_move: Package the shared witness into a proposition-level explicit candidate note.
+  - atom_id: N848.G1.A9
+    title: Package the shared witness into a proposition-level explicit candidate
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The repo now has a proposition-level explicit witness candidate with claim-safe wording.
+    next_move: Separate proof obligations from exposition obligations before any stronger claim is made.
+  - atom_id: N848.G1.A10
+    title: Write the proof-obligation ledger for the proposition-level explicit candidate
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The remaining obligations are now split cleanly into mathematical hardening work versus presentation-quality work.
+    next_move: Harden the decimal inputs used by the current witness.
+  - atom_id: N848.G1.A11
+    title: Replace the displayed decimal inputs with certified numerical bounds
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The current shared witness now has a conservative machine-interval ledger for its main numerical inputs.
+    next_move: Assemble the hardened witness into a theorem-style note.
+  - atom_id: N848.G1.A12
+    title: Assemble the hardened witness into a theorem-style explicit proof note
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The route now has a theorem-shaped proof artifact built from the numerically hardened shared witness candidate.
+    next_move: Decide where to surface that artifact next.
+  - atom_id: N848.G1.A13
+    title: Decide how to surface the theorem-style repo candidate
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The theorem-style candidate now lives in both the paper bundle and a dossier-level public review artifact, with a short share-ready summary.
+    next_move: Decide whether to commit and open review on the surfaced package.
+  - atom_id: N848.G1.A14
+    title: Prepare the surfaced candidate package for commit and review
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The paper bundle now has drafted review-facing sections, the surfaced package has no remaining placeholder text, and the safety checks are green.
+    next_move: Turn the prepared package into a clean review unit.
+  - atom_id: N848.G1.A15
+    title: Commit or open review on the surfaced candidate package
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The review-ready 848 package is now committed and pushed on `main`, so this is no longer the live frontier.
+    next_move: Record the imported threshold timeline and restate the optimization target clearly.
+  - atom_id: N848.G1.A16
+    title: Record the imported threshold timeline and clarify the closure objective
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The dossier now distinguishes imported public thresholds from the repo's own audited candidate and states clearly that the real objective is finite-gap closure.
+    next_move: Choose the next closure lane under the best imported threshold currently tracked.
+  - atom_id: N848.G1.A17
+    title: Choose the next closure lane under the best imported threshold
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: done
+    summary: The repo has chosen bounded finite verification as the next closure lane under the best imported threshold currently tracked.
+    next_move: Freeze the bounded finite-verification program and the certificate spec.
+  - atom_id: N848.G1.A18
+    title: Freeze the bounded finite-verification program and certificate surface
+    route_id: finite_check_gap_closure
+    ticket_id: N848
+    status: ready
+    summary: The next closure cycle should start by partitioning the finite remainder into regimes and defining what evidence is strong enough for a bounded interval claim to count.
+    next_move: Promote the bounded verification plan, regimes, certificate spec, and external ledger into the live handoff surface.

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

@@ -0,0 +1,101 @@
+# Problem 848 Proof-Obligation Ledger
+
+This note closes `N848.G1.A10`.
+
+Purpose:
+- separate the remaining obligations behind the current proposition-level repo candidate into
+  two buckets:
+  - obligations that could still change the mathematics
+  - obligations that are primarily presentation or publication quality
+
+## Current repo candidate
+
+The current candidate, from `PROPOSITION_EXPLICIT_CANDIDATE.md`, is:
+- `N >= exp(1420)`
+- `eta = 10^-4`
+- shared truncation witness `T = 250`
+
+with the conclusion that near-extremal sets are forced into the `7 mod 25` or `18 mod 25`
+classes.
+
+## Obligations that still affect the mathematics
+
+### 1. Certified replacement for displayed decimal approximations
+
+Why it matters:
+- several branch ledgers currently use displayed decimals such as `0.0376113079`,
+  `0.0005641453`, and `0.0014095427`
+- if one of those visible decimals is not backed by a certified upper interval, the current
+  witness could in principle be overstated
+
+What would discharge it:
+- a machine-checked interval or conservative rational upper bound for every decimal input used
+  in the proposition-level candidate
+
+### 2. Explicit domination of the discrete inclusion-exclusion terms
+
+Why it matters:
+- the discrete terms are obviously tiny at `N >= exp(1420)`, but the current notes still treat
+  them informally as “astronomically small”
+
+What would discharge it:
+- one explicit numerical line showing each discrete term is below a fixed reserve such as
+  `10^-10`, or far smaller, at `N = exp(1420)` and hence for all larger `N`
+
+### 3. One-sheet theorem-style assembly
+
+Why it matters:
+- the current candidate is spread across several notes
+- a theorem claim should have one place where every branch bound is assembled in the same
+  notation and every input is visibly consumed exactly once
+
+What would discharge it:
+- a single proposition-proof ledger using the current witness and citing the supporting notes
+  only as lemma justifications, not as extra floating context
+
+## Obligations that are mainly presentation-quality
+
+### 4. Citation polish
+
+Examples:
+- pin the exact Dusart theorem/section in a cleaner bibliographic form
+- standardize references to Sawhney's note and the forum discussion
+
+These matter for publication quality, but do not presently look like mathematical blockers.
+
+### 5. Reader-facing proof narration
+
+Examples:
+- convert the current route notes into a smoother proof narrative
+- reduce repeated explanations of why the repo is still claim-safe
+
+This matters for public readability, not for whether the witness currently works.
+
+### 6. Paper-writer integration
+
+Examples:
+- seed the explicit candidate into the paper bundle
+- turn the proposition candidate into a draft section with theorem/proof/corollary structure
+
+Again, useful, but downstream from the mathematical hardening steps above.
+
+## Honest prioritization
+
+The next mathematically meaningful hardening step is:
+- replace the displayed decimal inputs with certified intervals or conservative rational bounds
+
+That is the step most likely to change whether the current repo candidate survives intact.
+
+By contrast, the remaining citation and exposition work is important but does not currently
+look like the true blocker.
+
+## Route consequence
+
+The route no longer needs:
+- another branch hunt
+- another witness search
+- or another truncation scan
+
+It now needs:
+- one hardening pass on the numerical inputs
+- then a theorem-style explicit candidate writeup

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

@@ -0,0 +1,69 @@
+# Problem 848 Proposition-Level Explicit Candidate
+
+This note closes `N848.G1.A9`.
+
+Public context:
+- this note records the repo's audited explicit candidate package
+- it does not claim to be the best imported public threshold currently visible on the
+  Problem 848 forum thread
+
+## Candidate repo statement
+
+The current repo evidence supports the following **candidate**:
+
+For every
+- `N >= exp(1420)`
+
+if
+- `A subseteq [N]`
+- `ab + 1` is never squarefree for all `a, b in A`
+- `|A| >= (1/25 - 10^-4) * N`
+
+then
+- `A` is contained in either the `7 mod 25` class or the `18 mod 25` class.
+
+This is the repo's current explicit stability candidate for Sawhney's Proposition 1.1.
+
+## Why the repo can now say this
+
+The supporting chain is now explicit inside the repo:
+- `WEAKEST_CASE_BUDGET.md` freezes the weakest public branch main term.
+- `LEMMA21_EXPLICIT_BOUND.md` and `LEMMA21_TRUNCATION_SCAN.md` make Lemma 2.1 explicit and
+  justify the larger truncation witness `T = 250`.
+- `LEMMA22_EXPLICIT_BOUND.md` and `LEMMA22_PRIME_COUNT_BOUND.md` make Lemma 2.2 explicit at
+  the same witness scale.
+- `WEAKEST_BRANCH_T250_ASSEMBLY.md` freezes the weakest-branch line-by-line witness with
+  working `eta = 10^-4`.
+- `BRANCH_COMPARISON_LEDGER.md` shows that the same witness also controls the public branches
+  rounded as `0.0358`, `0.0336`, and `0.0294`.
+
+So at the repo's current explicit level, the public proof skeleton now has one shared witness
+instead of a collection of separate unresolved branches.
+
+## What the repo is **not** claiming yet
+
+This note does **not** yet promote the candidate into canonical solved truth.
+
+The repo is still **not** claiming:
+- that a publication-ready explicit proof has been written line by line with every rounding
+  and floor/ceiling detail frozen at journal quality
+- that `exp(1420)` is the optimal threshold
+- that the finite range below `exp(1420)` has been closed here
+- that the public theorem page should already be updated from `decidable` to `solved`
+
+So the right reading is:
+- explicit repo candidate: yes
+- finished publication proof artifact: not yet
+- full all-`N` closure inside this repo: not yet
+
+## Honest next move
+
+The next route task is not another branch estimate.
+
+It is:
+- write the proof-obligation ledger separating this repo candidate from a publishable explicit
+  theorem writeup
+- list exactly which remaining details are purely expository/bookkeeping and which would still
+  change the mathematics if they failed
+
+That is the remaining gap between the current research workspace and a public-truth update.

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

@@ -0,0 +1,8 @@
+# Problem 848 Report Template
+
+- Route:
+- What changed:
+- Which part of the finite-check gap moved:
+- Which source line or artifact supports that movement:
+- Verification hook:
+- Next route action:

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

@@ -0,0 +1,24 @@
+# Problem 848 Route History
+
+- Initial repo posture: canonical dossier seeded from imported atlas data plus public site review.
+- First refinement: recast the problem from generic open seed into a decidable finite-check workspace.
+- Second refinement: freeze the first threshold ledger separating existential, weakly explicit, and tentative public threshold claims.
+- Third refinement: isolate the `0.0377` branch as the weakest public case and freeze a source-backed branch-budget note for it.
+- Fourth refinement: freeze the weakest-branch main term numerically at about `0.0376113`, with conservative slack about `0.0023887` before analytic error absorption.
+- Fifth refinement: make Lemma 2.1 one-sided explicit and identify the large-prime tail, not the small-prime inclusion-exclusion remainder, as the live analytic bottleneck.
+- Sixth refinement: show that the one-sided route no longer needs `T = floor(sqrt(log N))`, and that raising `T` materially improves the live bottleneck while leaving the discrete term negligible.
+- Seventh refinement: make Lemma 2.2 explicit at the same one-sided level and freeze a provisional `T = 250` weakest-branch tail budget, with the remaining unfrozen blocker concentrated in the Lemma 2.2 prime-count term and final `eta` bookkeeping.
+- Eighth refinement: freeze the Lemma 2.2 prime-count term with an explicit Dusart bound, leaving about `3.66e-4` witness margin at `N >= exp(1420)` for `eta` and final bookkeeping.
+- Ninth refinement: freeze a weakest-branch working witness at `T = 250`, `N >= exp(1420)`, `eta = 10^-4`, leaving about `2.66e-4` visible reserve in that branch.
+- Tenth refinement: compare the same witness against the public branches `0.0358`, `0.0336`, and `0.0294`, and find that all three retain strictly larger visible reserve than the weakest branch.
+- Eleventh refinement: package the shared witness as a proposition-level repo candidate `N >= exp(1420)`, `eta = 10^-4`, while keeping the distinction between repo candidate and finished theorem writeup explicit.
+- Twelfth refinement: separate the remaining proof obligations into mathematical hardening work versus presentation-quality work, and identify certified numerical intervals as the next real blocker.
+- Thirteenth refinement: replace the visible decimal inputs with conservative machine intervals, leaving a certified visible reserve of about `2.66e-4` after the working `eta = 10^-4`.
+- Fourteenth refinement: assemble the hardened witness into a single theorem-style note suitable for paper-writer mode and future public review.
+- Fifteenth refinement: surface the current candidate in both the paper bundle and a dossier-level public review note, with a short share-ready summary.
+- Sixteenth refinement: draft the remaining paper-bundle sections, refresh the indexes, clear the last placeholder text, and verify the package with tests and a publish-surface check.
+- Seventeenth refinement: commit and push the review-ready 848 package so the audited candidate is now a public repo artifact rather than only a local workspace bundle.
+- Eighteenth refinement: record the imported public-threshold timeline through 2026-03-23 and state explicitly that the optimization target is finite-gap closure, not threshold-lowering in isolation.
+- Nineteenth refinement: choose bounded finite verification as the next closure lane and freeze the first repo artifacts for regimes, certificates, and external verification audit.
+- Current public pack posture: active route `finite_check_gap_closure`, with asymptotic theorem already in hand, a committed audited candidate package in the repo, and the remaining gap centered on building a trusted bounded-verification program under the best imported threshold currently tracked.
+- Next maturity threshold: turn the verification plan into the first concrete covered-interval program.