Private design-partner pilot now open.Apply
ZZeroRun
ZERORUN V0.4 · INTERNAL TECHNICAL QUALIFICATION

Evidence, bounded.

ZeroRun v0.4 passed the unchanged internal pytest and SymPy performance gates inside a narrow reviewed boundary. That is technical qualification—not a claim of universal speedup, external production reliability, or customer demand.

30 historical merged edits.10 PYTEST + 20 SYMPY · INTERNAL CORPUS
pytest p95 reduction78.415%10 historical merged patches
SymPy p95 reduction78.309%20 historical merged patches
Same-runner efficiency5.45×pytest 8.1657× · SymPy 5.4545×
Observed safety failures0stale success · shadow mismatch · cache conflict
FORMAL GATE

Both supported workloads passed.

  • ThresholdsAt least 5× same-runner compute efficiency and at least 50% p95 reduction.
  • pytest8.165714× compute efficiency and 78.415% p95 reduction.
  • SymPy5.454509× compute efficiency and 78.309% p95 reduction.
  • Safety observationsZero accepted stale-success events, shadow mismatches, or cache conflicts.
  • CandidateQualified runtime SHA: 0b3f1e7f3a3feb1b7e41f32704a3b558a3455e58.
SUPPORTED BOUNDARY

Reuse stays narrow.

  • PlatformLinux / amd64 only.
  • RuntimeImmutable OCI image identity pinned by digest.
  • SourceExplicit operator-reviewed source closure and content identity.
  • ResultPassing status for configured deterministic result-only targets; no stale artifact replay.
  • FallbackChanged, unknown, unsupported, or conflicting evidence executes fresh or is quarantined.
PYTEST RECEIPT

10 historical merged patches.

  • Direct total528,286.082 ms.
  • ZeroRun total64,695.641 ms.
  • Compute efficiency8.165714×.
  • p95 reduction78.415%.
  • Observed safety failures0.
SYMPY RECEIPT

20 historical merged patches.

  • Direct total3,788,203.578 ms.
  • ZeroRun total694,508.656 ms.
  • Compute efficiency5.454509×.
  • p95 reduction78.309%.
  • Post-edit hits60, with zero accepted shadow mismatches or cache conflicts.
PROVENANCE DISCLOSURE

The final SymPy receipt is explicit about two reviewed promotions.

Eighteen SymPy case rows come from the completed full 20-case execution. Two cases, PRs 20250 and 20590, use separately qualified exact-source-pair reviewed-equivalence evidence after the broad file-level harness conservatively invalidated all four partitions.

Those two reviewed decisions were bound to immutable pre-merge/merge source pairs and shadow-executed every claimed hit. Finalization replaced only those two rows, retained the other 18 unchanged rows, verified provenance, and ran the original aggregate and strict 5× / 50% gate unchanged. The public claim does not hide that methodology.

OPEN GATES

External proof starts now.

The internal 5× technical target is closed as passed. These are the milestones that remain genuinely open.

  1. 012–5 real developers or teamsUse ZeroRun on their own repositories and normal workflows.
  2. 02100+ genuine external requestsChronological edit → test requests with normal misses, failures and broad edits preserved.
  3. 03Design-partner reliabilityMeasure setup friction, bypasses, verification outcomes, support burden and repeated use.
  4. 04Commercial validationPricing, willingness to pay, retention, buyer ownership and paid pilots remain unproven.
PUBLICATION STATUS

Public product surface; private implementation.

floxy-21/zerorun publishes the product boundary, sanitized v0.4 evidence, pilot onboarding, roadmap and security channel.

The runtime implementation, benchmark harnesses, raw qualification receipts, private pilot build system and internal experimental history belong in the private implementation repository. Public documentation is not presented as reproducible source evidence.

NEXT EVIDENCE

Bring a real workload.

Request a private pilot