TEJL AI runtime

The software that turns LLMs into AI

200,000 €

Lifelong support included

Current status

Verified today: 21 capabilities, offline fixtures, reproducible by you in one day.

In soak since 13.11.2026: ledger (7 days), board (14 days), supervisor recovery. Results are published when they finish.

Not claimed: performance, financial return, fleet operation, desktop app.

Evidence date: 12.11.2026. All listed proofs concern offline fixtures or mocked boundaries. They do not establish live provider service, production credential isolation, multi-day operation or a deployed desktop.

Evidence ledger · PROVEN only

Verified capabilities

Each capability retains its evidence ID and proof scope.

CAP-033cards/workers
The harness rejects cards with missing checks, changed rule prefixes or an ownership scope that misses referenced files.
exit 0; SHA256 `928b8d649860`; 50 tests. Offline fixtures; scope limited to this claim.
CAP-037ledger/budget
The pool counts daily FreeLLMAPI usage and requests a smaller card when its tier token ceiling is exceeded.
exit 0; SHA256 `21b8fad6ba9a`; 6 tests. Offline fixtures; scope limited to this claim.
CAP-038ledger/budget
The attribution report records the served model, account, task class, size and actual acceptance without charging local blockers against provider success.
exit 0; SHA256 `3d740d28fd85`; 26 tests. Offline fixtures; scope limited to this claim.
CAP-039trace/receipts
The quality ledger joins fixture worker, trace and result evidence into repeatable per-class receipts.
exit 0; SHA256 `687fc2cf530d`; 18 tests. Offline fixtures; scope limited to this claim.
CAP-040routing/providers
The quality router recommends eligible models in shadow mode from acceptance and token evidence.
exit 0; SHA256 `1438b8812ac0`; 16 tests. Offline fixtures; scope limited to this claim.
CAP-041ledger/budget
The business reporter keeps money units distinct and rejects invalid or unsupported evidence.
exit 0; SHA256 `2b8b60bf5289`; 9 tests. Offline fixtures; scope limited to this claim.
CAP-042self-improvement
The incident reporter requires evidence for closure and detects expired waivers and repeated incidents.
exit 0; SHA256 `21a266e3409c`; 12 tests. Offline fixtures; scope limited to this claim.
CAP-043cards/workers
The cohort builder records whether a task class was explicitly declared or inferred.
exit 0; SHA256 `4190e6532da5`; 7 tests. Offline fixtures; scope limited to this claim.
CAP-045self-improvement
TIDE builds fixture cohorts from card outcomes and model evidence.
exit 0; SHA256 `94cf46fc5154`; 8 tests. Offline fixtures; scope limited to this claim.
CAP-046routing/providers
TIDE ranks fixture candidates using cohort evidence, tier limits and availability filters.
exit 0; SHA256 `de84019e9b07`; 11 tests. Offline fixtures; scope limited to this claim.
CAP-048self-improvement
TIDE produces promotion or demotion proposals with sample thresholds and confidence bounds.
exit 0; SHA256 `b8d3abaf33f5`; 5 tests. Offline fixtures; scope limited to this claim.
CAP-049self-improvement
TIDE evaluates canary qualification from measured cohort sample counts.
exit 0; SHA256 `683f2ac29df7`; 9 tests. Offline fixtures; scope limited to this claim.
CAP-050scheduling
The free-pool regulator raises or lowers concurrency from injected-clock feedback and holds a reduced limit after throttling.
exit 0; SHA256 `dc51b8c96c2d`; 1 explicit assertion function. Offline fixtures; scope limited to this claim.
CAP-051scheduling
The slot manager enforces a configured concurrent-attempt limit and ends a waiting request when its card is stopped.
exit 0; SHA256 `31d461d4f5ea`; 8 tests. Injected clock and fake sleep; no elapsed-time assertion or live provider.
CAP-052routing/providers
The pool router filters fixture models by tools, context, cost eligibility and cooldown state.
exit 0; SHA256 `66a2360ac265`; 18 tests. Offline fixtures; scope limited to this claim.
CAP-053grants/custody
The pool router’s SQLite reader refuses secret columns and tables outside its allowlist.
exit 0; SHA256 `b1fe735d8b36`; 7 tests. Offline fixtures; scope limited to this claim.
CAP-054observability/dashboard
The pool rollup totals fixture attempts, failures, limit events and tokens by model.
exit 0; SHA256 `fa97c80fc3eb`; 3 tests. Offline fixtures; scope limited to this claim.
CAP-055gates/review
The verification bot reruns a card check and refuses a claimed success when tests, scope or protected test bytes disagree.
exit 0; SHA256 `c92392fe6669`; 14 tests. Offline fixtures; scope limited to this claim.
CAP-057grove/VCS integration
Grove admission guards reject malformed place bindings, dirty workspaces and lagging mirrors in offline fixtures.
exit 0; SHA256 `5591b4db994a`; 13 tests. 13 unit tests; grove process calls are mocked and real-binary landing tests were not run.
CAP-060desktop HQ
The installed desktop launcher forwards only the resolved executable and environment and reports a generic refusal on failure.
exit 0; SHA256 `8eae2b9341a0`; 3 tests. 3 tests with resolution and process replacement mocked; native launch not exercised.
CAP-061install/ops
The desktop build wrapper constructs isolated job environments, validates build arguments and preserves compiler exit codes.
exit 0; SHA256 `05af07f57f34`; 18 tests. 18 tests; compiler and host checks mocked, shell argument forwarding exercised; no native build qualified.