Mixmod
Frontier supervisor plus local worker, with reported token reduction and a clear runtime caveat.
Local prep - Not published
Agent Orchestration Proof Kit
AIM Labs is a proof harness for agent orchestration. It turns local agent work into handoffs, receipts, ledgers, caveats, and public-safe proof surfaces before anyone claims a benchmark, adoption curve, or release.
- GitHub Pages: Not published
- PyPI: Not published
- External writes: Off
- Model calls: Off
Market signal
Mixmod proves the wedge is live. Aimee proves the category is broader.
Mixmod is the narrow benchmark reference: GPT-5.5 supervising a local Qwen worker on selected coding tasks. Aimee is the broader adjacent platform: memory, code graph, delegates, guardrails, and local services. AIM Labs sits in the proof layer: what happened, what was gated, what was redacted, and what can be shown safely.
Aimee
Local memory, code graph, delegates, model routing, and guardrails as a broader platform play.
AIM Labs
Receipts, Terrarium handoffs, validation reports, authority gates, and public-safe proof bundles.
Cost/context ledger
Token savings are not enough.
The packaged fixture agent_orchestration_proof_kit.json keeps
token counts beside wall-clock time, worker bytes, quality result, proof strength, and limitations.
That is the difference between a useful benchmark and a shiny number without a receipt.
Reported
Mixmod numbers are marked as reported because AIM Labs has not independently reproduced them.
Fixture
The AIM local proof-spine row proves the receipt shape, not live coding quality or token savings.
Next
Import a Mixmod-style run artifact, rerun what can be rerun, and label every metric by proof strength.
Non-claims
No victory lap from borrowed signal.
This kit does not claim AIM Labs beat Mixmod, does not claim public adoption, does not claim a PyPI release, and does not claim GitHub Pages is public. It prepares the local proof story and keeps the publish step separate.