Invitation: open Memory Tournament: review your setup before we run

[Codejain1]

Hi Hindsight team,

We're running an open tournament for AI memory systems, judged on ten competencies any agent needs from memory: recall, updates, time, multi-hop, abstention, learning from experience, acting on memory, forgetting on request, provenance/isolation, and cost at scale. Hindsight is one of 12 contestants, tested on its open-source edition, self-hosted.

Before anything runs, we'd like your team to check how we've set Hindsight up (version, settings, the LLM its writes use) in a private channel on our Discord: https://discord.gg/sugxNYp9rS
Fixes suggested before the freeze (9 Nov 2026) apply to the run.

Fair play: public benchmarks first (LongMemEval, LoCoMo, MemoryAgentBench, BEAM, MemoryArena); our own probe tests capped at 30% and reviewed; the same answer model for every system; two blind judges from other model families; raw per-question results published; a dispute window after each stage.

Disclosure: we build Horizon, which competes under the same rules. No reply is fine: we'll still test the open-source edition as published.

Thanks,
Kartik