SYSCALLAI // BUSINESS-SYSTEMS LAB
Reliable AI, even when the model fails.
End-to-end systems that read business documents faithfully, then gate every output through an adversarial check — document honesty in, an adversary on the way out.
See the systems →SCROLL
LLMs fail in nameable ways.
- F-01 Reversal Curse Learns "A is B" yet fails to recall "B is A".
- F-04 Confabulation Invents plausible facts to fill a gap.
- F-07 Prompt Injection Treats attacker text in retrieved content as a command — and acts on it.
What we do about it.
We don't wait for models to stop failing — we build the correction outside the model. Each failure mode gets a system that catches it before it can act: claim-level verification (Bastion), an independent adversarial validator (Nemesis), a deterministic inference-time interlock (OSM), causal-context compression (C3). The model can fail; the system holds.
Operational State Models — the category definition →
Building something that needs to hold? Start a conversation →