The ledger blocked a harmless ls, because the answer was already known
The guard only fires when the answer is already known, and its deny message is the fix. We have not measured whether it stopped the repeats.
Posts with this tag: 13
The guard only fires when the answer is already known, and its deny message is the fix. We have not measured whether it stopped the repeats.
A one-word last line carries no task. The crash list gives the session ids, and only two of its ten entries point to a transcript we can read from.
A shared frozen instinct plus a small owned layer per NPC, tested against a kill line we wrote before running anything. The brain passed the line and still lost to the script it came from.
NPC-mind · part 1Two decisions dated the same day disagree about whether the network ships. The game as built runs the bandit, which pays on measured costs and holds up under police and noise. The network returns only when a question needs more than three answers.
NPC-mind · part 5A frozen base brain copied from the script, an evolved readout on top of it, and a masked owned layer that each NPC changes during its life. We also describe the checks that stop a broken copy from passing.
NPC-mind · part 2Foresight lost to a one-step rule, a 12-word vocabulary carried no meaning, and a trail map reached 9 of 100 NPCs. The game log showed a threatened NPC whose needs did not move after 31 threats, so the network had nothing to learn from.
NPC-mind · part 4At tick 1000 we made one habit of the NPCs start to hurt. The script fell to 0.051 on the second half of its life, the frozen instinct to 0.211, and the learners recovered about 0.10 more than the instinct did.
NPC-mind · part 3A small learned memory predicted which files an agent would touch next at recall@5 0.599, against 0.534 for the usual rule. In a live A/B over 1,090 matched turns it changed nothing we could detect. Seven rules came out of the same week's rebuild and lock work.
BrainNet · part 7On our own vault, every network idea tied plain search at 23 of 29 and 17 of 30. Bridges, usage synapses, learned weights, Hebbian plasticity and a graph inside the model were all killed against a kill line set in advance.
BrainNet · part 5Spreading activation recovered 13.0 percent more held-out Wikipedia links on 1.18M articles, a local model found 42 extra hidden note links, and hub damping cut hub pages in paper answers from 22 slots to 9. The accuracy gain in brain-ask itself came from the reranker.
BrainNet · part 4Agents must recall before they hypothesize. The vault's rules, a walk tool, skill hooks and a per-repo network turn the memory into something an agent uses every turn, including on an older laptop without Mojo.
BrainNet · part 6brain-ask finds seed notes by meaning, spreads activation along synapses with personalized PageRank, then lets a cross-encoder pick five. On the Wikipedia network we cut a question from 5.3 s to 0.69 s by firing 2,618 neurons instead of 1,175,258.
BrainNet · part 3Our memory is a vault of Markdown notes. BrainNet reads it as a graph, with notes as neurons and links as synapses, then merges notes, papers, code and docs into one network of 1,640,443 neurons.
BrainNet · part 1