The latest articles on Forem by Dominik (@dominikj111). https://forem.com/dominikj111 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F1003726%2Fe007eb47-ceba-408f-8828-2d530b2d3b9b.png https://forem.com/dominikj111 en Dominik Fri, 17 Apr 2026 05:51:05 +0000 https://forem.com/dominikj111/llms-are-excellent-at-novelty-operations-reward-determinism-2g8l https://forem.com/dominikj111/llms-are-excellent-at-novelty-operations-reward-determinism-2g8l <p>Most production queries aren't novel — they're the same error signatures, the same workflow branches, the same resolution paths. Re-deriving that reasoning through a full model call every time is avoidable overhead.</p> <p><strong>Engram</strong> is a design proposal for a deterministic layer that sits in front of LLMs:</p> <ul> <li>Queries hit a confidence-weighted graph first</li> <li>High-confidence paths return answers directly — no model call</li> <li>Novel cases escalate to the LLM; confirmed answers write back as reusable paths</li> <li>The graph accumulates knowledge across sessions; model calls decrease over time</li> </ul> <p>The same architecture covers an agent mesh, a structured tool gateway with policy enforcement (guard-rails by architecture, not instruction), and persistent memory for LLM agents via MCP.</p> <p><strong>Early-stage</strong> — Phase 1 of 15 — published as a design proposal, not a product launch. Full architecture, trade-offs, and open questions in the article.</p> <h2> References </h2> <ul> <li><a href="https://dominikj111.github.io/blog/engram-deterministic-operations-layer-for-llm-agent-workflows/" rel="noopener noreferrer">Full article</a></li> <li><a href="https://dominikj111.github.io/engram/" rel="noopener noreferrer">Engram site &amp; simulations</a></li> </ul> agents ai architecture llm