A thousand tools, one small backpack
Put Codex, Claude Code, Hermes Agent, and DeepSeek-style Code Mode beside the same hypothetical 50,000-tool company and compare what actually reaches the model.
→
~/kforcode$ ls writing/
Writing
Field notes on production AI, agent runtimes, evaluation, retrieval, and ML systems. Written from implementation experience, with the trade-offs left in.
Open model accuracy is a harness problem
Open model accuracy is also a systems problem. Hermes Agent narrows tool choices, repairs safe mistakes, preserves context, and helps smaller models finish.
hermes-agent architecture: caching outranks compaction
The hermes-agent architecture at a glance: why the largest open-source agent harness ships both of its cheap context reducers switched off, and what that buys.
hermes-agent internals, one component at a time
A component-by-component study of hermes-agent internals: the loop, the prompt, the cache plan, compaction, state, approvals, and what each one costs.
The system layer behind AI products: what an agent runtime actually does
AI demos are easy; AI products are hard. The gap is a runtime — the shared execution layer that handles model routing, tool authorization, long-context state, memory, and fallback. Here is what it takes to run one at scale.
From 86% to 97%: engineering evidence retrieval for compliance agents
Compliance answers are only trustworthy if every claim cites the right evidence. Here is how hybrid retrieval, reranking, control-aware chunking, and citation-span ranking took evidence F1@10 from 86% to 97%.
One model instead of 23: consolidating a Document AI stack
A model per document type does not scale. Here is how a single layout-aware extraction architecture replaced 23+ document-specific models, cut OCR spend 70%, and made new document types cheap to add.
No posts match this filter.