CONTEXT ENGINEERING

Agent memory is five layers: history, compaction, semantic recall, working memory and files. How Mastra, LangGraph, OpenAI and Claude implement each.

Context engineering for agents, explained: caching, compaction, tool-result clearing, retrieval and subagents, with the exact Anthropic and OpenAI API knobs.

The text agents read is a control surface: bad MCP error messages cost more on stronger models, and context hygiene gains value with every model upgrade.

Agent Retrieval Bench isolates the part of coding-agent work most evals hide: did the agent find the right repository files before it started editing?

CodeNib treats repository context as compiled infrastructure, so coding agents stop rediscovering the same repo through grep and reads. What it builds.

The new MCP roadmap prioritizes progressive discovery for large tool catalogs and standardized agent identity. What changes for developers building MCP servers.

CLAUDE.md and AGENTS.md files keep growing because the reasoning behind each rule decays. A study of 1,867 repos names the mechanism and a comment-based fix.

Agent memory is moving into the model: in-backbone memory, parametric memory, reconstruction and transactional writes challenge external stores. What to do now.

AGENTS.md and CLAUDE.md configuration smells: a SCAM 2026 study of 100 top repos catalogs six, led by Lint Leakage and Context Bloat. What to fix in yours.

AGENTS.md and context files do not measurably change coding agent correctness in a 288-run Claude Code and Codex ablation. Failures were implementation skill.

OwlPath encodes code into an OWL2 ontology so coding agents can answer multi-hop structural queries with SPARQL, cutting tokens 28.8% on SWE-bench Pro.

Wiki skills add the missing graph layer to agent context: skills that link to each other let agents navigate knowledge instead of front-loading it all.

Anthropic cut over 80% of Claude Code's system prompt for Claude 5 models with no coding eval loss. The new context engineering rules and what HN thinks.

Anthropic removed over 80% of Claude Code's system prompt for Claude 5 models. Here is how the rules changed and what it means for your CLAUDE.md files, skills, and system prompts.

A deep dive into why fully autonomous AI coding agents degrade codebases over time, and what context engineering can actually fix.

SWE-Pruner Pro prunes coding-agent tool outputs inside the runtime before they become next-turn context. What it does, its limits, and the pattern to copy.

Harness engineering is where self-improving AI starts, Lilian Weng argues, not models rewriting their weights. What that means for developers building agents.

Fable 5 1M context workflows that actually work: whole-repo reviews, log archaeology, multi-doc synthesis - plus the honest math on when RAG still wins.

AI agent memory works best as a context ledger: source-linked, scoped, expiring memory that agents can inspect and users can audit, not magic recall.

Domain expertise is the agentic coding moat only when it becomes executable: examples, tests, DSLs, and review gates that turn tacit judgment into checks.