Context, memory & token cost

AI & agents › Context, memory & token cost · topic id ai/context-and-cost · 12 articles · generated, edit catalog.jsonl or notes/ instead

Cutting token spend, caching, context engineering and agent memory.

Picks to read (3)

• Understand Any Codebase Instantly (with up to 99.7% fewer tokens) — Agent Native
  understand-any-codebase-instantly-with-up-to-99-7-fewer-tokens · to-read ★ pick · 2026-11-07 · #token-cost · in issue 2026-11-07
• Context Engineering Is Dead. Long Live Cache Engineering. — Agent Native
  context-engineering-is-dead-long-live-cache-engineering · to-read ★ pick · 2026-11-06 · in issue 2026-11-06
  From a followed writer.
• Microsoft Dropped Claude Code Over Its Bill. Then Its Researchers Made It 31% Cheaper — Itai Spector
  microsoft-dropped-claude-code-over-its-bill-then-its-researchers-made · to-read ★ pick · 2026-11-04 · #claude-code #token-cost · in issue 2026-11-04

To read (9)

• Deep Research Agent That Remembers — Agent Native
  deep-research-agent-that-remembers · to-read · 2026-11-08 · in issue 2026-11-08
• 60–95% fewer tokens with Claude Code and Codex
  60-95-fewer-tokens-with-claude-code-and-codex · to-read · 2026-06-21 · #claude-code #codex #token-cost · in issue 2026-11-03
• GitHub: kytmanov/obsidian-llm-wiki-local
  gh-kytmanov-obsidian-llm-wiki-local · to-read · 2026-04-12 · #pkm #github-repo
• GitHub: IlyaGorsky/memory-toolkit
  gh-ilyagorsky-memory-toolkit · to-read · 2026-04-12 · #memory #github-repo
• Andrej Karpathy stopped using AI to write code he’s using it to build a second brain instead
  andrej-karpathy-stopped-using-ai-to-write-code-he-s-using-it-to-build · to-read · 2026-04-09 · #memory #pkm
• Claude Code’s second brain cuts token usage by 5x
  claude-code-s-second-brain-cuts-token-usage-by-5x · to-read · 2026-04-01 · #claude-code #token-cost #memory #pkm
• I cut my Claude Code API costs by 40% with one tool — @me_82386
  i-cut-my-claude-code-api-costs-by-40-with-one-tool · to-read · 2026-01-28 · #claude-code #token-cost
• RIP fine-tuning Stanford just blew it up with agentic context engineering
  rip-fine-tuning-stanford-just-blew-it-up-with-agentic-context-engineer · to-read · 2025-10-11
• The hidden costs of LangChain CrewAI PydanticAI and others why popular AI frameworks are failing
  the-hidden-costs-of-langchain-crewai-pydanticai-and-others-why-popular · to-read · 2025-07-15 · #token-cost #langchain