Episodes

  • Session Memory: Build Systems, Not Sessions in Claude Code

    Session Memory: Build Systems, Not Sessions in Claude Code
    Mar 8 2026
    Your Claude Code session hit 80% context. You compact or restart. Next session: "Can you read the project fil3es and understand the architecture?" Back to zero. Every time.This is not a Claude Code limitation. It is a workflow gap. After 110+ sessions across 4 projects, I built a free plugin that closes it - four commands that turn isolated sessions into a continuous development system.In this episode:- The gap between Claude Code's native MEMORY.md and actual session continuity- How 6+ existing memory plugins solve storage but miss the workflow problem- The four commands that create a session lifecycle: /project-status, /remember, /handoff, /context-stats- Why /handoff is the most important command - it generates continuity documents that the next session picks up automatically- How structured learnings with type, context, and tags create a searchable history- The difference between passive memory (read-only) and active session management- Zero dependencies: plain markdown and JSON, no MCP servers, no databases, no API keys- 60-second install that works on any Claude Code projectKey insight: recording what happened does not tell Claude what to do next. Indexing facts does not create session continuity. The missing piece is a lifecycle - a structured handoff between sessions that preserves momentum, tracks decisions, and suggests the next action. That is what separates a memory system from a session management system.The Claude Code Starter System plugin is free and open source on GitHub. It is the simplified version of the 650+ node knowledge graph system I use in production.---Read the full article: https://primeline.cc/blog/session-memory-pluginPlugin: github.com/primeline-ai/claude-code-starter-systemClaude Code Deep Dives is an AI-powered podcast by PrimeLine, generated with NotebookLM from in-depth technical articles. Each episode explores real production Claude Code setups.Website: primeline.ccX: @PrimeLineAI
    Show More Show Less
    19 mins
  • Session Management: Why I Stopped Restarting Claude Code

    Session Management: Why I Stopped Restarting Claude Code
    Mar 8 2026
    I injected a test marker into my auto-memory file mid-session. Next turn, Claude had no idea the file changed. That experiment changed everything about how I manage Claude Code sessions.Most developers restart Claude Code when context hits 80%. Feels productive. Costs $0.67 per restart in invisible cache rebuilds. Do that 10 times a day and you are burning $6.70 before writing a single line of code.In this episode:- The marker injection experiment that revealed how CLAUDE.md actually loads- What is in the cached prefix and what changes every turn- The real cost of session restarts vs. compaction (with actual numbers)- How the "prompt infusion" pattern works - system-reminder tags inject dynamic data without touching the cache- Why the first 3-5 turns of every new session are always the worst- The warm cache loop strategy that keeps your prefix alive between tasks- How handoff documents create session continuity without manual context loading- When restarting actually makes sense (and when it is pure waste)Key insight: Claude Code has two distinct layers in every session. The prefix (system prompt, tools, CLAUDE.md, rules) is frozen at session start and cached. The conversation layer (messages, tool results, hooks) changes every turn but never touches the prefix. Understanding this boundary is the difference between efficient sessions and expensive ones.Practical takeaway: use compaction instead of /clear, write handoff documents before closing sessions, and plan your tool set upfront. The warm cache strategy from this episode saves more money than any model switching optimization.Builds on Episode 1 (Prompt Caching) and leads into Episode 3 (Session Memory Plugin) which automates the handoff workflow.---Read the full article: https://primeline.cc/blog/session-managementClaude Code Deep Dives is an AI-powered podcast by PrimeLine, generated with NotebookLM from in-depth technical articles. Each episode explores real production Claude Code setups.Website: primeline.ccX: @PrimeLineAI
    Show More Show Less
    13 mins
  • Prompt Caching: The Hidden Rule Behind Every Claude Code Feature

    Prompt Caching: The Hidden Rule Behind Every Claude Code Feature
    Mar 8 2026

    AI-generated deep dive into Claude Code prompt caching - the hidden infrastructure rule that shapes every feature you use daily.


    Anthropic monitors prompt cache hit rate like infrastructure uptime. If it drops too low, they open an incident. Not a bug ticket - an incident. That single fact explains dozens of Claude Code design decisions that otherwise seem arbitrary.


    In this episode, we break down:


    - Why CLAUDE.md loads before your conversation (prefix stability)

    - How prompt caching works via byte-for-byte prefix matching

    - Why plan mode is a tool call, not a mode swap

    - The real cost of switching models mid-session (Opus to Haiku can be MORE expensive)

    - Why /clear destroys your cache but compaction preserves it

    - How system-reminder tags protect your cache hit rate

    - The 5-minute cache window and what happens when it expires

    - Why adding tools mid-session invalidates everything after that point


    Key insight: Claude Code assembles your prompt in a specific order - static system prompt and tools first, CLAUDE.md second, conversation messages last. This is not arbitrary. It is deliberately structured so the most stable content sits earliest in the prefix and stays cached across requests.


    The practical takeaway: stop rotating CLAUDE.md sections mid-session, stop adding tools on the fly, stop using /clear as a quick reset, and stop switching models when a task feels too small for Opus. Work with the prefix, not against it.


    This episode is based on the blog post at primeline.cc/blog/prompt-caching


    Claude Code Deep Dives is a podcast by PrimeLine (@PrimeLineAI) exploring real production setups for Claude Code - from memory systems and agent delegation to context management and self-correcting workflows.


    More at primeline.cc
    Show More Show Less
    22 mins