Claude Code can fully restore a conversation from local logs after you close the terminal—even remembering which branch your Git working directory was on. We think what's really worth noting here isn't that "it can continue," but that AI tools are starting to possess "working memory" like humans do.
What this is
Claude Code is Anthropic's command-line AI programming assistant—it chats with AI directly in your terminal to write code. Its session resume does two things: first, it stores the conversation to local storage as an append-only log; then, when you next type, it assembles the previous context and sends it to the model.
The technical key is a JSONL file (a line-by-line append log format where each line is an independent record). Every conversation turn, tool call, and config change is only added to the end—old content is never rewritten. When you use /rewind to jump back to an earlier message and continue, the old path doesn't disappear; both old and new paths coexist—like a small branching tree. When resuming, the system traces back along the pointers between messages, skipping records that don't belong to the current path.
Three usage modes: claude --continue picks up the most recent session in the current project; claude --resume <id> targets a specific session; /resume switches between sessions already opened. Sessions are stored separately by project, so different projects don't interfere with each other.
Industry view
Supporters see this as a key step in AI tools becoming "engineered." Using AI used to feel like talking to a stranger—every time you had to re-explain the background. With resumable sessions, AI finally has a chance to become a long-term collaboration partner. Some developers who have used the feature report that the time saved from not having to re-explain things is more noticeable than improvements from model upgrades.
But there are also sober dissenting voices. An independent developer commented on Hacker News: session resume is genuinely useful for short tasks, but on complex projects what engineers care about is whether the AI understands the entire codebase architecture—not whether it remembers where yesterday's conversation left off. Logs accumulate over time and eat disk space; the official product currently has no automatic cleanup policy. The bigger questions sit in data: do sessions stay local or in the cloud? Does the enterprise version include auditing? Anthropic hasn't been clear enough on these compliance issues—this is homework enterprises must fill in themselves before deployment.
Impact on regular people
For enterprise IT: Resumable sessions lower the barrier to deploying AI tools inside the enterprise, but IT departments need to start taking AI tool log retention, data sovereignty, and compliance audit seriously—this is no longer a personal tool, but part of enterprise assets.
For working professionals: For sales, operations, and content roles that require continuous follow-up, once an AI assistant has "working memory," its value will significantly exceed single Q&A. Not having to re-explain customer background or project progress every time—that alone is real efficiency.
For the consumer market: Consumer-side phone voice assistants haven't yet widely adopted this kind of persistent memory. Once they do, conversations with AI will no longer be "use and discard," and public debate around privacy boundaries will heat up—recall the controversies around smart speakers back then; they'll likely be replayed.