--- title: Claude Code's memory feature is mostly junk, and the... description: A growing consensus in the Claude Code community argues that its automatic memory feature does more harm than good, with one developer finding two-thirds of... canonical: https://daily.dev/posts/claude-code-s-memory-feature-is-mostly-junk-and-the-community-is-turning-it-off-9iriij284 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Claude Code's memory feature is mostly junk, and the community is turning it off | daily.dev og:description: A growing consensus in the Claude Code community argues that its automatic memory feature does more harm than good, with one developer finding two-thirds of... og:url: https://daily.dev/posts/claude-code-s-memory-feature-is-mostly-junk-and-the-community-is-turning-it-off-9iriij284 og:image: https://api.daily.dev/og/posts/9iRIiJ284.png og:image:alt: Claude Code's memory feature is mostly junk, and the community is turning it off og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Claude Code's memory feature is mostly junk, and the community is turning it off **[Trends](https://daily.dev/sources/trends)** · 2 min read · 5 upvotes · 1 comments ## Summary A growing consensus in the Claude Code community argues that its automatic memory feature does more harm than good, with one developer finding two-thirds of saved memories were stale, redundant, or actively harmful, and most never read. Creators of Pi and Flask argue coding agents don't need memory since code itself is ground truth; PwC's analysis of Anthropic's harness primitives (loop, memory, filesystem, bash) frames the real engineering work as the harness and infrastructure layers around models. Practical advice favors architecting codebases to eliminate failure categories, adding CI/lint checks, and writing detailed values-oriented CLAUDE.md files over relying on accumulated agent memory. ## Content The discourse around Claude Code has shifted from "how do I use this" to "how do I stop it from making things worse." The specific target right now: automatic memory. The argument against it is pretty damning. One developer audited his own saved memories across machines and found roughly two-thirds were either stale, redundant with existing `agent.md` files, or actively harmful. Most were never read at all. He's now disabled memory across his entire fleet. The theoretical backing comes from a conversation between the creators of Pi and Flask, who argue that code is the ground truth and coding agents don't need a separate memory system. Just bash tools and context loaded on demand. Memory as a feature implies the agent needs to remember things that your codebase should already be expressing. This connects to a broader framing getting traction in enterprise circles. PwC looked at Anthropic's harness primitives - the loop with memory, filesystem, and bash that Claude Code is built on - and the takeaway is that the three-layer model (business logic, harness, infrastructure) is where the real work lives. Frontier models collapsed the cost of writing custom code, but they didn't collapse the cost of running it reliably in production. The practical advice landing well: instead of leaning on memory, architect your codebase to eliminate failure categories. Add CI and lint checks to catch regressions. Write detailed, values-oriented `CLAUDE.md` files. The Uncle Bob framing getting cited here is that imposing human values on agents matters more than imposing discipline - which is a more interesting distinction than it sounds. A memory system that saves "always use tabs" is discipline. A `CLAUDE.md` that explains *why* the team made certain tradeoffs is values. The consensus forming: memory is a shortcut that creates technical debt in your agent setup the same way shortcuts create technical debt in your code. The developers getting the most out of Claude Code right now are the ones treating their context files as first-class artifacts, not the ones letting the agent accumulate notes on its own. ## Questions this post answers ### Should I turn off Claude Code's automatic memory feature? Many developers are disabling it because saved memories tend to be stale, redundant with existing agent.md files, or actively harmful, with one audit finding roughly two-thirds of saved memories fell into that category and most were never even read. The recommended alternative is maintaining detailed, values-oriented CLAUDE.md files instead of letting the agent accumulate its own notes. _Developers weighing AI agent memory tradeoffs can follow this debate as it evolves on daily.dev._ ### What should I put in a CLAUDE.md file instead of relying on agent memory? Write values-oriented content that explains why the team made certain tradeoffs, rather than discipline-style rules like 'always use tabs.' The distinction, credited to an Uncle Bob framing, is that imposing values on an agent matters more than imposing discipline, since values generalize while rigid rules just accumulate as technical debt in the agent setup. _Teams refining their AI coding workflows can track emerging best practices like this on daily.dev._ ## Community discussion Top comments from developers on daily.dev. **@trevorsuna** · 0 upvotes > The useful distinction is between durable project knowledge and accidental session residue. Versioned context files plus CI checks are reviewable and reproducible; automatic memories need expiry, provenance, and a clear way to see whether they were ever used, or they quietly become another stale config layer. ## Similar posts on daily.dev - [Claude Code learns from my mistakes between sessions now, and my setup runs itself](https://daily.dev/posts/claude-code-learns-from-my-mistakes-between-sessions-now-and-my-setup-runs-itself-xpfy38kzw) · XDA Developers · 3 upvotes · 0 comments - [AI Agent Memory: Why Agents Forget and How to Fix It in 2026](https://daily.dev/posts/ai-agent-memory-why-agents-forget-and-how-to-fix-it-in-2026-t3vsrqsah) · Alex CloudStar · 0 upvotes · 0 comments - [I gave Claude Code memory between sessions, and my setup started running itself](https://daily.dev/posts/i-gave-claude-code-memory-between-sessions-and-my-setup-started-running-itself-55iakitq2) · XDA Developers · 0 upvotes · 0 comments --- Tags: [#architecture](https://daily.dev/tags/architecture), [#ai-agents](https://daily.dev/tags/ai-agents), [#anthropic](https://daily.dev/tags/anthropic), [#prompt-engineering](https://daily.dev/tags/prompt-engineering), [#claude-code](https://daily.dev/tags/claude-code) [View this post on daily.dev](https://daily.dev/posts/claude-code-s-memory-feature-is-mostly-junk-and-the-community-is-turning-it-off-9iriij284) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Claude Code's memory feature is mostly junk, and the community is turning it off","url":"https://daily.dev/posts/claude-code-s-memory-feature-is-mostly-junk-and-the-community-is-turning-it-off-9iriij284","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/claude-code-s-memory-feature-is-mostly-junk-and-the-community-is-turning-it-off-9iriij284"},"datePublished":"2026-08-28T01:49:36.440Z","dateModified":"2026-08-28T01:50:18.210Z","description":"A growing consensus in the Claude Code community argues that its automatic memory feature does more harm than good, with one developer finding two-thirds of...","isAccessibleForFree":true,"articleSection":"Trends","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/claude-code-s-memory-feature-is-mostly-junk-and-the-community-is-turning-it-off-9iriij284","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":5},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"architecture,ai-agents,anthropic,prompt-engineering,claude-code","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"Claude Code's memory feature is mostly junk, and the community is turning it off"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/claude-code-s-memory-feature-is-mostly-junk-and-the-community-is-turning-it-off-9iriij284","comment":[{"@type":"Comment","text":"The useful distinction is between durable project knowledge and accidental session residue. Versioned context files plus CI checks are reviewable and reproducible; automatic memories need expiry, provenance, and a clear way to see whether they were ever used, or they quietly become another stale config layer.","datePublished":"2026-08-28T13:44:07.550Z","url":"https://daily.dev/posts/9iRIiJ284#c-TAlKmMAZZ","author":{"@type":"Person","name":"Trevor Suna","url":"https://daily.dev/trevorsuna","image":"https://media.daily.dev/image/upload/s--dZ7gXxpp--/f_auto/v1784081551/avatars/avatar_EMoP47rpuw8DNjhp6R1b6?_a=BAMAMicg0"}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/claude-code-s-memory-feature-is-mostly-junk-and-the-community-is-turning-it-off-9iriij284#faq","mainEntity":[{"@type":"Question","name":"Should I turn off Claude Code's automatic memory feature?","acceptedAnswer":{"@type":"Answer","text":"Many developers are disabling it because saved memories tend to be stale, redundant with existing agent.md files, or actively harmful, with one audit finding roughly two-thirds of saved memories fell into that category and most were never even read. The recommended alternative is maintaining detailed, values-oriented CLAUDE.md files instead of letting the agent accumulate its own notes. Developers weighing AI agent memory tradeoffs can follow this debate as it evolves on daily.dev."}},{"@type":"Question","name":"What should I put in a CLAUDE.md file instead of relying on agent memory?","acceptedAnswer":{"@type":"Answer","text":"Write values-oriented content that explains why the team made certain tradeoffs, rather than discipline-style rules like 'always use tabs.' The distinction, credited to an Uncle Bob framing, is that imposing values on an agent matters more than imposing discipline, since values generalize while rigid rules just accumulate as technical debt in the agent setup. Teams refining their AI coding workflows can track emerging best practices like this on daily.dev."}}]} ```