--- title: Anthropic wants Claude Code to be hackable. Here's... description: Anthropic is soliciting public feedback on unreleased Claude Code extensibility features, including Function Hooks for custom behavior via user-defined... canonical: https://daily.dev/posts/anthropic-wants-claude-code-to-be-hackable-here-s-what-s-missing-first--lewalogrn twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Anthropic wants Claude Code to be hackable. Here's what's missing first. | daily.dev og:description: Anthropic is soliciting public feedback on unreleased Claude Code extensibility features, including Function Hooks for custom behavior via user-defined... og:url: https://daily.dev/posts/anthropic-wants-claude-code-to-be-hackable-here-s-what-s-missing-first--lewalogrn og:image: https://api.daily.dev/og/posts/LEwalOgRn.png og:image:alt: Anthropic wants Claude Code to be hackable. Here's what's missing first. 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. # Anthropic wants Claude Code to be hackable. Here's what's missing first. **[Trends](https://daily.dev/sources/trends)** · 2 min read · 6 upvotes · 0 comments ## Summary Anthropic is soliciting public feedback on unreleased Claude Code extensibility features, including Function Hooks for custom behavior via user-defined functions, before shipping. Meanwhile, user Peter Yang highlights more pressing everyday friction: remote control not being on by default, context compaction requiring manual `/compact` commands, a possibly problematic 1M context window that discourages compaction, and browser automation lagging behind Codex. The piece argues Anthropic should clarify whether hooks address real workflow gaps or if basic usability bugs need fixing first. ## Content Anthropic is actively soliciting feedback on two unreleased Claude Code features: **Function Hooks** and a broader extensibility system that @bcherny calls "a little crazy, and very exciting." Neither has shipped yet, but the GitHub issue is open and the team wants input before they lock anything in. Meanwhile, the community isn't waiting. Power users have already built their own workarounds for the gaps that bother them most. **The context window problem is the loudest complaint.** @petergyang lists it as one of three major gaps: the 1M context window is "probably a mistake," compaction should be automatic, and remote control shouldn't require typing `/compact` manually. That last point has its own discourse thread — the short answer is: run `/compact` proactively, before you notice degraded responses, not after. One user built an Obsidian vault as external memory to stay under 4k tokens per session instead of loading everything upfront. **The verification gap is subtler but arguably more important.** Claude Code has a habit of declaring success after producing a plausible-looking fix without actually checking if it works. The `/goal` command pattern addresses this directly: you define measurable completion conditions (specific commands, exit codes, what shouldn't change), and a separate evaluator model checks the conversation evidence before Claude is allowed to stop. If evidence is missing, the evaluator turns the gap into Claude's next instruction. It's a stop hook, essentially, and it's the kind of thing that should probably be built into the tool. **The plugin ecosystem is already moving.** There's an open-source bridge that lets you delegate tasks to OpenAI Codex agents from inside Claude Code, with live progress streaming and session resumption. It runs on Claude Code's experimental Channels feature, which tells you something about how far ahead of the docs some of these builders are. The advanced features thread — custom sub-agents, skills, hooks, git worktrees for parallel sessions, `/rewind` for checkpointing — keeps surfacing as things "99% of devs skip." @lydiahallie is asking what should be surfaced better. The honest answer from the community seems to be: most of it. ## Questions this post answers ### What are Function Hooks in Claude Code? Function Hooks are an unreleased Claude Code extensibility feature that lets users customize and extend the tool's behavior through user-defined functions. Anthropic has published demo videos and opened a GitHub issue to gather public feedback before shipping, describing the broader hackability push as architecturally significant. _Developers deciding whether to adopt new Claude Code capabilities can follow how this extensibility feature develops on daily.dev._ ### Why do some users think Claude Code's 1M context window is a problem? A large 1M token context window can act as a footgun because it makes context compaction feel optional when it shouldn't be. Users assume the model can handle everything in that window, but the model eventually struggles anyway, and without a prompt to compact, nobody realizes they need to run the manual `/compact` command. _Anyone tuning AI coding workflows around context limits can track these tradeoffs on daily.dev._ ### How does Claude Code's browser automation compare to Codex for real tasks? Claude Code's browser use still lags behind Codex for practical tasks, such as editing a Google Doc, which is described as slow and unreliable. This is cited as one of several everyday friction points that matter more to typical users than upcoming extensibility features like Function Hooks. _Developers comparing Claude Code and Codex for real-world tasks can weigh these tradeoffs via daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 341 comments across x (as of 2026-09-10). **TL;DR:** Replies to the feature-discovery prompt turned into a mixed bag: some people shared genuinely useful hidden features (hooks, CLAUDE.md, /compact, remote control), while many others vented about usage limits, billing/support issues, account bans, and the harness lagging behind competitors like Codex and Cursor. **Sentiment:** 25% positive · 30% mixed · 45% skeptical **The case for** - Several people highlighted underused features that solved real pain points, like PostToolUse hooks replacing manual CLAUDE.md instructions, remote control for phone access, and /compact for controlling what context survives. - One person cited a concrete productivity comparison where a persistent session log and project memory led to 34% more PRs shipped over two weeks. - The Mac app and recent quality-of-life updates were praised as much improved from earlier versions. **The pushback** - Many commenters complained about usage limits, rate limits, and rapid consumption of subscription quota, with some checking usage more than their bank balance. - Multiple people reported unresolved account bans and slow or unhelpful customer support/billing disputes. - Several unfavorably compared speed, UI, and browser/task handling to Codex and Cursor. - One person described the model becoming more verbose and less instruction-following over time, pushing them toward switching tools entirely. - Concerns were raised that plugins silently consume context/usage even when not needed. **By community** - x (mixed): A mix of genuinely helpful feature tips sits alongside heavy venting about limits, billing, bans, and comparisons unfavorable to competitors. **Hottest debate:** Whether Claude Code's rate limits and pricing are reasonable value or a bait-and-switch, versus people praising specific underused features as genuinely useful. **Open questions** - Why do some features require both Thinking Mode and Verbose Output enabled just to see chain-of-thought? - Is there a way to trigger more than one user-only skill in a single prompt? - Will banked/saved weekly resets ever be supported instead of resets expiring immediately? **Highlights** > @lydiahallie the loop, not the model. 6 devs, vanilla claude code vs claude code + a long-lived session log + project-scoped memory. second group shipped 34% more PRs in two weeks, and reopened way fewer context-loss tickets. > — [elledynelabs on x · 2 points](https://x.com/elledynelabs/status/2097129994283880811) > @lydiahallie Hooks. I spent months putting "always run the linter" in CLAUDE.md before realising a PostToolUse hook just does it deterministically. They read as advanced in the docs, but they fix every instruction the model keeps forgetting. > — [KissonL on x · 1 points](https://x.com/KissonL/status/2097149329660363246) > @lydiahallie Idk, you've spent so much time and effort making claude more and more verbose, and less and less willing to actually follow instructions that i was still stuck on opus/sonnet 4.6 as the only actually usable model, until i gave up hope and finally switched to codex > — [siriusblack9999 on x · 2 points, 1 comments](https://x.com/siriusblack9999/status/2097090844209688727) > @lydiahallie Not a feature, a confession: I check my usage limit more often than my bank balance now. At least the bank offers overdraft. > — [thegitai on x · 1 points](https://x.com/thegitai/status/2097099872218517596) > @lydiahallie Didn’t know about split side view until today. Also didn’t know multiple chats could be going parallel, I thought it only accepted one chat at a time. > — [deydercintron on x · 1 points](https://x.com/deydercintron/status/2097106190865924415) **Source threads** - [x](https://x.com/lydiahallie/status/2097061523550322858) · 1 points · 341 comments ## Similar posts on daily.dev - [Claude Code felt overhyped until I discovered the workflow that makes it genuinely powerful](https://daily.dev/posts/claude-code-felt-overhyped-until-i-discovered-the-workflow-that-makes-it-genuinely-powerful-ellcs8hij) · XDA Developers · 0 upvotes · 0 comments --- Tags: [#devtools](https://daily.dev/tags/devtools), [#ai-agents](https://daily.dev/tags/ai-agents), [#anthropic](https://daily.dev/tags/anthropic), [#claude-code](https://daily.dev/tags/claude-code) [View this post on daily.dev](https://daily.dev/posts/anthropic-wants-claude-code-to-be-hackable-here-s-what-s-missing-first--lewalogrn) ```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":"Anthropic wants Claude Code to be hackable. Here's what's missing first.","url":"https://daily.dev/posts/anthropic-wants-claude-code-to-be-hackable-here-s-what-s-missing-first--lewalogrn","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/anthropic-wants-claude-code-to-be-hackable-here-s-what-s-missing-first--lewalogrn"},"datePublished":"2026-09-03T23:20:18.667Z","dateModified":"2026-09-10T20:20:13.671Z","description":"Anthropic is soliciting public feedback on unreleased Claude Code extensibility features, including Function Hooks for custom behavior via user-defined...","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":0,"discussionUrl":"https://daily.dev/posts/anthropic-wants-claude-code-to-be-hackable-here-s-what-s-missing-first--lewalogrn","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":6},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"devtools,ai-agents,anthropic,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":"Anthropic wants Claude Code to be hackable. Here's what's missing first."}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/anthropic-wants-claude-code-to-be-hackable-here-s-what-s-missing-first--lewalogrn#faq","mainEntity":[{"@type":"Question","name":"What are Function Hooks in Claude Code?","acceptedAnswer":{"@type":"Answer","text":"Function Hooks are an unreleased Claude Code extensibility feature that lets users customize and extend the tool's behavior through user-defined functions. Anthropic has published demo videos and opened a GitHub issue to gather public feedback before shipping, describing the broader hackability push as architecturally significant. Developers deciding whether to adopt new Claude Code capabilities can follow how this extensibility feature develops on daily.dev."}},{"@type":"Question","name":"Why do some users think Claude Code's 1M context window is a problem?","acceptedAnswer":{"@type":"Answer","text":"A large 1M token context window can act as a footgun because it makes context compaction feel optional when it shouldn't be. Users assume the model can handle everything in that window, but the model eventually struggles anyway, and without a prompt to compact, nobody realizes they need to run the manual `/compact` command. Anyone tuning AI coding workflows around context limits can track these tradeoffs on daily.dev."}},{"@type":"Question","name":"How does Claude Code's browser automation compare to Codex for real tasks?","acceptedAnswer":{"@type":"Answer","text":"Claude Code's browser use still lags behind Codex for practical tasks, such as editing a Google Doc, which is described as slow and unreliable. This is cited as one of several everyday friction points that matter more to typical users than upcoming extensibility features like Function Hooks. Developers comparing Claude Code and Codex for real-world tasks can weigh these tradeoffs via daily.dev."}}]} ```