[{"data":1,"prerenderedAt":45},["ShallowReactive",2],{"$f2xyuvmJ0Q9oH6aqgsULMtUcsiMP5LHr-Q6GeB1MUO-0":3},{"date":4,"generated_at":5,"picks":6,"candidates_scanned":43,"candidates_scored":44},"2026-04-16","2026-04-16T06:00:00.000000+00:00",[7,21,34],{"rank":8,"title":9,"source":10,"url":11,"category":12,"tldr":13,"score":14,"scores":15,"why":20},1,"v2.1.110","Claude Code Releases","https://github.com/anthropics/claude-code/releases/tag/v2.1.110","Release","- Run `/tui fullscreen` to switch to flicker-free rendering in the same conversation — finally fixes the flickering that made fullscreen hard to use in long sessions\n- Claude can now send mobile push notifications when Remote Control is enabled — your phone pings you when a long-running task finishes\n- The Write tool now tells Claude when you edit its proposed content in the IDE diff before accepting — Claude sees your manual edits and can reason about why you changed them\n- `/autocompact`, `/context`, `/exit`, and `/reload-plugins` now work from Remote Control (mobile/web) — full session control from your phone\n- `--resume`/`--continue` now resurrects unexpired scheduled tasks that were interrupted\n- Session recap is now enabled for Bedrock, Vertex, and Foundry users (previously only opt-in); opt out via `/config` or `CLAUDE_CODE_ENABLE_AWAY_SUMMARY=0`\n- Bash tool now enforces its documented maximum timeout instead of accepting arbitrarily large values",76,{"direct_claude_relevance":16,"practical_utility":17,"novelty":18,"source_credibility":19},33,22,8,13,"Push notifications and TUI mode are the standout additions: push notifications close the loop on background tasks without keeping a terminal open, and flicker-free TUI rendering makes fullscreen mode actually usable for long sessions. The Write tool IDE diff awareness is quietly high-value — Claude will now know when a human manually overrides its proposed edits mid-session, which changes how it reasons about subsequent steps. Novelty is penalized because this is the seventh consecutive day of Claude Code release coverage, but the push notification system and TUI mode are genuinely new capabilities not present in earlier releases.",{"rank":22,"title":23,"source":24,"url":25,"category":26,"tldr":27,"score":28,"scores":29,"why":33},2,"What's eating your Claude Code context window? I wrote a 500-line Python script to find out","Dev.to Claude","https://dev.to/geniej/whats-eating-your-claude-code-context-window-i-wrote-a-500-line-python-script-to-find-out-3oma","Tutorial","- `cc-healthcheck` is a single Python file, zero dependencies, that reads `~/.claude/` locally and tells you three things: what auto-loads into every session (CLAUDE.md chain, rules, skill frontmatter), whether your hooks are broken, and where last session's tokens actually went\n- Shows per-model token totals, cache hit ratio, and system-reminder injection counts — so you can see if your prompt cache is actually working\n- Catches common hook bugs: unquoted `|` operators, missing timeouts, and case-sensitivity traps that silently fail",61,{"direct_claude_relevance":30,"practical_utility":17,"novelty":31,"source_credibility":32},25,9,5,"Context budget visibility has been a gap in Claude Code's tooling — `/context` shows totals but not breakdown by source. This fills that gap with a zero-install script that reads your local session files and surfaces exactly which files and hooks are costing you tokens. The hook validation is especially useful: unquoted pipe operators in hooks are a documented corruption risk that Claude Code doesn't warn about. Dev.to credibility is modest, but the methodology is reproducible and the tool is on GitHub with no network calls.",{"rank":35,"title":36,"source":24,"url":37,"category":26,"tldr":38,"score":39,"scores":40,"why":42},3,"Architectural drift with Agentic coding. Here's what I built to fix it.","https://dev.to/deepcodersinc/agentic-coding-architectural-drift-heres-what-i-built-to-fix-it-4h2j","- `unkode` is a Claude Code skill that generates a `unkode.yaml` architecture map from your codebase — commit it to main as a versioned baseline\n- On each PR, it diffs the current module graph against that baseline and flags new cross-layer dependencies (\"Auth now imports from Billing — is that intended?\")\n- Solves a real agentic coding problem: Claude Code can add 14 imports across 8 files in one PR, and the diff is too big to catch the architectural implication by eyeballing it",52,{"direct_claude_relevance":17,"practical_utility":41,"novelty":31,"source_credibility":32},16,"Architectural drift is the under-discussed downside of high-velocity AI-assisted coding: Claude Code ships features fast, but the module boundaries that make a codebase maintainable erode one import at a time. Storing a YAML architecture map in the repo and diffing it per-PR is a simple, reviewable approach that doesn't require a SaaS dashboard. The unkode skill is open source and installable. Source credibility is low (personal Dev.to post), but the problem statement is real and the solution is concrete.",38,20,1776402242933]