Claude Code Daily Briefing - 2026-04-16

Release Summary

Version	Date	Key Changes
v2.1.110	4/15	`/tui` flicker-free fullscreen rendering, Push Notification tool, `autoScrollEnabled` config, `/plugin` tab improvements, `/doctor` MCP conflict warnings, `--resume` restores scheduled tasks, MCP hang/CPU fixes
v2.1.109	4/15	Extended-thinking indicator with rotating progress hint

Full release notes

New Features & Practical Usage

`/tui` — Flicker-Free Fullscreen Terminal Mode (v2.1.110)

v2.1.110 introduces the /tui command and tui setting for flicker-free fullscreen rendering. This eliminates the visual flickering that occurred during rapid output scrolling in long sessions, providing a much more stable visual experience.

# Switch to fullscreen TUI mode
/tui

# Enable by default in settings.json
# { "tui": true }

# Disable auto-scroll in fullscreen (for reviewing long output)
# { "autoScrollEnabled": false }

Ctrl+O has been simplified to toggle between normal and verbose transcript views only. A new option also lets you see Claude’s last response as commented context in your external editor. Particularly useful when reviewing large code changes or monitoring long-running agent tasks.

GitHub v2.1.110

Push Notifications — Mobile Alerts from Claude Code (v2.1.110)

v2.1.110 adds a Push Notification tool that sends alerts to your mobile device when long-running tasks complete or require your attention.

Previously, you had to keep watching the terminal or wire up third-party notification services. Now Claude Code can push notifications directly. Combined with Routines or /loop, you can monitor automated tasks on the go without staying glued to your terminal.

GitHub v2.1.110

Developer Workflow Tips

Claude Code + Codex Hybrid Workflow — Best of Both Worlds

A detailed comparison post on GeekNews (39 points) by a 14-year veteran Principal/Staff Engineer from a MAG7 company tested both tools extensively on a large Python/TypeScript project (~100 hours with Claude Code, ~20 hours with Codex).

Key findings:

Claude Code (Opus 4.6): Fast and interactive, but tends to ignore instructions and leave tasks incomplete. “Speeds through implementation without sufficient forethought.”
Codex (GPT-5.4): 3–4x slower but methodical and intentional. Rigorously follows instruction files (AGENTS.md) and performs spontaneous refactoring.

Community consensus — hybrid workflow:

Claude Code plan mode → Sonnet implementation → Codex review → repeat

Rather than going all-in on one tool, the most effective pattern is role-based division: Claude’s speed for rapid prototyping and implementation, Codex’s thoroughness for review and quality assurance.

GeekNews

`--resume` Now Restores Scheduled Tasks (v2.1.110)

In v2.1.110, --resume and --continue automatically restore unexpired scheduled tasks. Previously, resuming a session would lose any /loop or CronCreate schedules. Now your active monitoring workflows survive session restarts.

Combine with /recap to restore both “what was I working on” and “what was I watching” in a single session recovery.

Security & Limitations

Major Service Outage on April 15 — Claude.ai, API, and Claude Code Down Simultaneously

A major outage hit all Claude services on April 15, lasting approximately three hours with global impact.

Timeline:

10:42 AM ET: ~6,000 users reported issues (DownDetector)
10:53 AM ET: Anthropic acknowledged the issue
11:03 AM ET: First fix deployed, brief recovery
11:40 AM ET: Second failure — “Claude.ai and Platform are down”
1:42 PM ET: Full recovery confirmed

Login failures, chat interruptions, and usage limit glitches were reported. Already-logged-in Claude Code users could continue working, but new logins were broken. This is part of a recurring pattern of outages since early 2026, attributed to surging demand outpacing infrastructure scaling.

TechRadar | CNBC

Performance Decline Controversy — Default Effort Level Quietly Reduced to “Medium”

Fortune reports that Anthropic changed Claude’s default effort level from “high” to “medium” in early March, triggering widespread user complaints about degraded quality.

User complaints:

Increased instruction non-compliance, inappropriate shortcuts, more mistakes on complex tasks
Shift from “research-first” (gathering context before acting) to “edit-first” (reading less context, making more errors)
One analyst described Claude Code as having become “unusable for complex engineering tasks”

Anthropic’s response (Boris Cherny, Claude Code lead):

Changes were documented in the changelog, responding to user feedback about excessive token consumption
Reasoning capability wasn’t reduced, just made less visible to users
Enterprise users will be defaulted to high effort going forward

# Immediate fix: manually set effort level
claude --effort high

The backlash threatens Anthropic’s brand as a more transparent, user-aligned company — a key differentiator as a rumored IPO approaches.

Fortune

Ecosystem & Plugins

skills-cleaner — Track and Manage Claude Code Skill Usage

A new Claude Code plugin shared on GeekNews Show. It tracks active skill usage logs and provides a /profile-skills command to analyze current skill consumption patterns.

As skills accumulate, session startup token costs increase. This tool helps identify and remove unused skills to reduce token consumption — particularly valuable if you’re hitting Pro Max quota limits.

GitHub

Routines Daily Limits Confirmed — Per-Plan Breakdown

The daily execution limits for Routines (announced 4/14) have been confirmed: Pro gets 5 runs/day, Max gets 15, Team and Enterprise get 25. Currently in research preview. Supports 20+ trigger types including schedule, API, and GitHub events. Routine ownership is personal and cannot be shared with teams.