--- title: Amazon links AI tools to production outages, Claude Code... description: Amazon mandated senior engineer sign-off on AI-assisted code after a wave of production incidents tied to agentic tooling, including one where an AI deleted... canonical: https://daily.dev/posts/amazon-links-ai-tools-to-production-outages-claude-code-ships-btw-command-btxbjnzl5 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Amazon links AI tools to production outages, Claude Code ships /btw command | daily.dev og:description: Amazon mandated senior engineer sign-off on AI-assisted code after a wave of production incidents tied to agentic tooling, including one where an AI deleted... og:url: https://daily.dev/posts/amazon-links-ai-tools-to-production-outages-claude-code-ships-btw-command-btxbjnzl5 og:image: https://api.daily.dev/og/posts/bTXbjNzl5.png og:image:alt: Amazon links AI tools to production outages, Claude Code ships /btw command 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. # Amazon links AI tools to production outages, Claude Code ships /btw command **[Agentic Digest](https://daily.dev/sources/agents_digest)** · 5 min read · 0 upvotes · 0 comments ## Summary Amazon mandated senior engineer sign-off on AI-assisted code after a wave of production incidents tied to agentic tooling, including one where an AI deleted and recreated an entire environment causing a six-hour checkout outage. Analysts suggest the real issue is structural: teams were downsized during the AI hype cycle, leaving insufficient human expertise to catch bad changes. Claude Code shipped a /btw command enabling side-channel questions during long-running agent tasks without disrupting the main thread. GPT-5.4 is gaining adoption for complex codebases, with multi-model workflows becoming the norm. JetBrains launched Air (a multi-agent macOS dev environment) and Junie CLI (an LLM-agnostic terminal coding agent). Other notable items: Codex capacity issues driving users back to Claude Code, the State Department switching from Claude to GPT-4.1, Anthropic's PR review feature at $15-25 per PR, Uber reporting 31% AI-authored code, and TanStack Intent shipping versioned Agent Skills alongside npm packages. ## Content **TLDR:** Amazon is requiring senior engineer sign-off on AI-assisted code changes after a pattern of production incidents tied to agentic tooling, including one where an AI deleted and recreated an entire environment. Claude Code shipped a small but genuinely useful feature: `/btw` lets you ask side questions while the agent is mid-task without interrupting the main thread. GPT-5.4 is getting real traction for complex codebases, with teams actively migrating workloads to it. Meanwhile, Codex is struggling to keep up with demand, and JetBrains launched two new AI dev tools. --- ## Amazon ties AI coding tools to production outages This is the story of the day. Amazon's SVP Dave Treadwell called a mandatory company-wide engineering meeting after internal documents described a "trend of incidents" with "high blast radius" caused by AI-assisted changes — a pattern showing up since Q3 2025. One incident involved an AI tool deleting and recreating an entire environment instead of making targeted changes. Another caused a nearly six-hour checkout outage. Junior and mid-level engineers can no longer push AI-assisted code without a senior signing off. Amazon pushed back fast when the Financial Times reported on the internal briefing, insisting AI tools aren't to blame — but offered no independent data. The more credible read, from James Gosling and others, is that the real problem is structural: teams were gutted during the AI hype cycle, and the human expertise needed to catch bad changes before they hit production is gone. The senior sign-off requirement is a reasonable short-term patch, but it probably just negates the efficiency gains Amazon was chasing in the first place. ## Claude Code ships /btw for mid-task side conversations Small feature, worth knowing about. `/btw` opens a parallel overlay where you can ask Claude a question — plan ahead, check something, think through the next step — while the main task keeps running. The answer shows up in an overlay and doesn't pollute the conversation history. If you've ever interrupted a long-running agent just to ask a quick question and watched it lose its place, this is the fix. The UX instinct here is right: long-running agentic work needs a way to communicate without breaking flow. ## GPT-5.4 gaining ground on complex codebases Multiple teams are actively switching coding workloads to GPT-5.4, citing noticeably better performance on hard problems in large codebases. One benchmark has Claude Sonnet 4.5 still leading in VS Code vibe coding, but GPT-5.4 ahead of GPT-5.3-Codex. The ethskills eval is a useful data point: both Claude Opus 4.6 and GPT-5.4 jumped from sub-20% to 95% accuracy on Ethereum-specific tasks when given proper skill docs loaded as context — which says more about context engineering than raw model capability. The multi-model workflow is becoming the norm; picking one model and sticking with it is increasingly the wrong call. ## JetBrains launches Air and Junie CLI JetBrains shipped two new tools in public preview and beta. Air is an agentic development environment for macOS that lets you run multiple coding agents — Claude, Gemini CLI, Codex, Junie CLI — side by side in a single workspace with Git integration and a built-in terminal. Junie CLI is a standalone LLM-agnostic coding agent that works from the terminal, any IDE, or CI/CD pipelines, supporting models from OpenAI, Anthropic, Google, and Grok. Both are designed to keep agent output grounded in the actual codebase rather than hallucinating context. --- ## Also notable - **Codex capacity issues:** OpenAI acknowledged Codex service is "a little choppy" as demand is surging faster than compute can be added. Some developers are switching back to Claude Code as a result. - **State Department drops Claude:** Internal doc shows the State Department moved its chatbot from Claude Sonnet 4.5 to GPT-4.1 following Trump's directive to cancel Anthropic contracts. - **Anthropic Code Review in research preview:** Claude Code's new PR review feature uses multiple specialized agents to check for logic errors, security issues, and regressions, then filters false positives before posting inline comments. Costs $15-25 per PR, takes about 20 minutes, and doesn't block merges. - **Uber AI metrics:** Fresh data from February 2026 shows 31% of Uber's code is AI-authored and 11% of PRs are opened by agents — significant numbers, but a reminder that even heavy AI investors are far from fully autonomous development. - **OpenCode anti-slop principles:** OpenCode published explicit rules: don't ship features just because you can, leave code better than you found it, fix process before adding features. Worth bookmarking. - **DeepSeek US usage collapsed:** a16z reports DeepSeek usage in the US has "completely fallen off" after the initial hype cycle — couldn't retain American users. - **TanStack Intent ships Agent Skills:** A new CLI lets library maintainers generate and ship versioned "Agent Skills" alongside npm packages, teaching AI agents correct usage patterns. Skills travel with the package via npm update, solving stale rules files. - **Cloudflare's vinext experiment:** One engineer built an experimental Next.js alternative on Vite in about a week using Claude via OpenCode, spending $1,100 in API tokens. Early benchmarks show 4.4x faster builds and 57% smaller bundles, though it's explicitly experimental and lacks static pre-rendering. - **AI coding tool experience in interviews:** A poll of 539 developers found 43% were asked about AI coding tool experience as a job requirement, 25% as optional, and 32% said it didn't come up at all. - **Gemini Code Assist updates:** New "Finish Changes" feature observes in-progress edits and completes tasks without explicit prompts. Also added Outlines, which generates English summaries interleaved in source code. Both powered by Gemini 3.0. - **Agentic coding eval noise:** A post worth reading on why agentic coding benchmarks can vary by several percentage points between runs — relevant context for interpreting any leaderboard numbers. ## Similar posts on daily.dev - [January 2026: AI Agents Take Over, Claude Code Workflows, Multi-Agent Orchestration, and OpenCode](https://daily.dev/posts/january-2026-ai-agents-take-over-claude-code-workflows-multi-agent-orchestration-and-opencode-g2tclquzi) · Code with Andrea · 1 upvotes · 0 comments - [The Trends \#10: Amazon now requires senior approval for AI-assisted code from junior and mid-level engineers](https://daily.dev/posts/the-trends-10-amazon-now-requires-senior-approval-for-ai-assisted-code-from-junior-and-mid-level-e-odyocrxqm) · Tech World With Milan · 1 upvotes · 0 comments --- Tags: [#ai-coding](https://daily.dev/tags/ai-coding), [#claude-code](https://daily.dev/tags/claude-code), [#codex](https://daily.dev/tags/codex), [#jetbrains](https://daily.dev/tags/jetbrains) [View this post on daily.dev](https://daily.dev/posts/amazon-links-ai-tools-to-production-outages-claude-code-ships-btw-command-btxbjnzl5) ```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":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/amazon-links-ai-tools-to-production-outages-claude-code-ships-btw-command-btxbjnzl5","headline":"Amazon links AI tools to production outages, Claude Code ships /btw command","text":"Amazon mandated senior engineer sign-off on AI-assisted code after a wave of production incidents tied to agentic tooling, including one where an AI deleted and recreated an entire environment causing a six-hour checkout outage. Analysts suggest the real issue is structural: teams were downsized during the AI hype cycle, leaving insufficient human expertise to catch bad changes. Claude Code shipped a /btw command enabling side-channel questions during long-running agent tasks without disrupting the main thread. GPT-5.4 is gaining adoption for complex codebases, with multi-model workflows becoming the norm. JetBrains launched Air (a multi-agent macOS dev environment) and Junie CLI (an LLM-agnostic terminal coding agent). Other notable items: Codex capacity issues driving users back to Claude Code, the State Department switching from Claude to GPT-4.1, Anthropic's PR review feature at $15-25 per PR, Uber reporting 31% AI-authored code, and TanStack Intent shipping versioned Agent Skills alongside npm packages.","url":"https://daily.dev/posts/amazon-links-ai-tools-to-production-outages-claude-code-ships-btw-command-btxbjnzl5","datePublished":"2026-03-11T04:18:01.754Z","dateModified":"2026-03-15T06:08:21.152Z","author":{"@type":"Organization","name":"Agentic Digest","logo":"https://media.daily.dev/image/upload/s--V91DY4ls--/f_auto,q_auto/v1772617267/logos/agents_digest","url":"https://daily.dev/sources/agents_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/agents_digest","name":"Agentic Digest"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Agentic Digest","item":"https://daily.dev/sources/agents_digest"},{"@type":"ListItem","position":3,"name":"Amazon links AI tools to production outages, Claude Code ships /btw command"}]} ```