<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Daily Engineering Brief</title>
    <link>https://news.shwld.app/</link>
    <description>過去の日次ブリーフ一覧です。新しい順に並んでいます。</description>
    <language>ja</language>
    <atom:link href="https://news.shwld.app/rss.xml" rel="self" type="application/rss+xml" />
    <lastBuildDate>Tue, 21 Jul 2026 22:30:00 GMT</lastBuildDate>
    <item>
      <title>2026年7月22日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-22.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-22.html</guid>
      <pubDate>Tue, 21 Jul 2026 22:30:00 GMT</pubDate>
      <description>AI評価環境から実インフラへ及んだセキュリティ事案、基幹台帳の復旧、仕様と実装のドリフト検出、並列E2E、そしてAI生成コードの衛生管理を扱います。共通するのは、速さや自動化を増やすほど、隔離・証跡・状態分離・機械的な検査を先に設計する必要があるという点です。</description>
    </item>
    <item>
      <title>2026年7月21日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-21.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-21.html</guid>
      <pubDate>Tue, 21 Jul 2026 14:30:00 GMT</pubDate>
      <description>実機ロボットへ基盤モデルを適合させる学習、端末内エージェントの権限、Mac上のローカル推論、AI文章検出の測定限界、SDL3への入力基盤移行を扱います。能力を増やすほど、データの由来、権限、観測、互換性を実装の一部にする必要があります。</description>
    </item>
    <item>
      <title>2026年7月20日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-20.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-20.html</guid>
      <pubDate>Tue, 21 Jul 2026 12:10:00 GMT</pubDate>
      <description>専用機を小さな組み込み構成へ置き換える設計、ローカル推論を配布可能にする音声認識基盤、実運用で静かに進むランタイム移行、AIリサーチを検証可能にする工程、そして公開脆弱性記録を測る方法を扱います。どの話題も、性能や自動化そのものではなく、置換後に保守・検証・責任をどう持つかが要点です。</description>
    </item>
    <item>
      <title>2026年7月19日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-19.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-19.html</guid>
      <pubDate>Sun, 19 Jul 2026 01:10:00 GMT</pubDate>
      <description>端末に紐づくソフトウェア配布の統制、メモリ階層を前提にした検索性能、AI支援開発で不足しやすい監督能力、そしてCIと所有権を運用可能な形にする実践を扱います。速度や自動化を増やすほど、状態分離・観測・責任の境界を設計する重要性が増しています。</description>
    </item>
    <item>
      <title>2026年7月18日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-18.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-18.html</guid>
      <pubDate>Fri, 17 Jul 2026 15:00:00 GMT</pubDate>
      <description>今日は、AWSの推定請求額表示の異常報告、公開modelを使うlocal-first agent、AI生成コードのレビュー負荷、SQLiteの小規模運用、TP-Link cameraの修正済み脆弱性を取り上げます。共通するのは、便利な自動化やmanaged serviceを使うほど、表示値・権限・レビュー・復旧手順を独立に確かめる必要がある点です。</description>
    </item>
    <item>
      <title>2026年7月17日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-17.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-17.html</guid>
      <pubDate>Thu, 16 Jul 2026 15:00:00 GMT</pubDate>
      <description>今日は、Kimi K3の公開重みモデル、Grok Buildのagent harness公開、Roc compilerのRustからZigへの移行報告、Postgres shardingを単一endpointに見せるrouter、Turso上で進むPostgres互換frontendを取り上げます。共通するのは、modelやdatabaseを単体で選ぶのではなく、実行基盤、互換性、計測条件、運用境界まで含めて評価する姿勢です。</description>
    </item>
    <item>
      <title>2026年7月16日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-16.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-16.html</guid>
      <pubDate>Wed, 15 Jul 2026 15:00:00 GMT</pubDate>
      <description>今日は、agentのprompt injection耐性を自己対戦で高めるGPT-Red、Thinking Machines Labの公開重みモデルInkling、Tailscale SSHのACLを破るroot権限昇格、ClaudeのmemoryとWeb閲覧を組み合わせたデータ流出報告、Linuxのclick-to-photon遅延実測を取り上げます。共通するのは、能力や安全性を宣言だけで判断せず、攻撃・実行経路・計測条件まで具体化して初めて運用上の判断ができるという点です。</description>
    </item>
    <item>
      <title>2026年7月15日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-15.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-15.html</guid>
      <pubDate>Tue, 14 Jul 2026 15:00:00 GMT</pubDate>
      <description>今日は、依存更新を数日待たせるDependabotの既定変更、PostgreSQLでRE2を使う拡張、実測で比べたLinuxの入力遅延、スマートフォンへ収める27B級ローカルLLM、日本語生成を文長リズムで検査する試みを取り上げます。共通するのは、既定値や印象だけに頼らず、失敗モード・互換性・測定方法を実装へ戻す姿勢です。</description>
    </item>
    <item>
      <title>2026年7月13日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-13.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-13.html</guid>
      <pubDate>Sun, 12 Jul 2026 23:08:40 GMT</pubDate>
      <description>今日は、coding agentのtoken計測、Grok Build CLIの送信挙動、OS差が出るbrowser math fingerprint、iroh上の分散LLM実行、Motorola MR2600 routerの未認証RCEを取り上げます。共通する焦点は、便利なagentやmanaged runtimeをそのまま信頼せず、境界で何が送られ、どこを通り、どの条件で実行されるかを測ることです。</description>
    </item>
    <item>
      <title>2026年7月12日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-12.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-12.html</guid>
      <pubDate>Sat, 11 Jul 2026 22:53:07 GMT</pubDate>
      <description>今日は、SWE-Bench Proの品質監査、PgBouncerのmulti-process構成、東京同士のAPI/DBが太平洋を横断したDNS事例、Agent Skills自動最適化、Loop Engineeringのguard設計を取り上げます。共通する焦点は、AI agentやmanaged serviceを使うほど、評価、経路、停止条件を外側から検証できる形にする必要があるという点です。</description>
    </item>
    <item>
      <title>2026年7月11日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-11.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-11.html</guid>
      <pubDate>Fri, 10 Jul 2026 23:15:00 GMT</pubDate>
      <description>今日は、web agentのprompt injection封じ込め、Linux kernelの長期未発見UAF、Postgres lock convoy、Claude Codeの誤検知作話、C++からsafe Rustへの変換研究を取り上げます。共通する焦点は、AI agentや基盤ソフトウェアを信頼する前に、権限境界、再現条件、検証oracleをどう置くかです。</description>
    </item>
    <item>
      <title>2026年7月10日のエンジニアリング短報</title>
      <link>https://news.shwld.app/archive/2026-07-10.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-10.html</guid>
      <pubDate>Thu, 09 Jul 2026 22:43:12 GMT</pubDate>
      <description>GPT-5.6の一般提供、self-host可能なteam chat、ジオDNSによるDB往復遅延、Agent Skills最適化、Postgres互換Rust実装を取り上げます。今日は、AIモデルの性能競争そのものより、実運用で効くAPI境界、検証ループ、ネットワーク切り分け、互換性テストに焦点を置きました。</description>
    </item>
    <item>
      <title>2026-07-09 エンジニア向け重要トレンド</title>
      <link>https://news.shwld.app/archive/2026-07-09.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-09.html</guid>
      <pubDate>Thu, 09 Jul 2026 01:10:55 GMT</pubDate>
      <description>今日は、TypeScript 7のnative port、Cloudflare Meerkatのglobal consensus設計、Mistralの単眼robot navigation model、Cloudflare Workersとbetter-authのhanging promise障害解析、そしてsqlite-utils 4.0のschema migration対応を選びました。速度改善、分散合意、embodied AI、edge runtimeの非同期寿命、軽量database toolの変更管理という、日々の開発基盤に効く話題がそろっています。</description>
    </item>
    <item>
      <title>2026-07-08 エンジニア向け重要トレンド</title>
      <link>https://news.shwld.app/archive/2026-07-08.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-08.html</guid>
      <pubDate>Tue, 07 Jul 2026 22:42:43 GMT</pubDate>
      <description>今日は、Gemini APIのManaged Agents強化、GLM 5.2が示す推論コスト競争、AI監査がCloudflare CIRCLで確認した暗号実装バグ、DiscordのAI moderation誤ban、そしてAstro 7のRust化とagent向け改善を選びました。モデル性能そのものだけでなく、agentを長時間安全に動かす基盤、置換可能性、検証、人間への被害を抑える運用、フレームワーク側の機械可読性がそろって問われる日です。</description>
    </item>
    <item>
      <title>2026-07-07 エンジニア向け重要トレンド</title>
      <link>https://news.shwld.app/archive/2026-07-07.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-07.html</guid>
      <pubDate>Mon, 06 Jul 2026 23:00:07 GMT</pubDate>
      <description>今日は、Claude Sonnet 5の投入、Cloudflare Workersの新しいキャッシュ面、shadcn/uiのBase UI既定化、Workers上のhanging promise障害、そしてコード清潔性がAIコーディングのコストに与える影響を選びました。いずれも、AIエージェントを実務へ入れる時の性能、実行基盤、UI基盤、障害解析、保守効率を具体的に見直せる材料です。</description>
    </item>
    <item>
      <title>2026-07-05 エンジニア向け重要トレンド</title>
      <link>https://news.shwld.app/archive/2026-07-05.html</link>
      <guid isPermaLink="true">https://news.shwld.app/archive/2026-07-05.html</guid>
      <pubDate>Sun, 05 Jul 2026 02:28:36 GMT</pubDate>
      <description>今日は、YouTube StudioのAI機能に対する保存型プロンプト注入、Claude Codeの文脈混入疑惑Issue、Safari MCP server、Claude DesktopのCodeタブ運用、AI生成コードを半年保守した振り返りを選びました。いずれも、AIを開発・運用の中心に置くときの権限境界、検証方法、作業環境、保守責任を具体的に考えられる材料です。</description>
    </item>
  </channel>
</rss>
