{ "version": "https:\/\/jsonfeed.org\/version\/1.1", "title": "IT guy's blog: posts tagged agents", "_rss_description": "AI, IT, tech, software development, AI agents", "_rss_language": "en", "_itunes_email": "", "_itunes_categories_xml": "", "_itunes_image": "", "_itunes_explicit": "", "home_page_url": "https:\/\/egor.work\/tags\/agents\/", "feed_url": "https:\/\/egor.work\/tags\/agents\/json\/", "icon": "https:\/\/egor.work\/pictures\/userpic\/userpic@2x.jpg?1673209172", "authors": [ { "name": "Egor Zotov", "url": "https:\/\/egor.work\/", "avatar": "https:\/\/egor.work\/pictures\/userpic\/userpic@2x.jpg?1673209172" } ], "items": [ { "id": "18", "url": "https:\/\/egor.work\/all\/my-thoughts-about-the-junie-ai-agent\/", "title": "My thoughts about the Junie AI agent", "content_html": "

Supercharging My Development Workflow with AI Agent Junie<\/h2>\n

Over the past few weeks, I’ve been using AI agent Junie<\/a>, and I can say — I love it. It’s made my daily development smoother, faster, and far more efficient. Here’s why.<\/p>\n

\n\"Junie's\n<\/div>\n

Why I Like It<\/h2>\n
    \n
  1. It saves me time.
    \nJunie allows me to write more code in less time. It helps with code generation and speeds up repetitive tasks — I’m now much more productive.<\/li>\n<\/ol>\n
      \n
    1. It can execute commands and automatically fix issues.
      \nJunie comes with a built-in command shell. It can run commands, detect errors, and even fix them automatically — no need to leave the terminal or do everything manually.<\/li>\n<\/ol>\n
        \n
      1. It understands my code style.
        \nOne of the best parts is how Junie matches the style of my codebase. It doesn’t just dump generic code — it adapts to how things are already written.<\/li>\n<\/ol>\n
          \n
        1. It reuses tools I already use.
          \nJunie recognizes the existing patterns, libraries, and factories in my project. For example, when generating integration tests, it reuses the same tools and setups from other tests. It feels like it knows my project.<\/li>\n<\/ol>\n
            \n
          1. It writes great tests.
            \nI now write almost all of my tests using Junie. It’s especially good at generating realistic and complete test cases that fit seamlessly into the existing test suite.<\/li>\n<\/ol>\n

            The Downside<\/h2>\n

            The only real drawback is the lack of transparency around credit usage. You don’t get a clear breakdown of how many credits each request consumes or how many you have left. You only get a warning when you’re running low, which can be annoying.<\/p>\n

            Tips & Tricks for Using Junie Effectively<\/h2>\n

            Here are a few tricks I’ve discovered that make working with Junie even more effective:<\/p>\n

              \n
            1. Ask Junie to write code without comments.
              \nIf you prefer clean, uncommented code (especially for production use), just ask Junie explicitly to omit comments. It will follow your instructions precisely.<\/li>\n<\/ol>\n
                \n
              1. Provide the full Docker command path when running tests inside the Docker container.
                \nIf your application runs inside Docker, include the full path to your test command (e. g., docker exec app php bin\/phpunit). Junie will use it to run tests inside the correct container, avoiding issues with Dockerized services like the database.<\/li>\n<\/ol>\n
                  \n
                1. Tell Junie not to mock specific classes.
                  \nIf you prefer using real objects instead of mocks for certain parts of your code, simply say so. Junie respects that and will write tests using the actual implementations instead of generating mocks.<\/li>\n<\/ol>\n

                  Final Thoughts<\/h2>\n

                  Despite that, I highly recommend trying AI agent Junie, especially for writing tests. It’s become an essential part of my workflow, and I can’t imagine working without it now.<\/p>\n", "date_published": "2025-07-23T14:22:21+03:00", "date_modified": "2025-07-23T21:43:44+03:00", "tags": [ "agents", "AI" ], "image": "https:\/\/egor.work\/pictures\/Screenshot-2025-07-23-at-22.41.26.png", "_date_published_rfc2822": "Wed, 23 Jul 2025 14:22:21 +0300", "_rss_guid_is_permalink": "false", "_rss_guid": "18", "_e2_data": { "is_favourite": false, "links_required": [], "og_images": [ "https:\/\/egor.work\/pictures\/Screenshot-2025-07-23-at-22.41.26.png" ] } } ], "_e2_version": 4171, "_e2_ua_string": "Aegea 11.4 (v4171)" }