Hello! 👋

Our original plan for this week was to feature a well-known indie developer and founder of SwiftLee, Antoine van der Lee.. But life happens, and we need to postpone it to issue #17.

So this week we have Anton Gubarenko. Senior iOS Engineer / Contractor with 10+ years of experience.

In this week’s issue, he shares his top tools to enhance AI development, along with deep dives into the tools, their features, and his thoughts.

Follow him on X/Twitter & LinkedIn and read his Newsletter.

Sponsor 🤝

Build Hub: your infrastructure upgrade for GitHub Actions

Keep GitHub Actions for CI. Run on mobile-optimized build machines from Bitrise. No migration, no maintenance: just faster builds.

• M4 Pro Macs and granular Linux sizing (4-192 vCPU)
• Pre-configured stacks for iOS, Android, React Native, Flutter
• 30-50% faster builds vs GitHub Hosted Runners

Try Build Hub free

5 Tools to Enhance AI Development

AI is becoming a more essential part of software development than ever, changing the way we work and study.

With the variety of AI interaction methods available, we each choose what fits us best. Some developers prefer the command-line interface (CLI) for its speed and control, while others are waiting for robust AI integration in their IDEs (salute to Xcode!).

Personally, I’ve found advantages in both the terminal and in GUI tools that display AI-generated code changes in a friendly way. Here are a few tools that have been helping me boost my productivity with AI, whether via the CLI or a sleek IDE-like interface.

Conductor – Orchestrate Multiple Claude Sessions

Conductor is a macOS application that lets you run multiple sessions in parallel on your codebase. If we can run parallel agents - why not?! Think of it as a “project manager” for AI coders – each AI agent works on its own isolated copy of your repository, and Conductor gives you a birds-eye view to supervise them. You can see at a glance what each agent is working on and then review and merge their changes, all within one interface.

Key features:

• Parallel Agents: Spin up several coding agents simultaneously to tackle different tasks. Conductor supports running multiple instances of Claude at once, effectively letting you work on multiple features in parallel. This can dramatically speed up development, as an AI “team” handles various parts of your project concurrently.

• Isolated Workspaces: Under the hood, Conductor uses Git worktrees to give each AI its own branch of your repo. In practice, that means each agent works in an isolated environment with a fresh copy of the code, so their edits never conflict. You don’t have to worry about one AI overwriting another’s changes until you deliberately merge them.

This is how I faced the Git Worktree feature, that were introduced long ago and now shines in new colors.

• Integrated Code Review: Conductor provides a built-in diff viewer and merge tool for reviewing AI-generated code. As the agents produce code, you can inspect the diffs in real time, comment or make adjustments, and merge the changes back into your main branch when you’re satisfied. It’s like a mini IDE focused on AI contributions, ensuring you stay in control of what goes into the codebase.

• Streamlined Workflow: To get started, you just add your repo to Conductor, deploy a few agents, and hit “Cmd+N”. The app will clone your project locally, launch the AI agents, and then show you who’s working on what, which parts need attention, and any pending reviews. In addition, Conductor comes with features like custom scripts, automated testing integration, and TODO tracking to help formalize how the AI agents work on your code (so they follow your plans and pass your tests). It’s a powerful way to coordinate multiple AI helpers in your development workflow.

Website: Conductor Docs – includes installation instructions and a feature tour.

Codex Monitor – AI Agent Command Center for Your Codebases

Codex Monitor is an open-source desktop app that acts as a command center for AI coding agents (specifically geared toward OpenAI’s Codex or GPT-based code assistants). It’s available on macOS, Windows, and Linux, and provides a unified UI to orchestrate any number of AI “agent” sessions across any number of projects. In simpler terms, if you’re using AI to help write or refactor code, Codex Monitor puts everything in one place: you can chat with the AI, run commands, view file changes, and manage multiple projects, all without juggling between your editor, terminal, and browser.

For me it’s “Conductor for Codex”. Missed and beloved.

Key features:

• Multi-Project Orchestration: Codex Monitor can connect to multiple code repositories at once and spin up an AI agent for each. It essentially runs a local AI server (or connects to OpenAI’s API) for each project and remembers the conversation threads for each. All your active projects and AI threads are listed in a sidebar, so you can switch context easily. If you’re working on a backend and frontend repo simultaneously, for example, Codex Monitor keeps both AI conversations and their code changes organized in one place.

• Conversation + Terminal in One: The interface is designed like a co-pilot IDE. You have a chat area where you converse with the AI model about your code, and right next to it is a live terminal/command interface. You can ask the AI to perform a task, it might propose some code or a shell command, and you can approve and run that command directly from the app. The context (current directory, recent command history, etc.) is shared, so the AI always knows the state of the project. It feels like pair-programming with an assistant that can also execute code when needed.

• Git Integration and Issue Tracking: Codex Monitor deeply integrates with Git and GitHub to streamline code changes. It has a panel that shows Git diffs and commit logs in real time – as the AI writes code or you execute commands, you can see the changes being made to your files and how they differ from the last commit. It even hooks into GitHub issues: if you link it to your repo’s issues, you can see them in the app, possibly even letting the AI reference them or mark them as it fixes things. This way, your AI assistant is aware of your tasks/bugs (from issue tracker) and you see all relevant info in one window.

• Live Diff & File Navigation: One standout feature is the live diff view. As the AI modifies files, Codex Monitor shows a live “beautiful” diff with syntax highlighting, so you can literally watch the code being changed and review it visually. You don’t have to manually open a diff tool or check git – it’s right there, updated in real time. Additionally, there’s a built-in file tree browser with search functionality, which makes it easy to open any file in your project to see context or inspect what the AI did. In essence, you get an IDE-like explorer and diff tool focusing on the AI’s changes.

• Skills and Prompt Library: For power users, Codex Monitor supports custom “skills” (plugins or predefined commands) and prompt templates. The app can autocomplete these skills or file path references as you type in the chat box. This means you can extend the AI’s functionality with your own scripts or prompt macros, and the interface will help you use them. For example, you might have a skill for running tests or for seeding a database – you can invoke it easily and the AI knows to use that skill. It’s a neat way to augment the AI with tools specific to your workflow.

In fact, the creators describe Codex Monitor as a “next-generation IDE” for AI-assisted development – without a text editor built in, because you might not even need one. The idea is that the AI and these integrated tools handle the heavy lifting, and you review/respond rather than writing code in a traditional editor.

Website: CodexMonitor.app – learn more and find the link to the GitHub repository for downloads.

iTerm2 – A Terminal that Supercharges Your CLI

If you try to use any CLI, a lot of things will start to bother you. cd to folder, runt multiple CLI to compare or simultaneous use. Terminal might need an upgrade.

iTerm2 is a popular open-source terminal emulator for macOS, often seen as a major upgrade over the default macOS Terminal app. It doesn’t have built-in AI itself (aside from an optional plugin, mentioned below), but it’s an indispensable tool if you like interacting with AI via command-line tools. iTerm2 offers tons of developer-friendly features that make working in the terminal more efficient and pleasant, which indirectly boosts your productivity when using CLI-based AI tools or scripts.

Key features:

• Split Panes & Tabs: You can split your iTerm2 window into multiple panes (both horizontally and vertically) to run several sessions side by side. This is great for multitasking – for example, you might run a local development server in one pane, an AI CLI assistant (like an OpenAI API script) in another, and some logs or tests in a third. Everything stays visible at once.

• Hotkey Window: iTerm2 has a configurable hotkey window that drops down like a console so you can access a terminal instantly from anywhere. With a single keystroke (which you set, e.g. ⌘+* or a custom combo), a terminal slides into view. This is perfect for quick AI queries or running a shell command without switching apps – a huge time-saver during development.

• Search & Navigation: The terminal output is searchable with Cmd+F, and iTerm2 highlights all matches like a web browser would. This makes it easy to find that error message or piece of AI response text in a long scrollback. There’s also an Instant Replay feature (essentially a time machine for your terminal) that lets you scroll back to see past output even if it’s scrolled off screen. Never miss an AI model’s earlier answer or a compilation error again.

• Clipboard History & Mouseless Copy: iTerm2 keeps a history of text you’ve copied or pasted in the terminal, accessible with a hotkey. If you copied an AI command or code snippet a few steps back, you can pull it from the paste history without flipping through your clipboard manager. Additionally, selecting text in iTerm2 can automatically copy it (so you don’t need to hit ⌘+C), which is a small but nice convenience.

• Notifications on Long Tasks: Ever kick off a lengthy job (like installing packages or running an AI training script) and then get distracted? iTerm2 can send a macOS notification when a terminal job is done or if it produces certain output . For example, you can set it to alert you when a process finishes or if a keyword (like “Error” or “Done”) appears. This way, if you run an AI code-generation script, you’ll know exactly when it’s completed its work.

• Optional AI Integration: Recently, iTerm2 introduced an optional AI Chat that brings generative AI assistance right into your terminal. It offers a chat window with a LLM that can optionally interact with iTerm2’s windows and share the session data.

Website: iTerm2 Features – full list of iTerm2’s capabilities and downloads.

ClaudeUsageBar – Track Your Claude AI Usage in Real-Time

Really don’t know why it’s not advertised a lot. Most popular command is obviously /usage and getting this info is critical. So if you’re using Claude, you might know it has usage limits – for example, a 5-hour message window and a weekly message limit on the Pro plan. ClaudeUsageBar is a free, open-source macOS menu bar app that helps you monitor those usage limits at a glance . It lives in your Mac’s menu bar (the top status bar) and displays how much of your Claude usage quota you’ve consumed, so you never hit the limit unexpectedly mid-conversation.

Maybe, setup is a little bit tricky (with Cookie copy-pasting), but it worth it.

Key features:

• Live Usage Meter: Once logged in, the app icon in the menu bar shows a percentage or progress bar of your current usage. Click it, and you’ll see detailed stats: how much of your 5-hour session limit is used and how much of your 7-day weekly limit is used. For example, you might see “Session: 67% used (resets at 3:45 PM)” and “Weekly: 82% used (resets on Jan 31)” in a dropdown . This real-time tracking means you always know how much Claude time you have left.

• Smart Notifications: ClaudeUsageBar will gently remind you as you approach your usage caps. It sends notifications at key thresholds – 25%, 50%, 75%, and 90% of your limit – so you get plenty of warning before you hit 100% . No more surprises where Claude suddenly refuses to respond because you unknowingly blew past the 5-hour limit; this tool makes sure you’re informed well in advance.

• Reset Countdown: The app also displays when your usage limits will reset. You’ll always know the exact time your 5-hour window renews, or how many days until your weekly quota refreshes. This is super handy for planning. For instance, if you’re at 90% usage and see there are 30 minutes until reset, you might wait for it to reset. Conversely, if you have a big coding task to do with Claude and only 1 hour left on the timer, you’ll know to maybe hold off or switch accounts. Planning your work around these limits becomes much easier.

• Privacy First & Open Source: Concerned about logging into a third-party app? ClaudeUsageBar is built with privacy in mind. It doesn’t collect or send any of your data to external servers – in fact, it simply uses your existing Claude session cookie to fetch your usage directly from Claude’s API, and all of that stays local on your machine. The app is fully open-source, so you (or any developer) can inspect the code on GitHub to verify it’s safe. It’s also lightweight (under 5 MB) and won’t hog resources, so you can let it run in the background continuously.

Website: ClaudeUsageBar – download and source code are available (macOS 12+).

MacWhisper - Voice Interface for AI (the underrated productivity multiplier)

MacWhisper is a macOS app that runs OpenAI’s Whisper speech-to-text models locally on your machine and converts your voice into text in real time. In practice, it becomes a universal dictation layer for every AI tool you use: ChatGPT, Claude, Codex, terminal prompts, commit messages, documentation, even Xcode comments.

And honestly, this is probably the most unexpected productivity boost in this entire list.

When I first heard that many developers like to speak with console instead of typing made me shocked, but after trying it - that is really helpful in some cases.

Because the biggest bottleneck in AI workflows is no longer thinking - it’s typing.

Key features:

• Local processing (very important): Everything runs locally on your Mac using Whisper models. No audio is uploaded anywhere. This makes it safe to dictate proprietary code or client context - a big deal compared to cloud dictation.

• Works everywhere. MacWhisper just outputs text into the active text field. That means it works with: ChatGPT / Claude web UI, Claude CLI / Codex CLI in iTerm2, Git commit messages, Pull request descriptions, Documentation, Xcode comments, Even Slack or Linear tickets. It becomes a system-wide developer dictation tool.

• Multiple model sizes: You can choose smaller models for speed or larger ones for accuracy. On Apple Silicon Macs the medium model is already extremely usable.

• Fast correction workflow: You speak → text appears → you lightly edit → send.

Website: MacWhisper – GumRoad website with Free and Paid versions.

Connect with Anton! 👇

Follow him on X/Twitter & LinkedIn and read his Newsletter.