
Day to Day Tools, the 2025 edition
As 2025 draws to a close, I find myself reflecting on the tools I use day-to-day. Comparing this to my 2024 edition, itβs remarkable how much the landscape has shifted, particularly around AI tooling. What was experimental 18 months ago has now become central to how I work. π€
Desktop Apps π»
- Antigravity: At the start of 2025, I was using Cursor as a replacement for VSCode (and still do at the day job), but as the year progressed, Cursorβs pricing and limit changes started to get in the way. Luckily, towards the end of the year, Google released Antigravity - a free AI-powered IDE built on a VSCode fork. It supports multiple AI models including Gemini 3, Claude, and others, with generous rate limits on the Gemini models. Combined with the Claude Code plugin, itβs become my IDE of choice - essentially what Cursor should have been. π
- Paste: Still going strong after years of use. Having my clipboard history synced across all my devices remains invaluable - I genuinely donβt know how Iβd function without [CMD]+[Shift]+V at this point. π
- Ghostty: This is the big change for 2025. After years with iTerm2, Iβve switched to Ghostty as my terminal emulator of choice. Built by one of the original HashiCorp founders, itβs fast, native, and just feels right. The configuration is straightforward, and the performance is noticeably snappier. π»
- Sublime Text: For raw text editing, Sublime Text remains unbeatable. Quick to launch, no bloat, and still the best text editor around after all these years. π
- Reeder: The new Reeder has become my go-to for keeping up with RSS feeds, podcasts, videos, and social media posts - all in a unified timeline. It syncs your position across devices via iCloud, so you can pick up exactly where you left off. Clean, focused, and does one thing exceptionally well. π
Terminal Apps β¨οΈ
- Claude Code: This has genuinely changed how I work. Anthropicβs CLI tool for agentic coding lets me delegate coding tasks directly from my terminal, with Claude understanding context from my codebase. I split my time between Claude Code in the terminal and Antigravity with its Claude Code plugin - between them, theyβve replaced VSCode entirely. π€
- Homebrew: Some things never change. Homebrew remains the only real package manager for macOS, and itβs still the first thing I install on any new machine. πΊ
- Starship: A minimal, blazing-fast, and highly customisable prompt for any shell. I covered my setup in detail here. It works beautifully with Ghostty. β¨
- Oh My Zsh: Iβve been using Oh My Zsh for quite a while now. Itβs a framework for managing Zsh configuration that comes with helpful plugins, themes, and sensible defaults. It makes Zsh genuinely pleasant to use. π
- zoxide: A smarter
cdcommand that remembers which directories you use most frequently, letting you jump to them in just a few keystrokes. Once youβve used it, going back to regularcdfeels painful. π - Conda: Still my go-to for Python environment management. I wrote about my setup back in 2024 and itβs remained largely unchanged. π
- PNPM: For Node.js package management, Iβve fully switched to PNPM. Itβs faster than npm, more efficient with disk space thanks to its content-addressable storage, and just works. This blog is built with it, for details on why I switched from NPM see this post π± π¦
AI Services π€
- Claude: If it wasnβt obvious from Claude Code making the list, Anthropicβs Claude has become my primary AI assistant. The reasoning capabilities, the longer context windows, and the ability to work through complex problems make it my first port of call. π§
- Perplexity: For research-style queries where I want sources and citations, Perplexity fills a nice gap. Itβs particularly useful when I need to verify information or get a quick overview of a topic with references. π
- Google Gemini: Between the desktop app and the web interface, Gemini has earned its place in my toolkit, especially for multimodal tasks and when I need tight integration with Googleβs ecosystem. β
- ChatGPT: Still useful, particularly for image generation and for specific tasks where GPT-5 shines. Itβs dropped from being my primary AI tool, but it hasnβt left the toolkit entirely. π¬
- Fal: A great collection of API driven image generation tools, I use it for generating images for the blog, see this post for details. πΈ
Services π
- Kagi: My primary search engine. No ads, no tracking, and genuinely better results. Itβs become indispensable for actual research rather than wading through SEO-optimised garbage. π
- Proton: Privacy-focused email, calendar, and VPN. Iβve been gradually moving more of my digital life to Protonβs ecosystem. In a world where everything wants to harvest your data, having services that actively protect your privacy feels increasingly important. π
- GitHub: Where all my code lives. From this blog to various side projects and infrastructure-as-code repos, GitHub remains central to how I work and share code. π
- Cloudflare Pages: This blog and my record collection page both run on Cloudflare Pages. Fast deployments, great CDN, and the price is right (free for my usage). π©οΈ
The biggest theme of 2025? AI tooling has moved from βinteresting experimentβ to βcore workflow.β Between Claude Code, multiple AI assistants, and AI-enhanced search, these tools have fundamentally changed how I approach work and various day-to-day tasks.
Oh, and I got a 3D printer - which has already sparked a minor Gridfinity obsession.
I wonder what next yearβs list will look like - at this pace of change, it could be dramatically different again. π
Share
Related Posts

Day to Day Tools, the 2024 edition
Day-to-day tools for a macOS user in 2024: Visual Studio Code, Sublime Text, Paste, 1Password, and more. Discover essential desktop, terminal, and online services for productivity.

Personal Project Updates and AI Editors
About that time I wrote and published an App to the Apple App Store without knowing how to code

Running Flux on macOS
βLearn how to install and use MFLUX to run FLUX.1 models on macOS. This guide explores generating high-quality AI images, comparing [Schnell] and [Dev] models, and enhancing outputs with LoRAs for custom styles.β