My AI Stack (June 2025)

16 Jun, 2025

This is my regular monthly post to track my AI tool ecosystem.
Here is the May post.

No changes to last month's stack.

Current Stack Overview

• Core LLM Subscriptions ($240/month total)

  • Anthropic Claude Max ($200/month)
    • I use Claude Code for all my vibe coding and managing my Obsidian knowledge base. Still extremely happy with this; mostly thanks to Opus 4.
    • If you are not yet a hardcore Claude Code user, you should definitely check out the testimonials at r/ClaudeAI. Honestly, I'm a bit worried that Anthropic is probably losing money on this and that they might jack up the prices or reduce the limits.
  • OpenAI ChatGPT Plus ($20/month)
    • At this point, I'm only using o3 for image trivia stuff (where is this picture taken, how much is this boat), and 4o for random image creation here and there. I really don't see myself canceling my OpenAI subscription anytime soon (especially since we are supposed to be weeks/months away from GPT 5), but I'm definitely not a heavy user, and I'm pretty sure that it would be cheaper for me to use their API at this point. They are also supposed to enable o3-pro for Plus users, so looking forward to that.
  • Google Gemini Advanced ($20/month)
    • Google still has the best deep research. Mostly sticking to this subscription for anything that involves online research. Otherwise, my Gemini usage has gone down significantly.

• AI-Powered Products ($10/month active)

  • GitHub Copilot ($10/month) - Still good for the occasional Cline/Roo calls with Gemini pro.
  • Warp Terminal (Free plan) - I've been using less Warp since Claude Code is excellent at handling all my terminal commands. I still like the visuals but I'm much less likely to get their subscription at the moment.
  • Perplexity Pro ($20/month, yearly) - I think I used Perplexity 3 or 4 times in the last month. Their new "create with deep research" option is cool but Gemini is still better at that. Keeping this around since I got the annual subscription.

Things I'm Looking Forward to Testing/Playing With

I'm still not a heavy MCP user. Here is a list of things I want to play around with this month:

• Zen MCP for calling other LLMs
• Context7 and DeepWiki for getting up-to-date documentation on packages
• Puppeteer and Playwright for testing browser apps
• Workflows-MCP for building deterministic workflows

Also, these Claude Code tools for scaling agents seem really cool but I didn't have time to test them:

• Claude Composer for adjusting agent configs
• Claude Swarm for spinning up multiple agents at once
• Uzi for getting better results with agents (using @k)

I still haven't played around with the GitHub integrations of AI agents (i.e., Codex, Jules and Claude GitHub Actions). I first need to figure out (and get comfortable with) how to divide my tasks across multiple agents properly.

This Month's Highlights

• Claude 4 Opus: Enough has been said. Really happy with the model (though I still miss Sonnet 3.7's dedication to please)
• Willow Voice: Not paying for it yet, but the early experience shows promise for replacing my previous Superwhisper setup.