How I Actually Use AI Daily: Astra, Claude 5.5, Blender & More

I used to think AI progress was overhyped until I tried building my daughter's imagination into reality six months ago. Back then, Claude Opus 4.6 couldn't handle the camera movements in Blender to create the animated scenes I needed. It failed over and over. Today, with ChatGPT Astra and Claude Opus 5.5, I'm watching Wild Rebellion—Lilly's character designs—come to life with full animation and voice sync. This isn't speculation. This is what AI actually does for me right now.

My Current AI Stack

Over the past year, I've assembled a practical AI toolkit that handles everything from creative projects to everyday annoyances. Here's where I actually spend my money and attention:

• ChatGPT Astra ($100/month plan) — Handles my most important work: the Wild Rebellion animated episodes, live stream setups, and anything I can't afford to fail
• Claude Code ($20/year) — My go-to for building actual projects, like the Roblox version of Wild Rebellion
• Claude Opus 5.5 (via Claude Code) — Unbelievably fast compared to Astra for code-heavy tasks
• Antigravity (free with Google Pro) — Quick generic questions I don't want to waste resources on
• MiniMax (discontinued starter plan) — Powers all audio and music generation for Wild Rebellion
• Grok (free tier) — On-phone generic questions when I'm out and about
• Muse Spark contributor tier (free) — Powers my Hermes agent setup

ChatGPT Astra vs Claude Opus 5.5: The Real Difference

The transcript compares these two frontier models directly, and the answer is clear: both are incredible, but they serve different purposes.

For creative, vision-driven work like animated episodes where I need lip-sync, scene composition, and artistic direction, Astra is my choice. It's still handling the Wild Rebellion episodes because that work is irreplaceable to me.

But when Opus 5.5 dropped, I was shocked. It's not just on par with Astra—it's significantly faster. I had zero confidence that Claude Code could build a functional Roblox game from Lilly's drawings. With Opus 5.5, it happened. The speed difference is real and matters for iteration-heavy work.

Quick Comparison Table

Real Projects AI Actually Built

Wild Rebellion: From Drawings to Animation

Lilly's been creating character designs for a project called Wild Rebellion. Six months ago, this wasn't possible with available AI. Opus 4.6 couldn't handle Blender's camera movements—it kept getting in its own way and failing to create the scenes I wanted.

Today with Astra, I'm rendering full episodes with animated characters. Episode two will have even more animation and voice work where you can actually see the lips move. The same tool that's been around nearly 20 years (Blender) now works because AI finally caught up to its complexity.

Roblox Game Built in One Night

Using Claude Code with Opus 5.5, I built a functional Roblox version of Wild Rebellion. I had zero confidence this would work. The $20 yearly plan became "super, super valuable" with Opus 5.5's performance. We're testing it tonight.

This wouldn't have been possible with older Claude models. The combination of Codex's agentic capabilities and Opus 5.5's speed made it happen.

Automated Amazon Returns

This one surprised me. My wife's Amazfit devices broke past warranty. Instead of spending hours on customer service, I asked ChatGPT to handle it. It found our Amazon purchase history, requested return authorizations, coordinated with the company, and got us new Helio and BIP-6 replacements shipped. Amazfit even honored the warranty slightly past the one-year window.

AI didn't just answer questions—it completed a real transaction that saved me hours.

The Subscription Reality

I want to be transparent about costs because the transcript covers this directly:

• The $100/month ChatGPT plan gives me three weekly usage resets. Switching from high to medium effort doesn't save tokens if you're in an individual thread—you lose cached reads and may actually spend more.
• Claude Code at $20/year is "the best bang for the buck" for building things. When Opus 5.5 dropped, it became dramatically more valuable.
• MiniMax's discontinued $100/year starter plan was "an amazing deal" for audio generation. If I can stay on it, I will.
• The free tiers (Grok, Antigravity, Muse Spark) handle generic questions and agentic experimentation without costing anything.

AI in Everyday Life

Beyond big projects, AI handles annoyances I used to just tolerate:

Mac Spotlight search is "garbage" and "doesn't work half the time." Antigravity searches my file system instantly. It found my MiniMax API key across multiple system locations when Spotlight would have failed.

When I burned through Gemini usage for the week, Claude Sonnet 4.6 handled generic questions without issue. These tools have become infrastructure.

What AI Can't Do (Yet)

Meta Spark has "no good use case" for me despite being powerful. Grok's free tier is useful for quick phone questions, but not worth paying for compared to what Astra offers. Muse Spark's contributor tier works for Hermes setup, but lacks computer use capability, which limits its appeal.

The frontier models aren't magic—they're tools with specific strengths. Matching the right model to the right task matters more than chasing the latest release.

Why I'm Bullish Despite the Doom

I've heard all the AI doomsday predictions from various CEOs. The "p-doom" percentages feel arbitrary. As Elon Musk said (and I agree): we have a 100% chance of death regardless. That hasn't stopped me from building things.

My portfolio reflects this conviction. Years ago, I bet against Intel and on NVIDIA and AMD. NVIDIA is now my largest holding. Tech companies account for roughly 90% of S&P 500 gains over the last two decades.

But more than money: my daughter is losing her imagination as she gets older, just like I lost mine. I can't bring Pokemon to life in my backyard anymore. But I can bring her characters to life using AI. In a world full of consumption, building something we're excited about together means more than any investment return.

The Verdict

Six months ago, I couldn't build what I'm building today. The progress isn't theoretical—it's in my finished episodes, my working Roblox game, and my completed Amazon returns. If you're asking whether AI is actually useful or just hype, the answer is in what you can now build that you couldn't build before.

With Astra and Opus 5.5, we have two amazing frontier models at incredible pricing. I've never been more excited about what's possible. The tools we have today would have seemed like science fiction a few years ago.

Source: What AI Actually Does for Me: Codex, Claude, Blender & Everyday Tasks