The Day My Terminal Became Mission Control
Joona Nuutinen I’m a freelance web consultant and musician. I’ve been an entrepreneur for 23 years. Today felt like science fiction.
It started with a simple frustration: I needed to email a project brief to a subcontractor. The brief was ready, the files were organized, but switching to a browser, logging into email, composing the message — that friction was enough to make me think, “what if I just built a Gmail integration instead?”
So I did. OAuth2 authentication, full CLI tool — read, search, send, reply, triage — all from the terminal. Twenty minutes later I sent that brief without ever leaving my workflow.
That alone would have been a productive day. But then the momentum kicked in.
Five APIs in one session
By the end of the day, I had wired up five Google API integrations into my personal “Life Management” system — a terminal-based command center I’ve been building with Claude Code:
- Google Calendar — find free slots, create events, check what’s coming
- Gmail — read, search, send, triage email without opening the app or browser
- YouTube Analytics — pull view counts, audience demographics, traffic sources, growth trends
- YouTube Data — video metadata, comment counts, channel stats
- Google Ads — campaign performance, budget management, geographic breakdowns
Each one follows the same pattern: OAuth2 auth module + CLI tool with subcommands. Once you’ve built one, the next takes half the time.
Cross-platform advertising analysis — from the terminal
Here’s where it got interesting. I’m an electronic pop artist with a YouTube channel (60K+ subscribers). I’ve been running video ad campaigns for a couple of years, evolving my strategy through trial and error. But I’d never had all the data in one place.
With YouTube Analytics and Google Ads both accessible from the terminal, I could finally cross-reference the two. I pulled monthly view duration data from YouTube and matched it against campaign spend and geographic targeting from Google Ads.
Three distinct phases emerged:
| Phase | Strategy | Avg View Duration | Cost per Watch Hour |
|---|---|---|---|
| Volume-first | Broad targeting, maximize views | 22 seconds | €1.97 |
| Geographic | Target specific countries | 18 seconds | €1.24 |
| Engagement | European audiences, quality signals | 55 seconds | €0.36 |
The numbers told a clear story: chasing view counts was expensive and shallow. Targeting for engagement — longer watch times in markets that matter — costs a fifth as much per watch hour and produces 2.5x longer viewing sessions.
But the real discovery was in the organic data.
The algorithm noticed
When I queried YouTube’s traffic source breakdown month by month, something jumped out. Starting around the same time I shifted to the engagement-focused strategy (which was based on an interview with Gemini, and rooted in specific music ad professional videos, sharing the recipes):
- Algorithmic recommendations grew from ~50/month to 2,000+
- Subscriber-driven views exploded from ~100/month to nearly 9,000
- Organic share of total views went from 2-4% to over 10%
Higher engagement signals from paid campaigns appear to be teaching YouTube’s algorithm that the content is worth recommending. The paid strategy is generating organic compound interest.
Live budget optimization via API
With the analysis complete, I executed campaign budget changes directly through the Google Ads API. Increased budgets on the highest-engagement campaigns, decreased spend on underperformers, and effectively paused campaigns tied to the old volume strategy.
Total daily budget went down. Expected impact went up. All executed from a Python script in the terminal.
Why this matters beyond the novelty
I’m not sharing this to show off the tooling. The real insight is about friction and avoidance.
As a freelancer whose values prioritize creativity and freedom over financial admin, I’ve historically avoided the tedious stuff — billing, email, analytics, campaign management. Not because I can’t do it, but because the context-switching cost is too high. Every browser tab is a trap. Every login screen is a reason to do something else instead.
Building these integrations into a unified terminal workflow removes the friction. When checking ad performance is as easy as typing python ads.py campaigns, it stops being a thing I avoid and becomes a thing I just do. When sending an email is a one-liner, I actually send it.
The compound effect of removing friction from a dozen small tasks is enormous. Today I:
- Organized and sent a complete outsourcing package for a client project
- Built a Gmail integration and immediately used it for real work
- Authenticated two separate Google Cloud projects for different accounts
- Performed a data-driven advertising analysis combining two platforms
- Executed live campaign optimizations based on that analysis
- Set up a personal finances review session
All without opening a browser. All from one terminal window. All with Claude Code as my copilot.
The tools
Everything is built on a simple pattern:
- Auth module: OAuth2 flow with local callback server, token persistence
- CLI tool: Subcommands for each operation, JSON output option for piping
- Claude Code: The reasoning layer that ties it all together
The individual tools are straightforward Node.js and Python scripts. The magic isn’t in any single integration — it’s in having them all accessible from the same context, with an AI that can reason across all of them simultaneously.
What’s next
The system keeps growing. Email triage with AI-powered filtering. Automated billing from time tracking data. A LinkedIn posting API for sharing work like this without the context switch. A financial dashboard pulling from all these data sources.
Each new integration makes the next one easier and the whole system more valuable. It’s compound returns on automation.
Today felt like the future. Not because the technology is new — APIs and CLI tools are ancient. But because the combination of unified terminal access + AI reasoning turned a day of “admin tasks I’d normally avoid” into the most productive workday I’ve had in months.
I build things at the intersection of web technology, music, and AI. Currently working on Netura Campaign Manager, an AI-powered advertising tool, and making electronic rock as Unzyme.