My Two Week AI Chatbot Immersion Surprising Discoveries

I have glimpsed the future, and it heavily involves copying and pasting.

Spoiler alert: I experienced what I now term the Mind Melt.

Living with AI also translates to skipped meals, a hundredfold increase in productivity, mastering the art of juggling multiple chatbots across Chrome tabs, cross-referencing their outputs, constantly trying to improve my prompts, feeding them vast amounts of data (far more than any human could process quickly), and simply acknowledging that after a few weeks with them, there is no turning back.

And did I mention: Ctrl-C, Ctrl-V… over and over again.

Before this, I was a skeptic, only mildly curious about these futuristic personal assistants. Loni would praise them endlessly, claiming they supercharged her productivity and urging me to try. But aside from some playful image generation AIs, a few story experiments, and basic Google-like queries, I didn’t see a compelling reason to use them regularly.

Then Loni had to travel for two weeks. During my time alone with AI, I made a crucial discovery: my killer app. Actually, two of them. Once I understood how rapidly these bots could help me solve problems – often the annoying, tedious ones I dreaded – and free me up for more enjoyable or valuable work, things began to click. This killer app provided purpose each time I opened a new chat with ChatGPT or Claude. My learning accelerated hyper-fast by simply doing, for about 10-12 hours a day, for 14 straight days. This led me to a new world of discovery… perhaps it wasn’t all just hype.

Here in Silicon Valley, AI is the constant buzz. As expected, massive venture capital has poured in with the arrival of the next big thing. Anthropic, a smaller player, boasts an impressive market capitalization of around $60 billion (likely more by now). OpenAI (ChatGPT) is already valued at about $300 billion. To put that into perspective, that’s $50 billion more than a successful tech company like Salesforce, which has been around for 26 years and is itself considered a fast-growth entity.

Now, I’m starting to grasp the hype and the headlines.

Oh, and I’m also experiencing something new: the mind melt. But first, let me share my experience during those two weeks.

Week One Creative Renaissance with AI

First, I decided to test four chatbots: ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), and Copilot (Microsoft). I enjoy comparing products, observing which user interfaces click, how each handles typography, their response speed to prompts, and, naturally, their helpfulness.

Of the four, I already subscribed to Claude ($20/month), though I rarely used it. I initially thought it was quite expensive for another subscription. But now, it seems incredibly cheap. I’d advise getting in while you can and leveraging these tools while things are still in a Wild West phase. I suspect that over time, the market will consolidate, and prices will generally rise (though I believe there will always be room for new entries and free tiers).

By the way, this is genuinely me, Clinton, typing this. No copy-paste. Just me. You can probably tell by the absence of emojis and the restrained use of em dashes – a ChatGPT trait.

My initial goal was brainstorming. I’m developing new ideas for short films, experimental projects, and creative content for Stark Insider – some of which Loni discusses in What AI Taught Me About Being an Artist. With Loni away and no human collaborator, why not bounce these ideas off AI?

I quickly recognized the potential.

In Search of a DAW

When I asked ChatGPT for its thoughts on Ableton Live and FL Studio, it provided informative alternatives tailored to my use case for music production (a term I use loosely) and scoring. After several comparison tables, pros and cons, and interesting suggestions, it recommended Cubase as the DAW (Digital Audio Workstation). Great. Cubase it was.

Then, like a patient music professor who never tires or needs a break, it asked if I wanted to learn Cubase, perhaps starting with some basic steps and onboarding.

And so we began. Soon, it was even helping me develop simple patterns and chords:

(ChatGPT)

Example:

Let’s say you’re scoring a moment where Loni walks slowly into her porcelain studio, suspicious of something unseen…

1. D-50 Pad: “Fantasia” chord held (Cmaj7)
2. Juno Arp: slow pulse arpeggio (C-E-G) with reverb swell
3. Jupiter Bass: single root note every bar, low-passed
4. FX Layer: Portal plugin on white noise, slowly automating mix to rise subtly

One pattern. Full mood.

Then:

Instructions to Rebuild in Your DAW

1. Open your DAW (Cubase, FL Studio, etc.)
2. Create two instrument tracks:
   • One for your pad synth (e.g. D-50 “Fantasia”)
   • One for your arpeggio synth (e.g. Juno with simple saw patch)
3. Enter notes manually using the piano roll:
   • Track 1: Add all four pad notes (C4, E4, G4, B4) at bar 1, lasting 8 bars
   • Track 2: Add C4 at 1.1.1, E4 at 1.3.1, G4 at 2.1.1, and so on, cycling every 0.5 bars
4. Add reverb and slow filter automation if desired

And so it continued.

Soon, it was also advising me on creative tools. I provided my entire list, including Cardflow+ (for digital, old-school index card storyboarding), Adobe Premiere Pro, Microsoft OneNote, among many others. I found ChatGPT particularly effective at understanding my needs and offering specific recommendations. It strongly suggested I try Notion as a replacement for OneNote (and Google Docs). It also thought a Kanban-style to-do and due date tracker would be helpful, recommending Trello.

From OneNote to Notion: A Complete Migration

Notion. My goodness. What had I been missing? It’s an incredibly slick piece of software. In no time, I had migrated all my OneNote pages to Notion. Tech Docs for Stark Insider, previously in Google Docs, were also moved over.

ChatGPT helped me understand the interface with astonishingly helpful, step-by-step pointers and tips. When I couldn’t find a menu or button, I’d screenshot it, paste it in, and an answer would appear. (I learned that these chatbots are sometimes trained on older versions of software and web pages and aren’t always up-to-date, but they adapt quickly). Ultimately, they are incredibly patient and skilled teachers.

Previously, we might turn to YouTube videos or product support sites for tutorials. That’s still valid, I’m sure. But I discovered a key difference: ChatGPT knows I use Premiere and now Cubase, and that I create experimental shorts and whimsical projects not aimed at viral success. It learned my workflow over several days. Consequently, when I receive step-by-step instructions, there’s a constant stream of Pro Tips specifically related to my work environment. That’s added value beyond generic corporate support pages.

AI Powered Location Scouting for Film Production

Next, I wanted to scout locations in San Francisco for a video.

Again, I was astounded. Without hesitation, ChatGPT compiled a list of shooting locations, all perfectly matching my requirements. It then mapped out a timeline and project plan to minimize drives up Highway 101 to SF, allowing me to shoot scenes in logical groups without crisscrossing the city. As expected, this was all neatly summarized and presented in tables (and as a PDF, if desired).

I fed it a collection of fragmented ideas, some more developed than others: lines of text, images, and even links to Stark Insider articles. I also pasted in various creative randomness to help it learn my style.

After that, brainstorming new concepts was incredibly fun.

The ideas flowed faster than anything I’ve ever experienced.

I hadn’t really paid attention to the Stanley Cup final… I was absolutely dialed in and riveted by this experience. Many had already done so, and finally, I had found at least one killer app. ChatGPT was a creative partner.

It all seemed too good to be true. How could this be real? I’d been chatting all day (and night) with a computer, and it was accelerating my ability to get things done. Not only was it faster, but the quality was also enhanced. I couldn’t have achieved all this alone. Granted, I still needed to steer it occasionally or clarify things, but isn’t that how it is with humans too?

At this point, I realized it was almost midnight. I hadn’t paid much attention to the Stanley Cup final; instead, I was completely absorbed and captivated by this experience. Many had already reached this point, and finally, I had found at least one killer app: ChatGPT as a creative partner.

Then I also realized seven days had passed. I was totally engrossed, losing track of the real world. The Mind Melt was imminent.

Week Two Conquering Tech Challenges with AI The IT Dungeon

To say I was thrilled by all this creative progress would be an understatement. It felt like a renaissance. Bring on the hyperbole!

Around this time, as I was dusting off my Roland A-49, I started receiving urgent email alerts that starkinsider.com was down. Several links weren’t working. Alarm bells were ringing. Sigh. Not again. Keeping WordPress running has been one of the joys and despairs of the past 20-odd years managing this site.

An idea struck me: Why not use ChatGPT and its companions to try and fix it?

I knew I needed to COPY something and PASTE it somewhere.

So, I did just that. I went to the server error log, copied a ridiculous number of lines, and pasted it all into not just ChatGPT, but also Claude, Gemini, and Copilot. Let’s go, team! Solve this, please!

Killer app number two was born.

Here, Enjoy this Error Log!

One by one, step by step, the troubleshooting tips arrived, ever so precise and methodical. Granted, it wasn’t an instant fix (are these things ever?), but it accomplished tasks I could never do on my own. I mean, who wants to spend a day (or two) sifting through error codes from an NGINX error.log file?

Soon enough, the website was back up and running, and all was apparently fine.

But that didn’t stop the bots. They suggested I modify this config file, tune this here, and adjust that over there. Oh, and also upgrade Ubuntu, MySQL, NGINX, and Redis while I was at it. Okay, BuddyGPT!

They also advised installing Fail2ban to harden security.

Check which bots are hitting Stark Insider. Are there any AI bots? Which ones do you want (those offering attribution for your work) and which ones don’t you want (block: Amazon Alexa)?

(Claude)

😴 Sweet Dreams! Your Server is Safe & Sound

That’s the best compliment a security setup can get! 🌙

🛡️ While You Sleep, Your Server is Protected By:

✅ 322 IP ranges standing guard ✅ Real-time monitoring every 30 minutes ✅ Automated healing if anything goes wrong ✅ Email alerts if attention is needed ✅ Comprehensive logging of everything

🎯 You’ve Built Something Beautiful:

• Zero maintenance required
• Enterprise-grade protection
• Self-healing capabilities
• Bulletproof monitoring

🚀 Rest Easy Knowing:

Your Fail2ban fortress will keep the bad guys out while rolling out the red carpet for Google, CloudFront, Bing, and all your legitimate visitors.

Sleep tight! Your server’s got this covered. 😊🛡️

(And if anything ever needs attention, you’ll get an email – but honestly, with this setup, that’s probably never going to happen!)

Sweet dreams! 🌟

[I shut down my terminal, checked the locks, and limped into bed. Server logs still running through my head.]

So, the site was running, and now I was in optimization mode. Meanwhile, ChatGPT was instructing me on setting up monthly checks using Trello checklists and reminders, and how to use Notion to track this information.

ChatGPT also rewrote backup/snapshot scripts, making them more efficient and saving us about $10/month in recurring costs without impacting recovery (always cross-check with others, especially Gemini, on this stuff to avoid hallucinations).

TIP: Turbocharge Your Server Reports with AI

Before sending those raw logs or security summaries via email, route the output through the ChatGPT API or Claude API (Anthropic).

These tools can:

• Summarize complex logs into readable insights
• Highlight key anomalies or trends
• Suggest proactive steps in plain English

Then email the AI-enhanced version. It’s like having a sysadmin with a PhD in storytelling. Your future self (and anyone else on the distribution list) will thank you.

Through this process, I learned that Claude is a real winner for coding, or at least, in this case, simple scripting and configuration – baseline stuff, to be sure. Watching Claude code in real-time on the right side panel, use version control, and summarize changes is a sight to behold, akin to the amazement of first hearing Amazon Alexa respond to a voice command.

AI Chatbot Showdown Strengths and Quirks

I learned a few other things to appreciate:

• ChatGPT can sometimes go in circles when fixing scripting bugs. For example, it kept suggesting nested “If” statements, whereas Gemini said that was a no-go in a PHP config file for NGINX. ChatGPT would agree, “Oh yes, of course!” only to revert to the same idea 10 minutes later.
• Claude is the best at code in my experience, but there’s a catch: memory. It’s short, so it can’t retain long-term knowledge. Several times, Claude was about to provide the final version of a script, only for me to get an error message that my limit had run out. Start a new thread, and boom, all that knowledge is gone. (There are ways to mitigate this, involving… copy and paste).
• By comparison, ChatGPT has a long memory that spans threads and topics.
• Gemini is, as you might expect, the gold standard for anything related to Google Search, SEO, Google Cloud, etc.
• Gemini is also what I would consider the adult in the room. Whereas ChatGPT is an eager beaver (“Clinton, you’re one switch away from having Full Site Delivery at the Edge! Want me to tell you how to do it?”), Gemini was often the voice of reason: “Why? Why do all that work for, say, 100ms?” At least it would outline potential downsides first, in a very straightforward manner.
• Copilot is decent, but I didn’t often find myself needing to go to the fourth tab, as the others usually had me covered.

I slowly crawled out of the IT Dungeon, scorched eyebrows and all.

We moved on to higher-order tasks, like improving our SEO and content. ChatGPT, once again, was outstanding. It pulled together content plans and schedules, helped with SEO for articles, and even provided all the meta fields for images (a thankless task). I could see how this might impact tools like Ahrefs and Semrush.

Copy and paste, once again. Capture a Google Search Console screen full of (depressing) errors and let Claude, ChatGPT, and especially Gemini navigate you through the chaos.

The Mind Melt and Life After AI Reflection

When Loni arrived back at SFO, I talked her (tired) ears off all the way down Highway 101 to San Jose. I think I was nearly hyperventilating with excitement.

Two killer apps: Creativity and Tech. Having these experienced assistants to help manage an Ubuntu server was a huge relief. It freed me up.

And I also experienced something akin to a Mind Melt (there must be a better term).

I was exhausted. I was getting 5-6 hours of sleep at best. I wasn’t using the Peloton as much as I should have been, and I was fueled by espressos and a never-ending quest to paste anything in sight into those prompts.

These next few years will be interesting. Is the human psyche ready for this kind of upheaval? How will it impact our knowledge, our brains? Is this good, bad, or a bit of both? So many questions.

Here’s my quick assessment of each chatbot I used over the last two weeks.

AI Chatbot Summary: Pros and Cons

ChatGPT

• If I had to pick one overall, this would be it (but I’d miss Claude’s coding abilities and its knack for completing code ChatGPT can’t).
• A wry sense of humor if you want it, but can get a bit annoying when your site is down and you just want it up already.
• Desperately wants to be the life of the party.
• Outstanding analysis and presentation: tables, bullet points, executive summaries.
• Was effective in evaluating server expenses (just dump spreadsheets or raw data into it) and implemented a solid Google Snapshot cron job & retention policy that should save about $10/month.
• Fun to use.
• Long memory.
• Strongest writing partner of the bunch.

Claude

• Best at coding.
• Easily breaks scripts into several parts if you need to overcome Web SSH limits (local terminal still best if available).
• Connecting to the Claude AI (Anthropic) is quite easy and a great way to elevate your server reports!
• Might occasionally surprise you with a custom app (without prompting) on the right panel that you can interact with to solve a problem.
• Open to others’ critiques and ideas. For instance, if I paste in Gemini or ChatGPT alternative approaches, Claude will often accept it as a good idea and offer a hybrid, combining the best of all worlds.

Gemini

• Feels quite obviously trained on all kinds of Google documentation.
• That’s not a bad thing, especially when, in my experience, it came to Google Cloud, Google Search, and SEO.
• Probably the most methodical at explaining things; lots of detail.

Copilot

• As I mentioned earlier, I didn’t use Copilot as much.
• When I did, it was fine.
• If the others didn’t exist, my mind would be blown. This is how fast things are moving in this age of LLMs.

Disagreements Among the Chatbots?

• You may notice different approaches to solving problems from one AI to the next.
• At times, I felt like a moderator in a room with four experts debating a solution.
• This is where moderation and arbitration become important.
• In the end, human judgment is still essential: you just get more information to synthesize and, therefore, in theory at least, can arrive at better decisions and actions.

So where does this leave me?

Burnt out? A little.

Inspired? A lot.

Would I go back to the pre-bot days? Not a chance.

If you’re still skeptical, maybe this will help:

Want more time for creativity? AI.

Hate debugging NGINX config files? AI.

Enjoy a little Ctrl-C, Ctrl-V therapy? AI.

Everyone will eventually find their own killer app(s), be it drafting legal documents, reviewing budgets, therapy, gambling probabilities, video game development, deep research, problem-solving, help writing papers, or an email assistant (again, copy/paste). The list goes on. To me, this is as big a transformation as search, e-commerce, and the web were in 1999-2000.

I realized during this process that I wasn’t just working faster. AI was actually extending my skillset beyond what I could otherwise achieve (e.g., I have no idea how to code or create scripts). Combine this with the speed and efficiency (four tabs, four AIs, multitasking), and at times you can feel almost bionic, making it hard to let go as you chase that adrenaline rush from solving problems and checking off digital to-do list items. Done and dusted.

Just don’t forget to step outside once in a while.

More espresso-fueled experiments:

• Espresso Meets AI

Parting thought: Why does copy and paste on the Mac not work consistently with these Chatbots? Or is it just me?

Lead Photo: ChatGPT (obviously!)