I’ve seen the future and it looks a lot like cut-and-paste.
Spoiler: I entered what I now call the Mind Melt.
Living with AI also means skipping meals, multiplying productivity 100-fold, learning to juggle multiple chatbots across Chrome tabs, cross-checking one against the other, striving to up my prompt game, feeding it copious amounts of data and content (far more than a human could digest in the blink of an eye) and simply accepting that after working with them for a few weeks there’s no going back.
And did I mention: Ctrl-C, Ctrl-V… Ctrl-C, Ctrl-V… Ctrl-C, Ctrl-V… Ctrl-C, Ctrl-V… Ctrl-C, Ctrl-V… Ctrl-C, Ctrl-V…
Before all this I was a skeptic and only partially interested in the idea of these personal assistants of the future. Loni would rave all about them. They helped to accelerate her productivity she said. I should try it! But aside from silly little Gen AIs for images, and a story here and there and perhaps Google-like questions (short prompts) I really wasn’t finding a need to return.
Then Loni had to travel for two weeks. During my time alone with just AI I discovered something important: my killer app. Two killer apps in fact. Once I realized how fast these bots could help me solve problems — often annoying and tedious ones that I loathed addressing — and enable me to spend more time on the fun stuff or the valuable stuff then things started to click. The killer app gave me reason and purpose whenever I opened a new chat with ChatGPT or Claude. My hyper-accelerating learning by just doing and doing for about 10-12 hours a day for 14 days straight led me to a whole new world of discovery… perhaps this wasn’t all just hype.
Here in Silicon Valley that’s all we hear about these days. AI this and AI that. Huge amounts of venture have flowed as you might expect when the next big thing rolls into town. Anthropic, one of the smaller players, has an impressive market capitalization of about $60B (probably more by the time you read this). OpenAI (ChatGPT) is about $300B (!) in market cap already. To put that in perspective, that’s $50B more than a successful tech company like Salesforce which has been around for 26 years and, itself, considered a fast growth mover.
Now I’m beginning to understand all the fawning and headlines.
Oh, I’m also experiencing something new: the mind melt. But first, my experience during those two weeks.
Week 1 The Discovery: Storyboards, short film ideas, scoring and creative exploration
First, I decided to try out four chatbots: ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google) and Copilot (Microsoft). I really enjoy comparing products. Seeing which user interfaces resonate, and how each handles typography, in addition to how fast they would respond to prompts. And, of course, how helpful they were or were not in helping.
Of the four I already subscribed to Claude ($20/month), but didn’t use it much. I thought that was pretty expensive for yet another subscription. But, boy, today, it seems downright cheap. I say get in while you can and put these to work while things are still in the Wild West mode. If my suspicion is right over time the market will consolidate and prices will generally go up (though I think there’s always room for entry and free tiers).
Oh, by the way this is me. This is truly me, Clinton, sitting here typing. No copy/paste. Just me. Really me! You can possibly tell by the lack of emojis and lack of overuse of em dashes — ChatGPT.
My first order of business was basically brainstorming. I’m working on new ideas for short films, experimental stuff and creative projects for Stark Insider — some of which Loni reflects on in What AI Taught Me About Being an Artist. Why not see if I could bounce these around with AI while Loni was on the road and I had no human partner to do so?
Quickly I realized the potential.
In Search of a DAW
When I asked what ChatGPT thought about Ableton Live and FL Studio, it informatively provided alternatives and matched them to exactly my use case for music production (using that term lightly) and scoring (same). After a bunch of pretty comparison tables, pros and cons, and all sorts of interesting suggestions it arrived at Cubase as the DAW (Digital Audio Workstation). Great. Cubase it is.
Then, like a patient music professor who never tires or eats, it asked me if I wanted to learn how to use Cubase, and maybe some basic steps and onboarding.
Off we went. Soon enough, it was even helping 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…
D-50 Pad: “Fantasia” chord held (Cmaj7)
Juno Arp: slow pulse arpeggio (C-E-G) with reverb swell
Jupiter Bass: single root note every bar, low-passed
FX Layer: Portal plugin on white noise, slowly automating mix to rise subtly
One pattern. Full mood.
Then:
Instructions to Rebuild in Your DAW
Open your DAW (Cubase, FL Studio, etc.)
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)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 barsAdd reverb and slow filter automation if desired
And so it went.
Soon it was also advising me on what creative tools to use.
I dumped it my entire list including things like Cardflow+ (for digital storyboarding old school using index cards), Adobe Premiere Pro, Microsoft OneNote, among so many others. I found ChatGPT particularly effective at understanding my needs and rounding out a slew of specific recommendations. In particular it really wanted me to try Notion as a OneNote (and Google Docs) replacement. Also it thought a Kanban style to-do and due date tracker would help. Trello was the pick.
From OneNote to Notion: Total Migration
Notion. My god. What have I been missing. That is one slick piece of software. In a jiffy I had migrated all my OneNote pages into Notion. Tech Docs that used to be in Google Docs to track Stark Insider tech stuff were also moved over.
ChatGPT helped me understand the interface with astonishingly helpful pointers and tips. All step-by-step. When I couldn’t find a menu or a button, I’d screen shot it and paste it and an answer would come back (sometimes I learned these chatbots are trained on older versions of software and web pages and aren’t quite up-to-date, but they quickly learn and adjust). In the end, these are incredibly patient and skilled teachers and instructors.
In the past we might go to a YouTube video, or the product’s support/help site and look for tutorials. That’s still valid I’m sure. But I discovered a key difference. ChatGPT knows I use Premiere, and now Cubase, and make experimental shorts and silly things that aim to be anything but viral. It knows my workflow because I taught it over several days. Hence, when I get step-by-step instructions on something there’s a constant stream of Pro Tips as well that relate specifically to my work environment. That’s added value beyond just generic corporate support pages.
Location Scouting for Film Production
Then I wanted to scout locations in San Francisco for a video.
Again, my jaw was nearly on the floor. Without hesitation, ChatGPT amassed scores of places to shoot, all exactly matching my requirements. Then it mapped out a timeline and project plan so I could minimize drives up 101 to SF and efficiently shoot scenes in logical groupings without having to criss-cross all over town. As you might expect this was all neatly summarized and presented in tables (and PDF should I want).
I fed it a bunch of fragmented ideas, some more developed than others. Lines of text. Images. And even links to Stark Insider articles and pages. I’d also paste in all sorts of creative randomness so it would get to know my style.
After that, brainstorming on new concepts was a hoot.
The ideas flowed faster than anything I’ve experienced in my life.
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.
Really, it all seemed too good to be true–how can this be real? I’ve been chatting all day (and night) with a computer and it’s accelerating my ability to get stuff done. Not only is it all faster, but the quality is upped as well. I could not achieve all of this alone. Granted, I still needed to sometimes steer it in the right direction or clarify something, but that’s how it is with humans too no?
At this point I realized it was almost midnight. I hadn’t really paid much attention to the Stanley Cup final and instead was absolutely dialed in and rivetted by this experience. Many had already done so, and, finally, I had found at least one killer app. ChatGPT was a creative partner.
Then I also realized 7 days had passed. I was all absorbed and losing track of the real world. The Mind Melt was soon to follow.
Week 2 The Descent: The IT Dungeon
To say I was thrilled by all this creative progress would be an understatement. It was a renaissance. Bring on the hyperbole!
About this time, as I was dusting off my Roland A-49, I started receiving urgent email alerts that starkinsider.com was down. A bunch of links weren’t working. Bells were going off. Sigh. Not again. Keeping WordPress running has been one of the joys and despairs of the past 20 or so years running this site.
An idea hit me. Why not use ChatGPT and friends to try and fix it?
I knew that I needed to COPY something and PASTE it somewhere.
So I did just that. Went to the server error log. And copied a ridiculous number of lines and just pasted all that into not just ChatGPT, but also Claude, Gemini and Copilot. Let’s go boys! Solve this thing please!
Killer app #2 was born.
Here, Enjoy this Error Log!
One by one, step by step, the troubleshooting tips came in. Ever so precise and methodical. Granted it wasn’t an instant fix (are these things ever?), but it did things I could never do on my own. I mean, who wants to waste a day (or two) sifting through error codes from an NGINX error.log file?
Soon enough, this web site was righted again and all was apparently fine.
But that didn’t stop the bots. I should modify this config file, tune this here and that over there. Oh, also upgrade Ubuntu, MySQL and NGINX and Redis while you’re at it. Oh, ok BuddyGPT!
Also, they tell me, install Fail2ban to harden security.
Check which bots are hitting Stark Insider. Are there any AI bots? Which ones do you want (ones that offer attribution for your work) and which ones don’t you want (block: Amazon Alexa)?
So the site was running, and now I was in optimization mode. Meantime, ChatGPT was telling me how to setup monthly checks using Trello checklists and reminders. How to use Notion to keep track of this stuff.
ChatGPT also re-wrote backup/snapshot scripts that were more efficient and saved us about $10/month in recurring costs without impacting recovery (always cross-check with others, especially Gemini on this stuff to avoid those hallucinations).
TIP: Turbocharge Your Server Reports with AI
Before firing off 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 EnglishThen email the AI-enhanced version. It’s like having a sysadmin with a PhD in storytelling. Your future self (and anyone else on the distro list) will thank you.
Through this process I learned that Claude is a real winner when it comes to coding. Or, at least, in this case, simple scripting and configuration. The baseline stuff to be sure. Watching Claude code in real time on the right side panel and use version control and summarize all the changes from one to the other is a sight to behold. Sort of like the first time Amazon Alexa responded to a voice command level amazement.
ChatGPT vs Claude vs Gemini vs Copilot
I learned a few other things to like:
ChatGPT can sometimes go in circulars when it comes to fixing scripting bugs. For example, it kept suggesting nested “If” statements whereas Gemini said that was a no-no in a php config file for NGINX. It agreed, oh yes of course! Only to come back around 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-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’s ways to help that which involves… copy and paste).
By comparison, ChatGPT has a long memory that cuts across threads and topics
Gemini is absolutely the gold standard as you might expect 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 as in: 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 lay out the potential downsides first. In a very straightforward manner.
Copilot is decent, but I don’t often find myself making to the fourth tab as the others have me covered
I slowly crawled out of the IT Dungeon, scorched eyebrows and all.
We were on to higher order tasks. Like improving our SEO and content. ChatGPT once again was just aces. It pulled together content plans and schedules. Helped with SEO for articles, even so much as providing every meta fields for images (a thankless task). I could see how this might impact Ahrefs and Semrush potentially.
Copy and paste once again. Capture a Google Search Console screen full of (depressing) errors and let Claude and ChatGPT and Gemini (especially) navigate your way through all the chaos.
The Mind Melt & Reflection
When Loni arrived back at SFO, I talked her (tired) ears off back down 101 to San Jose. I think I was nearly hyperventilating with excitement.
Two killer apps: Creativity and Tech. Now having all 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 expression?).
I was exhausted. I was getting 5-6 hours sleep at best. Wasn’t using the Peloton as much as I should be, 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 sort of upheaval? How will it impact our knowledge, our brains? Is this good or bot or a bit of both? So many questions.
Here’s my quick assessment of each of the chatbots 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 the coding of Claude and its ability to complete code that ChatGPT can’t)
A wry sense of humor if you want, 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, exec summaries
Was effective in evaluating server expenses (just dump spreadsheets into it, or raw data) and implemented solid Google Snapshot cron job & retention policy that should save about $10/month in costs
Fun to use
Long memory
Strongest writing partner of the bunchClaude
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 up-level your server reports!
Might occasionally surprise you with a custom app (without prompt) on the right panel that you can interact with to solve a problem
Open to other’s critiques and ideas. For instance, if I paste in Gemini or ChatGPT alternative approaches to a problem, Claude will often accept it as a good idea and offer a hybrid, combining the best of all worldsGemini
Feels quite obviously trained on Google documentation of all kinds
That’s not a bad thing in the least especially when, in my experience, it came to Google Cloud, Google Search, and SEO
Probably the most methodical at explaining things; lots of detailCopilot
Like 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 LLMsDisagreements Among the Chatbots?
You may notice different approaches to solving problems from on AI to the next
At times, I felt like a moderator in a room with 4 experts debating a solution
This is where moderation and arbitration becomes important
In the end, human judgement is still essential in these situations: you just get more information to synthesize and, therefore, in theory at least, you 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’ll 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 for real estate or banking. Or reviewing budgets. Or therapy. Or gambling probabilities. Or video game development. Or deep research and problem solving. Or help writing papers. Or an email assistant (again, copy/paste). Or… the list goes on. To me, this is as big a transformation as search and e-commerce and the web was back in 1999-2000.
I realized during this process that I wasn’t just working faster. AI was actually extending my skillset above and beyond what I could otherwise achieve (i.e. I have no idea how to code or create scripts). Combine this with the speed and efficiency (four tabs, four AIs, multi-tasking) and at times you can feel even bionic, making it hard to let go as you just want more and more of that adrenaline rush that comes with solving problems and checking off those digital post-it checkboxes. Done and dusted.
Just don’t forget to step outside once in a while.
More espresso-fueled experiments:
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!)