Two Screens. Two Worlds. One Tool.
# categorized RSS feeds,
# custom layouts, and
# pagination — from a
# conversation.
$ claude "Add a new blog post
with category filtering,
RSS feeds, and update
the index"
✓ Created logs/aimplified.html
✓ Updated logs/index.html
✓ Updated logs/feed.xml
✓ Updated logs/feed-tech.xml
# Production code.
# From a prompt.
On the left, production code. On the right, a graphic I turned into stickers and sent to people I care about. Same tool. Same afternoon. Entirely different domains.
That is the part most people are not seeing yet.
The Amplification Nobody Expected
The narrative around AI right now is narrow. It is a coding assistant. It is a chatbot. It is a search engine with opinions. The discourse stays locked inside the tech industry's frame because that is where the money is, and money shapes the conversation.
But the actual utility is logarithmic. It does not add to what you can do. It multiplies it. And it multiplies it across domains that have nothing to do with software.
In the past year, AI has taught me how to:
Soil Science
How to make biochar. How to mix it with worm castings. How to set up and maintain a worm farm from scratch. The entire composting pipeline — from kitchen scraps to soil amendment.
Apothecary
Lotions. Body wash. Candles. Room sprays. Wax melts. Full formulations with ingredient ratios, safety guidelines, and curing times.
Creative Design
Custom graphics turned into physical stickers. Visual assets for websites and products. Brand design iterations that used to require a designer and three rounds of revisions.
Production Software
Full-stack applications. Jekyll sites with custom architectures. API integrations. RSS feeds. The kind of work that used to require a team and a timeline.
That is not a list of hobbies. That is one person operating across four completely unrelated disciplines — at a level of competence that would have previously required either years of study or hiring specialists in each domain.
The New Specialization
People keep asking what they should learn to stay relevant. The answer is not a programming language. It is not a framework. It is not prompt engineering, whatever that means this week.
The new specialization is yielding AI. Learning how to collaborate with it. How to direct it. How to know when its output is good and when it is hallucinating. How to take a vague instinct — "I want to make candles" — and turn it into a structured learning path with actionable steps, sourced information, and tested formulations.
That is a skill. And it is a skill that compounds. Because once you learn how to learn with AI in one domain, you can do it in every domain. The meta-skill transfers. The biochar research taught me patterns I used when learning apothecary. The apothecary work refined my prompting patterns, which made the software development faster. It is a flywheel.
The Collaboration Gap
Most people I talk to still think of AI as a tool you use to replace a Google search. Ask a question, get an answer, move on. That is like using a car to sit in your driveway and listen to the radio.
The actual leverage comes from collaboration. Extended back-and-forth. Iterating on formulations. Debugging a worm bin the same way you debug an application — describing symptoms, testing hypotheses, adjusting variables. AI is not an oracle. It is a thinking partner that never gets tired and has read more than you ever will.
The gap between people who understand this and people who do not is going to define the next decade of careers. Not because AI replaces jobs — that framing is lazy and mostly wrong. Because AI amplifies the person using it. A curious person with AI access becomes dangerous. A builder with AI access becomes a team. A learner with AI access compresses years into weeks.
Log Scale
I keep coming back to the logarithmic framing because linear thinking undersells it.
If AI made you 10% better at one thing, that would be useful. Interesting. Forgettable. What it actually does is make you 10x capable across an unbounded number of things. And those capabilities stack. The person who learned biochar and apothecary and software development and graphic design is not four times as capable. They are operating in a different category entirely. They see connections between domains that specialists never will.
The compound effect is the part nobody is talking about. Every new domain you learn with AI teaches you to learn better with AI. The returns accelerate. That is not addition. That is not even multiplication. That is logarithmic amplification.
The Hedge
Now that I am two years ahead of where most people are with this technology, I do not mind admitting something: I am using AI to hedge against AI.
The apothecary work. The soil science. The worm farming. The candles and lotions and body wash. None of that is accidental. Those are physical products. Tangible things that require hands, raw materials, and the kind of embodied knowledge that no language model can replicate. You cannot 3D-print a lotion. You cannot deploy a candle to production.
AI can teach me the formulations. It can help me iterate on ratios, troubleshoot emulsification failures, optimize curing times. But it cannot pour the wax. It cannot harvest the castings. It cannot hand a bar of soap to a neighbor. The final mile is irreducibly physical.
That is the hedge. Use the amplifier to accelerate into domains where the amplifier itself has no foothold. Learn faster than the market can commoditize. Build skills and products that exist in the real world, where the moat is not code but craft.
The irony is clean: the best defense against AI disruption is AI-accelerated competence in things AI cannot do.
#AImplified.
The Protocol: Stop thinking of AI as a tool for your job. Start thinking of it as an amplifier for your curiosity. The people who will thrive are not the ones who use AI to work faster at one thing — they are the ones who use it to become capable across many things. The new specialization is not a skill. It is a multiplier. And the smartest play is to aim that multiplier at things no model can touch.
