In less than an hour, I had all of it.

I scrolled through the output and my first thought wasn't “wow.” It was: okay, so where does this actually go?

That question is the whole story of tech right now.

The meat used to be the hard part

For twenty years, “the meat” — the plan, the code, the copy, the design, the system — was the expensive part. You paid people for it. You waited for it. You hired agencies, freelancers, consultants, entire departments just to produce the thing.

That constraint is gone.

Any of us can now ask an LLM to write the business plan, design the brand, draft the CRM logic, build the product inventory, generate a content calendar, spin up a website. The meat is cheap. The meat is everywhere. Nobody's impressed by the meat anymore.

The new scarcity is the box

Once anyone can generate infinite output, the bottleneck moves. It's no longer “can I produce this.” It's “where does this live, and does it stay usable once it's there.”

That's why the tools gaining ground right now aren't the ones generating the content. They're the ones built to hold it.

• GitHub and VSCode — the box for your code.
• Canva — the box for your designs.
• Notion — the box for your databases and workflows.
• Obsidian — the box for your notes and thinking.

I wrote about this shift in Notion Just Gone Geek. Notion didn't get interesting because it got better at generating content. It got interesting because it started acting on what you drop into it, without you duct-taping five other tools together to make that happen. Same lesson I learned when I built a tool with Claude specifically so I'd never have to ask Claude the same thing twice — the model reasons once, something else has to hold and run the result, forever after.

Think of AI right now as the world's fastest, most capable cook. It'll produce a five-course meal on demand, any cuisine, any hour. But a meal with nowhere to be plated is just a mess on the counter. The box is the plate. The shelf. The fridge. The thing that turns output into something you can actually use tomorrow, and the week after that.

What this means if you're building

If you're a founder wondering what's left to build in an age where AI produces the meat for free, here's my honest answer: stop trying to out-produce the model. Build the container instead.

Smol OS, the foundation product under Power of Smol, is my own attempt at exactly this. I'm not trying to out-write or out-strategise what a solopreneur could already ask an LLM to generate for her business. I'm trying to build the shelf sturdy enough to hold it, so it doesn't evaporate into forty browser tabs and a voice note to herself three days later.

I don't think I've fully cracked it yet. The box has to be sturdy enough to trust and simple enough that people actually use it, and I'm still finding that balance with real solopreneurs, not just in my own head.

If you had to build one box right now — one place to hold and organise AI output for a specific kind of person — what would you build it for, and who's it for?

#AIWorkflow #BuildingInPublic #PowerOfSmol

Tier 1: Claude for thinking This is where the real work happens. When I wanted to build a personal finance system, I didn't ask Claude to build it. I asked Claude to help me think through what I actually needed, the folder structure, the monthly operating skills, a sensible way to keep my balance sheet updated. Claude helped me design the blueprint. That blueprint becomes the instruction set for everything downstream. So this is where I invest time getting it right.

Tier 2: Gemini for research and comparison Once I have a direction, I don't need deep reasoning anymore. I need fast, reliable information. Gemini handles my desk research. Comparing tools. Pulling facts. Understanding what something actually does before I commit to it. Its context window is large, it's cheaper for this kind of work, and frankly it's better suited to it than Claude is. Right tool, right job.

Tier 3: Gemma running locally, for free This is the part I'm most excited about, and the part most people skip entirely. I run Gemma 4 (e2b) locally via Ollama. On its own, a small local model isn't impressive. But I've been building modelfiles from what I've worked out with Claude, distilling the reasoning, the structure, the decisions into a reusable skill that Gemma can execute. No API. No cloud. No ongoing cost. The effort is front-loaded. Once the logic is right, Gemma just runs it. Indefinitely. I'm still building this out, the library is small, the system isn't fully automated yet. But even half-built, it's already changed how I approach the question of where AI effort should actually go.

I'm not saying everyone needs three models. But if you're defaulting to one for everything, you're either overpaying, underusing what's available, or both.

I'm curious: if you're using more than one AI right now, what's your actual decision rule for which task goes where? Do you even have one?

#AIWorkflow #LocalLLM #Ollama #AILiteracy #BuildingInPublic