Welcome to another issue of Empire of One.

The newsletter for builders who want leverage, systems, and AI, without hiring a team to get there.

If this helps, forward it to someone still pricing out a hire instead of a tool.

In this issue:

1. Why Anthropic's new pricing for Claude Sonnet 5 matters more than the model itself

2. Why cheaper agents change what hiring means for a business of one

3. What to actually do with cheaper leverage before the price goes back up

Anthropic released a new model this week. That's not the story.

Most people will read it as a benchmark update. Another leaderboard shuffle. Which model codes better, which one reasons longer, which one wins this month's chart.

That framing misses what actually changed.

The real shift is the price of running an agent just dropped below what a lot of solo operators were still budgeting for a part-time hire. Not the price of a chatbot. The price of something that can plan, use tools, and finish a task without you sitting over it.

Most solo operators still treat AI like a nice-to-have expense. Meanwhile the actual cost of the alternative, a person's time, keeps moving in exactly one direction.

Every model release gets covered the same way.

A benchmark chart. A capability demo. A debate about whether it's smarter than the last one.

None of that tells a solo operator anything useful, because the question was never which model is smartest.

The question is what does a unit of agentic work cost now, compared to what it cost you six months ago, and compared to what a person would charge you for the same output.

Chasing the smartest model is a hobby. Watching the cost curve is a strategy.

Here's what actually happened, according to TechCrunch.

Anthropic's new Claude Sonnet 5 is priced at $2 per million input tokens and $10 per million output tokens through August 31. That's cheaper than Opus 4.8, GPT-5.5, and Gemini 3.1 Pro. After August, it moves to $3 and $15, still on the low end for what it does.

What it does is the part that matters. Sonnet 5 can make plans, use tools like browsers and terminals, and run autonomously at a level that, just a few months ago, required larger and more expensive models.

Anthropic's own framing is blunt about it: agentic capability is now the baseline expectation at every price tier. The competition isn't about who can reason best anymore. It's about who can do it cheapest without falling apart.

That's the actual headline. Leverage keeps getting cheaper faster than headcount ever will.

You don't need to switch every workflow to a new model this week.

You do need to look at the launch pricing window closing on August 31 and ask a specific question: is there one task you're still paying a person to do, or still avoiding entirely because it felt too expensive to automate, that this price point now makes worth testing?

Pick one. Not five. One task with a clear input and a clear output.

Run it. Compare the actual cost and the actual quality against what a hire would have run you for the same output. Let the numbers make the decision instead of the hype cycle.

A business should increase your agency, not consume it.

Every time the cost of agentic work drops, that's one more piece of the business that can run on a system instead of on you. Not because the model is impressive. Because the unit economics finally make sense for someone operating alone.

The founders who benefit from this aren't the ones chasing every new release. They're the ones who already know which task they'd hand off the moment it got cheap enough.

Be that person before the price goes back up.

Before you upgrade any workflow to a new model, price out the same task with a human hire first.

If the agent is cheaper and the quality holds, that's your leverage signal. Not the benchmark score, not the demo video. The actual number next to the actual number.