I have recently been reading Accelerando by Charles Stross. While it was written in 2005, some of it's ideas resonate strongly with today's era of LLMs and Agents.

Manfred: The Human-Agent Hybrid

In the (earlier portions of) the book, Manfred, who is super into tech, is exponentially more capable than his peers because he really leans into his technology. He has hundreds (or thousands) of "agents" running simultaneously that go out on their own to:

  • Do research, gather up-to-the-minute news
  • Run experiments
  • Confer & share information with other agents
  • Develop marketing / exploitation plans
  • Communicate with other external agents / APIs behind the scenes
  • And lots more

The idea is, this human's "meat brain" becomes only one small component of the system that encompasses Manfred, and his public image. He is just one small node in the system that is "Manfred".

Today’s Reality: Intelligence on Tap

We kind of have a version of "intelligence on tap" now with LLMs. We can make a simple API call with some input context, and receive back intelligent analysis and output.

We do have some narrowly defined agents (deep research, coding, simple workflow automations) that we can download or define ourselves - and if we give them good inputs and effective tools, they will make decent decisions and provide decent outputs.

Today, if you have some system design skills, you can make these agents yourself pretty easily. I myself have made a very useful coding agent, an email assistant agent, and a few others.

OpenClaw lets you create a multi-use agent that runs on your local machine (but uses API calls for the actual LLM completions - usually to Claude), and has skills which are text descriptions on how the agent can use a specific service (via API calls or whatever).

Agents for Goal Achievement

These agents only really work well when you give them a specific narrowly defined goal. When you do, and you set an agent running on it, they help you achieve more, and help you automate parts of your work.

Finding utility in agents really depends on you defining a specific goal, and providing the agent with enough tools and context for it to start providing value.

The Ultimate Goal: Making Money

At the foundation of most personal and business endeavors is the goal to make money. Money is what gives us:

  1. Security and stability
  2. Luxury items
  3. Flexibility to spend time on passion projects
  4. Makes social interactions easier

Using Agents to Generate Income

What kinds of money-making tasks can agents help with?

  • Finding exploits to systems
  • Finding arbitrage opportunities
  • Making well informed bets on prediction markets
  • Making entertainment / educational content for consumption by others
  • Bring a scarce resource to someone who is willing to pay for it

The possibilities expand as the agents’ competencies and autonomy improve.

Parallels With Employment and Company Structures

The way we envision agents closely mirrors traditional employment:

  • As a company, you have many humans (agents) that work for you
  • You have a centralized goal (generally profit-making)
  • You want your agents to help you to attain more profit (by providing, improving, doing research)
  • It is a director's duty to list out the specific job roles for each agent in the system

In summary - to have a company:

  • You need the shared goal (profit-making)
  • You need a way to attain that goal (a plan)
  • And you need instructions for every agent/component that will work to help attain that goal

I am an agent

As a sole owner and operator of my business, I play the role of both a director, and an agent:

  • Director: The planner who decides what strategies to pursue and sets goals (with the ultimate goal of profit-generation)
  • Agent: The executor who performs research, analysis, and task completion using various tools

Occasionally, my director self puts my agent self to work on building new tools that make parts of the process more efficient, enhancing my overall capacity.

Conclusion

The age of intelligent agents is here, and it's a toolset anyone can harness with curiosity and effort. By thinking like both a director and an agent - setting clear goals and building or adopting smart assistants to help you can exponentially increase your capacity to create, earn, and innovate. Embracing this paradigm will redefine productivity, creativity, and even what it means to be human in the years ahead.