We are building the next 100 tomorrows today
Technology is moving faster than our institutions, faster than our ethics, faster than most people can process.
We are on the cusp of a future we cannot yet imagine. The rate of progress in generative systems, reasoning, and agentic behavior is unlike anything in recorded history. What used to take teams of people years can now be prototyped in days.
I am not afraid of the speed. I am in awe of it. The danger was never moving too fast. The danger is moving fast while remaining spiritually and philosophically asleep.
The physical world is finite and the living world is fragile, and the next generation will inherit whatever we decide to extract or preserve while we race to ship. Sustainability is not a marketing word for us. It means refusing to externalize costs onto people, places, and species who have no seat at the table. The same intelligence that can optimize a supply chain to the last mile can also be used to waste less, pollute less, and ask better questions before we build. The tools now exist to solve problems that have plagued humanity for centuries, and, just as readily, to create new ones at unprecedented scale. We get to decide which version of intelligence we unleash.
Many fear that AI could turn against humans. What should be feared is the risk of few humans controlling that AI. How the future looks entirely depends on whose hands the power ends up in, and what values will be encoded into the systems that are on track to control the majority of the modern world.
Power Struggle
There are 342 million people in the United States, and billions more beyond it. These are people and businesses with creative and revolutionary ideas that could change the world, if they have access to the tools and opportunity to execute on them. AI has the potential to provide that opportunity faster and more broadly than anything in history, but it is increasingly being funneled through choke points. The future power structure is being built right now. If we get this wrong, the consequences will be global and lasting.
Protections Of Distribution
One protection against over reaching power is mass distribution. If the power is spread amongst many people, there is a greater likelihood of balance. The more companies that can adopt automation and AI powered systems, the more evenly competition increases and power is distributed. The more AI model providers that are available to choose from, the less control over information and access that can be exerted by any one provider. However, widespread distribution is facing a major obstacle. Implementation is still too complicated.
Technical Challenges
To build even a small specialized agent today you have to set up a harness, manage the memory and context that feed it, understand tool calling, stand up MCP servers, and build connections into every service your workflow needs. You have to author the right skills to tell the agent precisely which tools to use, in what order, and how, build safety layers and guard rails that allow autonomy while preventing catastrophe, and then troubleshoot constantly when it starts drifting off track. This is at least one full-time specialty engineering job for a small business, and a team for anything more. People that are not experienced developers are already at a significant disadvantage.
If you are a large enterprise, you hire teams of experts to build and maintain all of this. If you are anyone else, you do without.
Single Provider Solutions
Many large labs are also coming to the same conclusion. Anthropic, OpenAI, and Perplexity have all released dashboards and apps that abstract much of the complexity to increase adoption. They are sending teams into large organizations to implement their technology stacks into the companies.
However, these solutions require betting everything on a single provider. The specialized agents, workflows, training, customized automations, all of it is now captive to models and infrastructure from a single company. This is incredibly high risk, especially so early in the game. We have already seen models get temporarily disabled by government intervention, and access to certain capabilities being restricted to government approved ‘trusted partners’. This is only the beginning of the fight for power. Full commitment to one provider without a backup plan leaves exposure to becoming collateral damage.
Expanding Access
True access means more than a usable interface. It means open source models that anyone can study, modify, and run on their own terms, so that the foundations of intelligence are not held behind closed doors. It means access to the equipment and compute required to train, fine tune, and deploy those models, so that capability is not determined by who can afford the largest data center. And it means the freedom to innovate and experiment, to try ideas that fail, to build things that seem impractical, without asking permission from the handful of entities that currently control the infrastructure. When the models, the equipment, and the freedom to experiment are all distributed, the result is a far larger and more resilient ecosystem than any single provider could build alone.
What We Are Building
AgentPMT exists to lower one of the barriers to distribution by giving individuals and small teams access to powerful AI automation infrastructure without the prohibitive price tag. The premise is simple: no code, no technical background required, only a clear sense of the process to be completed. There is no vendor lock in, no model lock in, no long term contracts. We believe the best version of the future requires widespread access to knowledge and technology, and that is what we strive toward every day.