Your product is not the center of the universe: think about the person and their agent, not the product

The article, written by Andy Peatling on October 19, 2026 (about a 4-minute read), argues that established products start from a problematic assumption: they believe they are the center of their users' universe and that everything should go through them. The author calls this a product-centric view, in which the product sits at the center and everything else orbits around it.

According to the piece, this mindset has four typical traits: expecting people to come to the product to do their work, building integrations so other products connect directly to it, asking users for context and preferences they have already shared elsewhere, and trying to control access to AI by building its own agent inside the product.

Peatling believes this view has an expiration date. Once a person's agent can work across all of their tools, visiting each product one at a time will start to feel like friction. He says coding agents and personal agents are bringing that date much closer.

In response, he proposes a person-centric view: at the center are the person and the agent they choose. The agent brings in each product's data, interface and context when needed, and over time builds up memory and preferences across all of them, so the user doesn't have to start from scratch. This means changing the defaults: products don't wait for a visit but are available wherever the agent works; they don't build custom integrations but open their capabilities to any agent; they don't ask for context but use what the agent already knows; and they don't create their own siloed agent but connect to one.

The author acknowledges an obvious concern: that this approach turns the product into a commodity, a mere bag of tools at an agent's disposal. He argues it doesn't have to be that way, and that the conversation with the agent can be seen as a new canvas where the product still contributes what the agent can't do on its own. He lists several avenues:

• Its interface as interactive components: previews, editors, customization options or questions that appear inside the agent's conversation right when they are needed. The article's example is a full design editor running inside the conversation, showing only the controls needed at that moment, served via MCP.

• Its expertise: skills that teach the agent to do the work the way the product would.

• Its moat: data, content or network that the agent cannot get anywhere else and that force it to turn to the product.

• Its follow-through: the agent can draft the work, but the product hosts it, publishes it, collects payment and provides support when something breaks.

The piece raises two reasonable objections. The first is the financial cost of AI: the best models are still expensive, which limits who can use them. Peatling considers this a short-term concern. He argues that open-source models are crossing the usefulness threshold; although they may lag behind the best for a while, small versions will soon be good enough to run capable agents on a laptop, almost for free. He also mentions that free personal agents, such as Meta's Muse, could bring the financial cost to zero, though he warns that paying with your data is a cost in itself. Either way, he believes the cost barrier will shrink.

The second objection is the discoverability gap. Citing his recent work on a WordPress benchmark, he says agents know WordPress well but rarely choose it, stressing that being known is not the same as being chosen. He admits he doesn't yet have a complete answer, but would bet on products that present themselves in a person-centric way.

The conclusion sums up the idea: the person is the center of their own universe, and the products that endure will be the ones that show up best within it. The author invites anyone who disagrees to reach out to him on X.

Editor's note: this is an opinion piece with a personal thesis and bet, not a study with original data; the claims about the future of agents, the cost of open models and discoverability are the author's predictions or judgments.