AI power users are here and the case for agent-friendly software is materializing. Aaron Levie on agents outnumbering people 1000x: "If you start to imagine that we all have to build software for agents… we spend as much time now thinking about the agent interface to our tool as... we do the human interface." "If you have a hundred or a thousand times more agents than people, then your software has to be built for agents… it's going to be through an API or a CLI or MCP." "What if you give a coding agent access to your SaaS tools and a coding agent access to your knowledge work, workflows and context?" "It can actually code its way or use APIs through whatever task it's trying to achieve." "That appears to be like a paradigm that is starting to compound… and I actually think it kind of makes sense as the ultimate manifestation of this stuff." Aaron Levieshow more

coding its way through whatever task it needs is exactly the line that should make security teams nervous

an agent using apis through whatever task it needs is just a script with worse error handling

Agent friendly usually means an API you already had, documented properly

The interesting part is that agents may kill the UI moat. If an agent interacts with every SaaS through APIs, the product with the best interface no longer necessarily wins. The product with the best infrastructure does.

sure, aaron, whatever you say

Aaron Levie wants software built for agents. The real bottleneck isn't building it, it's the average human tolerance for being managed by a bot that hallucinates your name twice before lunch.

The 1000x number changes what a feature costs, not just who uses it. Human traffic follows attention, so a rough edge on a rare path stays cheap because few people ever reach it. An agent reads the docs and exercises every path uniformly. The long tail stops being a tail.

If agents outnumber people 1,000 to 1, what do they do after they can call an API? Talking to software is coordination. An economy starts when they can price a job, settle it, and carry a reputation for the outcome. That’s the piece after “more agents.” 👇

the honest answer is, most companies still haven't figured out how to design for human-machine collaboration. till then, agent-friendly interfaces will keep being a kludge

This is where the agent economy gets interesting. When agents become the primary software users, reliable machine-readable context becomes infrastructure. @ekosproject is building exactly that layer for enterprise knowledge.

So are we still bearish on knowledge work replacement at scale?

damn

Agents humans soon 🔥 1000x scale is wild

If you are still using other peoples software, you are not an AI power user. Entry level at best.

@ekosproject

If agents outnumber people 1000x, they cannot share one laptop, one filesystem, and one set of credentials. Isolated compute is the actual interface. A VM per agent, or the blast radius scales with the agents.

The API, CLI and MCP point is the practical tell: when agents become the main users, documentation stops being a nice-to-have and becomes part of the product.

Most product teams are still treating APIs as an afterthought while building for human UI. If agents are outnumbering users 1000x, your API and context layer are your primary product surface area.

This is the bigger shift: software won’t just need a good human interface. It’ll need a good interface for machines that can actually use it.

Agent interfaces may end up looking more like API contracts than UIs.

Would love to collab by providing an infrastructure for your agents and the proper tools and SDKs Is this something we can discuss with the right person at a16z ?

the ui for humans is becoming the legacy layer, api is the new focus

yeah the api is becoming the real frontend for agents

1000x agents means 1000x rate limits and quota fights. agent-friendly software mostly means idempotent writes and a spend ceiling per caller, not a prettier API.

The future is bright for America and her allies. Semper Fidelis

Building the agent interface as a co-equal to the human interface is the part most teams still bolt on last. At 1000x the agent count, does that ratio flip which interface gets the primary design pass first?

Implies separate schemas and action endpoints, not a wrapped human UI

was built for agents I had never built a CLI before but it was obvious that agents would interact with the opensource code via CLI very fun coding and architecting for agents instead of the normal human which color should i make the button hmmm 🤔🤔🤔