For decades, software was designed around one assumption:
A human would be the one using it.
That assumption is starting to break.
We are enteri...
For further actions, you may consider blocking this person and/or reporting abuse
This really changed how I think about APIs.
Until recently I'd mostly think about them as something another developer would consume. Now I'm starting to realize that increasingly the "developer" on the other end is going to be an AI agent making decisions on behalf of a human.
One thought I kept having while reading this is that we're probably moving toward a world where good documentation, structured data, observability and clear interfaces become even more valuable than flashy UI.
It also makes me wonder whether we'll eventually start talking about "agent experience" (AX?) the same way we talk about UX and DX today. We already optimize for users and developers. Optimizing for reliable delegation almost feels like the next layer.
Really interesting perspective. It's one of those shifts that's obvious once someone points it out.
Thanks Aryan! Definitely. AX does sound catchy!
For years, we optimized software for clicks. Now we may need to optimize software for delegation." — This is the absolute mic-drop quote of the year, Julien. Completely agree with your entire breakdown.
The shift toward the M2M (Machine-to-Machine) infrastructure means our metrics for success have inverted. When an AI shopping agent or an IDE agent fetches an endpoint, it completely skips the client-side UI layer. It doesn't care about pixel polish; it cares about reliable context retrieval.
This is exactly why MCP (Model Context Protocol) is gaining such massive gravity. It's becoming the universal OS for this delegation layer.
To address your brilliant checklist at the end ("Is the interface designed for human navigation or agent retrieval?"), we've been building a framework around what we call the Agent Readiness Score (ARS). It basically measures how machine-readable, schema-compliant, and safe a target DOM is before an active agent ingests it. If the site has dynamic rendering blind spots or hidden prompt injections, the agent's delegation path breaks.
We are truly moving into an era of Generative Engine Optimization (GEO). Fantastic piece, man. Def hit a nerve with this one!
Nice! Thanks for validating my points and sharing your cool project!
Another perspective on this is the growing integration of “Product features” into the main UI of models. Recently, ChatGPT launched Sites and Claude with Claude Design. Initially, every product started incorporating AI into its products, but with more powerful models and labs developing their own harnesses, we now have the concept of a generic personal agent that attempts to centralize every problem into its own UI and ecosystem or connectors/plugins.
It appears that the products that align with this new trend are those that focus on extending the personal agent with tools, external data sources, and infrastructure. As you mentioned, the UI and UX matter less if the personal agent is the one managing the entire experience.
I really like this framing. I have also noticed this growing integration of Product features into the UI of models.
I completely agree with the idea of considering AI agents as a new category of stakeholders with their own requirements. One aspect that caught my attention was the emphasis on reliability and safety mechanisms in agentic systems. I think it's crucial to balance the level of autonomy given to agents with the need for human oversight and control. What are some potential strategies for implementing effective human-in-loop controls in agent-first systems, and how can we ensure that these controls are scalable and efficient? I'd love to hear more about the author's thoughts on this matter, particularly in regards to the role of approval systems and policy engines in maintaining trust and reliability.
Thanks Luis.
A fundamental is clear observability of the agent traces. You can't fix something if you don't know how it happened.
From a UX perspective, I also like to build in manual fallbacks if the agent loop fails, when applicable. I also like to give control of the user in validating AI outputs or not and voting on relevance, which becomes valuable input data to improve future outputs.
Making sure the agent has the strict minimum access privileges and tools required to operate is important.
How are you approaching this?
the observability point is the one i keep coming back to. when a human navigates a UI, they correct course naturally — confusion is visible, errors show up in behavior. agents don't do that unless you explicitly build it in. so observability becomes not just a debugging tool but the primary feedback loop.
thinking more about where 'human in loop' checkpoints make sense in practice. upstream policy feels cheaper than downstream audit but the latency cost can break the delegation value entirely. how are you thinking about that tradeoff in the systems you've been working on?
Agree Mudassir. Clear observability is a fundamental for agentic development.
The tradeoff lies in the criticality and SLAs of the system I would say.
Yeah potentially excited, but "really" excited only once I see usable patterns, tools, frameworks, "best practices" and standards emerging to enable all of this - right now it all sounds pretty vague and abstract ...
How would you (re)design your app to make it "operable" by AI agents - rely more on a 'discoverable' REST API, rather than a fancy UI? Of course AI agents can scrape and operate your web UI, but that seems circuitous when it could more simply call your (REST) API ...
Fair point. The idea here is exposing data that is optimized for agent retrieval. There are several frameworks being adopted out there but it still feels abstract at the moment I agree.
Yoi kangae
thanks!