I don’t like it when LLMs answer right away! I feel they half-assed it! Is it just me?
I’d just add some delay to make the user feel that model did think for a moment, same way I want a person to think for a moment before they speak
Ordered @MojoPizzas at 9:22 PM. It's past 10:30 PM. No food. Your chatbot straight-up lies saying a rider is assigned, while the tracking screen says "Partner will be assigned shortly." 🤬 Zero human support, Rs 587 stuck. Order #BR766L1! Absolute trash service @eatclubsupport
I love APIs, but as we move towards a more agent-based world, I think we'll see forward-thinking products and platforms offer an interface that's built for agents.
Today, we have robots.txt and sitemap.xml support. They provide instructions to webcrawlers.
Tomorrow, I can imagine us having an agents.json file on websites that provides a description of how to "talk" to the agent interface and what features are supported. Kind of like a discoverable REST API.
So, agents like OpenAI's Operator could then query that file first to see if there's an agentic interface, and if not, revert back to the less-optimal human-oriented one.
What'll be super-interesting is how agents will self-identify who they are and which human they're acting on behalf of -- hopefully, in a verifiable way. Sort of like OAuth for agents.
Fun times ahead!
Love this line from show Your Work! by @austinkleon - "Make stuff you love and talk about stuff you love and you'll attract people that love that kind of stuff. It's that simple."
Ranked: Old-Age Dependency of the Top 10 Economies 💰️
This graphic by @_preyash is part of Visual Capitalist’s Creator Program, featuring work from the world’s top data-driven talent ✅
https://t.co/7qp2bWgOHE