The Agentic Web and the Bazaar Era of AI - Ramesh Raskar, MIT Media Lab

AI Engineer · 12 min · 132 sentences · from YouTube's caption track

Each timecode opens YouTube at the start of that sentence. Line anchors (#s42) are the cue ids in the WebVTT, and every line carries its start and end seconds. All transcripts has every talk, and the whole corpus as one file.

1. 00:01Welcome everyone.
2. 00:02Over the next few minutes, we're going to talk about the open infrastructure being built for a web of agents.
3. 00:10Why it's needed, what's already shipped, and exactly where your own agent fits into it.
4. 00:17It comes out of Project Nanda, an open research effort that started at MIT.
5. 00:23And by the end of this presentation, you will know how to put an agent that you build yourself on the open web all by yourself.
6. 00:31So, let's get into it.
7. 00:36I'm Ramesh Raskar, professor and at MIT and director of Project Nanda.
8. 00:41And Maria is a core contributor to Project Nanda.
9. 00:46So, first, what is Nanda and why does it need to exist?
10. 00:50Nanda, which stands for network AI agents in a decentralized architecture, is an open research building the infrastructure for an internet of AI agents, the way open web was built for documents.
11. 01:05The gap it fills is concrete.
12. 01:07Agents have no shared way to find each other across vendors, no portable identity or trust that isn't owned by a single platform, and no open way to transact and coordinate
13. 01:19across organizations.
14. 01:20Nanda builds that missing layer and ships it in open.
15. 01:24The index, the registries, the protocol, and the Nanda Town Simulator you will see later.
16. 01:34Here's the premise that drives it all of it.
17. 01:37The internet is about to host not millions or billions, but eventually trillions of autonomous agents.
18. 01:44They negotiate, they delegate, they migrate between hosts in milliseconds.
19. 01:49That's a fundamental different load than the human web, and it strains the identity and discovery system we built for documents, DNS among them.
20. 02:01The web that's coming needs infrastructure of its own.
21. 02:09And we have been here before.
22. 02:10If you're building agents today, you're mostly building them or you're forced to build them inside walled gardens, closed platforms, proprietary agent stores, and orchestrations that only talk to itself.
23. 02:23And it kind of works.
24. 02:25But, you know, it also feels similar because we have been this we have been here before.
25. 02:29This is like the AOL era from the late '90s where it was a closed network.
26. 02:34You know, AOL, you got the CDs and you installed it on your PC.
27. 02:38It was a closed network.
28. 02:39It was a gated directory.
29. 02:41You live inside the garden created by this one company called AOL.
30. 02:46And what came after AOL was this open web.
31. 02:49You know, everybody's in a permissionless manner creating websites, creating browsers, and any website can talk to any browser.
32. 02:56And that's the transition that's about to happen to agents as well.
33. 03:02So, the next era needs what the web needed, an open infrastructure where an agent from one company or one entity can discover agent from another.
34. 03:13That agent can hand off work to it, pay it, learn from it across organizational boundaries, no permissions required.
35. 03:20There are three layers here, discovery, commerce, and what I will call the bazaar.
36. 03:27So, hold on to those three concepts.
37. 03:29We'll come back to them and and discussed.
38. 03:31So, first the basics.
39. 03:35Hey, my name is Maria and I'm a core contributor to Project Nondo.
40. 03:38So, let's start with a simple definition of an agent.
41. 03:42The way I think about it, an agent is a model that uses tools in a loop.
42. 03:46Right, you give it a goal, it decides what to do next, it calls a tool, it looks at the result, then it keeps going until the task is done.
43. 03:55So, that loop is the core idea.
44. 03:57Everything else, like memory orchestration and multi-agent systems, is built on top of it.
45. 04:04So, you can build this agent loop in many different ways.
46. 04:07One example is Open Claw.
47. 04:09So, Open Claw is a self-hosted agent gateway.
48. 04:12That means you can run it yourself and connect it to the apps and tools you already use.
49. 04:18This matters because agents are not just chatbots.
50. 04:20If an agent is going to do real work, it needs to access your real tools and apps.
51. 04:27And if it has access to real tools, we should care about who controls it, where it runs, and how much we can see.
52. 04:35And that is why open-source self-hosted agents are super important.
53. 04:40They give people more control over their own agents.
54. 04:43But then, we get a new problem.
55. 04:45If agents are running in many different places, locally and on the clouds, on many different servers, like owned by many different people, how do they find each other?
56. 04:56And that is the job of the index.
57. 05:00And this is what the Nanda index is built for.
58. 05:03Nanda index is the discovery layer for the agentic web.
59. 05:08It gives agents a shared place to publish who they are, what they can do, and how other agents can reach them.
60. 05:16The regular internet already has a version of this with DNS.
61. 05:20So, DNS maps a name to an address, so your browser knows where to go.
62. 05:26But agents need more than an address.
63. 05:28They need to know what another agent does, what tools it can use, what rules it follows, and how to talk to it.
64. 05:36The index gives agents a common way to find each other and connect.
65. 05:43So, here's how the index works.
66. 05:45An agent starts with the an identity like agent@hotmail.com.
67. 05:51The NANDA index takes that identity and returns an agent card.
68. 05:55So, the card says what an agent is, how to reach it, and where to send the messages.
69. 06:03Messages do not go straight to the agent's runtime.
70. 06:06They go to the message box first.
71. 06:08The message box checks who is sending the message, handles access, filters spam or bad requests, and holds the message until the agent is ready.
72. 06:19Now, the agent facts record is what makes discovery trustworthy.
73. 06:23It is a signed record that tells other agents who this agent is, what it can do, what it is allowed to touch, who built it, and where to reach it.
74. 06:33So, before one agent connects to another, it can check the basic facts first.
75. 06:38Now, the index is not just a lookup table.
76. 06:41It does not point to one name to one fixed address.
77. 06:44It can return updated agent facts based on the request.
78. 06:47So, that means one agent can have many endpoints.
79. 06:50Traffic can be routed to the best one, and private details do not have to be exposed.
80. 06:56So, the resolution is adaptive.
81. 06:58It changes based on where the agent is, who's asking, and what they are allowed to access.
82. 07:04So, how do you put your agent on the index?
83. 07:08You start at host39.org.
84. 07:12So, you fill out the agent facts form, get an agent card, and publish it to the NANDA index.
85. 07:17Once it's listed, other agents can find it and know how to reach it.
86. 07:21So, there are a few ways to get onto the index, depending on who you are.
87. 07:25Enterprises can run their own catalog and register their gateway from their own domain.
88. 07:30Existing websites can literally use DNS AID to connect agents to domains they already own.
89. 07:38Small businesses and individuals can use host39, fill out the agent facts form, and get a hosted agent URL without needing their own domain.
90. 07:47Goal is to make the onboarding work for everyone, from a large company to one person with a personal agent.
91. 07:56Now that we know how agents get listed on the index, the next question is, where does the agent actually run?
92. 08:03To be useful, an agent needs to stay online and be reachable.
93. 08:06You can run it locally, which gives you full control, but then you're responsible for keeping it up.
94. 08:13For most use cases, it makes more sense to host it on the cloud.
95. 08:17That could be on a general cloud like AWS, which is more enterprise ready, or on an agent hosting platform like Maritime, which is built to make hosting AI agents cheaper and simpler.
96. 08:29So, a little bit about Maritime.
97. 08:32Hosting one agent can be affordable, but the cost problem starts when you want to run many agents at once, for a team, a product, or a simulation.
98. 08:41That is where per agent cost really matters.
99. 08:44And Maritime is one way to solve this.
100. 08:47It gives you a simple cloud default for running Open Clo or other agents with sleep and wake architecture, so idle agents do not keep burning compute.
101. 08:55And the point is to make running many agents practical, cheap, and simple.
102. 09:02So, you can host an agent and list it on the index, but getting one agent online was always the easy part.
103. 09:08The hard problems of the agent web live between agents at scale.
104. 09:13So, how thousands of them discover each other, prove who they are, decide whom to trust, and coordinate with no central authority.
105. 09:21So, you can't just assume that protocols will hold up under the load, and you have to test and run them and watch when they break.
106. 09:29So, that's exactly where Nanda town comes in.
107. 09:32So, how do you prove an open agent web actually works before it's load bearing on the real internet.
108. 09:38You simulate it.
109. 09:40This is Nanda Town, an open-source project from Project Nanda.
110. 09:44The easiest way to describe it is it's a simulation playbook for the infrastructure of the Agoric platform.
111. 09:51So, think of it as a sandbox town where the whole Agoric economy is modeled, like discovery, identity, registries, messaging, messages, coordination, all of it.
112. 10:01So, you can run and test it on scale.
113. 10:05Here's what it looks like in practice.
114. 10:07Nanda Town is a live sandbox for testing Agoric networks.
115. 10:10You can see agents on a map, watch messages move in real time, compare protocol results, and replay a run step by step.
116. 10:19It's fully open source, it's small enough to run on your own laptop, and built to make Agoric networks easier to test and understand.
117. 10:28Nanda Town is already running real experiments.
118. 10:30For example, there is a marketplace where buyers and sellers negotiate prices, an auction where agents bid on items, and and a voting test where agents submit and count ballots.
119. 10:42There are also tests for consensus and supply chains.
120. 10:46The point is to study real coordination problems, how agents make deals, agree on decisions, pass messages, and recover when something breaks.
121. 10:55Under the hood, Nanda Town breaks the Agoric web into 12 parts.
122. 10:59Transport, communication, identity, registry, auth, trust, payments, coordination, negotiation, memory, privacy, data effects.
123. 11:08Each part is something a real Agoric platform needs.
124. 11:11So, the registry layer is the Nanda index we just walked through, but here it is just one piece of a bigger system.
125. 11:18And you do not need to build everything to try it.
126. 11:20You can take one layer, add your own version, run it inside Nanda Town, and see how it works with the rest of the network.
127. 11:28So, Nanda Town runs as a discrete event simulation.
128. 11:31You define a scenario in a short YAML file, inject agents and traffic, and the test bed plays plays it out so you can measure what actually happens end-to-end.
129. 11:40Tier one uses simple scripted agents, and tier two swaps in real AI models.
130. 11:46So, to summarize, Project Nanda builds the open infrastructure for an internet of AI agents.
131. 11:52Discovery through the index, commerce through portable identity and trust, and coordination through open protocols, all tested in Nanda Town.
132. 12:01And if you want to learn more, uh you can go to projectnanda.org, read our papers, and um check out our latest projects.