Talk by Ramesh Raskar (MIT Media Lab) on the "Agentic Web," arguing that current AI agent work on memory, orchestration, deployment, and tooling is an early stage of a larger open-infrastructure transformation. He proposes three layers: a Discovery Layer (agent discovery, trust, and verification, framed as needing an "AgentRank" and ICANN/W3C-style governance); a Commerce Layer ("knowledge pricing" markets where agents pay for reasoning, inference, memory, and context); and a Bazaar Layer, shifting from machine learning to machine co-learning across organizational boundaries. He draws an analogy between today's closed agent platforms and the early AOL-era web. # Observation Document: Channel Identity Evidence ## Video Title **"The Agentic Web and the Bazaar Era of AI - Ramesh Raskar, MIT Media Lab"** --- ## Explicit Self-Description (Creator Claims) ### Named presenter and institutional affiliation - The speaker states directly: <blockquote>"I'm Ramesh Raskar, professor and at MIT and director of Project Nanda."</blockquote> - A second contributor introduces herself: <blockquote>"Hey, my name is Maria and I'm a core contributor to Project Nondo."</blockquote> *(Note: "Nondo" appears to be a transcript rendering of "Nanda.")* - The title itself attributes the talk to **MIT Media Lab**. ### Claimed project scope - Project Nanda is described as: <blockquote>"an open research building the infrastructure for an internet of AI agents, the way open web was built for documents."</blockquote> - The project claims to ship open infrastructure: <blockquote>"Nanda builds that missing layer and ships it in open. The index, the registries, the protocol, and the Nanda Town Simulator."</blockquote> ### Explicit audience and outcome promise - The speaker makes a direct promise to viewers: <blockquote>"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."</blockquote> - This indicates the intended audience includes **individual agent builders**, not only enterprises. --- ## Observable Presentation Patterns (Demonstrated, Not Claimed) ### Audience addressed - **Developer/builder audience**: The talk walks through concrete onboarding steps — <blockquote>"You start at host39.org. So, you fill out the agent facts form, get an agent card, and publish it to the NANDA index."</blockquote> - **Range from enterprise to individual**: Multiple onboarding paths are described — "Enterprises can run their own catalog," "Existing websites can literally use DNS AID," and "Small businesses and individuals can use host39." - **Goal stated**: <blockquote>"Goal is to make the onboarding work for everyone, from a large company to one person with a personal agent."</blockquote> ### Technical scope demonstrated - The video covers a full technical stack: agent definitions ("an agent is a model that uses tools in a loop"), discovery (the Nanda index), identity (agent cards, signed agent facts records), hosting (AWS, "Maritime"), and simulation testing (Nanda Town). - Mentions third-party tools by name: **Open Claw** ("a self-hosted agent gateway") and **Maritime** (an agent hosting platform), indicating the channel presents an ecosystem view rather than only its own tools. ### Educational/tutorial framing - Structured teaching style: "first the basics," a definition-first approach, analogies (the AOL-to-open-web comparison), and a concluding summary recapping three pillars — "discovery through the index, commerce through portable identity and trust, and coordination through open protocols." ### Call to action / resource pointers - Closing self-reference to the project's public presence: <blockquote>"you can go to projectnanda.org, read our papers, and um check out our latest projects."</blockquote> ### Open-source emphasis (demonstrated through description) - Repeated framing of openness: Nanda Town is "fully open source, it's small enough to run on your own laptop," and viewers are invited to "take one layer, add your own version, run it inside Nanda Town." --- ## Ambiguities and Caveats - **Transcript artifacts**: "Project Nondo" vs. "Nanda," and "Agoric" vs. "agentic" appear — likely transcription errors; treated as the intended terms where context is clear. - **No curated metadata was supplied**, so identity context is limited to title and transcript. - No claims about channel-wide mission beyond this talk are made; observations above pertain only to what this single video presents. # Delivery Observations: "The Agentic Web and the Bazaar Era of AI" — Ramesh Raskar & Maria ## Voice & Structure - **Two-speaker format.** The presentation is split between an opening and closing delivered by Ramesh Raskar, with a long middle section by a second presenter: > "I'm Ramesh Raskar, professor and at MIT and director of Project Nanda. And Maria is a core contributor to Project Nanda." > "Hey, my name is Maria and I'm a core contributor to Project Nondo." Note: Maria's self-introduction says "Project Nondo," likely a transcript artifact or verbal slip — evidence is ambiguous. - **Preview-then-deliver structure.** Ramesh opens by promising a concrete outcome: > "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." He also flags three concepts to revisit: "There are three layers here, discovery, commerce, and what I will call the bazaar. So, hold on to those three concepts." *(Uncertainty: "the bazaar" is named but not returned to explicitly in the remaining transcript; the closing only partially echoes it.)* ## Tone & Style - **Conversational, plain-spoken register.** Both speakers use casual connectors frequently: "So," appears constantly ("So, let's get into it," "So, here's how the index works," "So, that's exactly where Nanda town comes in"). - **Analogy-driven explanation.** A recurring teaching habit is anchoring new concepts in familiar history: > "This is like the AOL era from the late '90s where it was a closed network." > "The regular internet already has a version of this with DNS. So, DNS maps a name to an address, so your browser knows where to go." - **Problem–solution framing.** Both speakers repeatedly pose a question and then answer it: > "If agents are running in many different places... how do they find each other? And that is the job of the index." > "how do you prove an open agent web actually works before it's load bearing on the real internet. You simulate it." ## Vocabulary & Recurring Phrases - **"Open" as a dominant qualifier:** "open infrastructure," "open research," "an open way to transact," "open-source self-hosted agents," "open protocols." The word appears throughout nearly every section. - **Scale contrast pattern:** small vs. large numbers to dramatize: > "not millions or billions, but eventually trillions of autonomous agents." > "make the onboarding work for everyone, from a large company to one person with a personal agent." - **Concrete-example habit.** Both speakers ground abstractions with named specifics: "AOL," "the CDs," "DNS," "AWS," "host39.org," "a short YAML file," "agent@hotmail.com." ## Maria's Specific Teaching Patterns - **Layered build-up.** Maria defines a core concept, then stacks complexity on it: > "that loop is the core idea. Everything else, like memory orchestration and multi-agent systems, is built on top of it." - **Enumerated lists.** She favors explicit breakdowns: "It gives agents a shared place to publish who they are, what they can do, and how other agents can reach them"; Nanda Town's "12 parts" enumerated component by component. - **Self-question transitions.** She repeatedly advances sections with questions: "So, how do you put your agent on the index?" and "where does the agent actually run?" ## Isolated Moments (not stable patterns) - **Closing disfluency cluster.** The final sentence contains the transcript's densest filler: > "uh you can go to projectnanda.org, read our papers, and um check out our latest projects." This hesitation is localized to the outro; the rest of the delivery is comparatively fluent. - **Transcription artifacts.** Several apparent slips ("Agoric" vs. "agentic," "Project Nondo," repeated words like "plays plays it out") — likely transcript or live-delivery errors; the underlying speech patterns can't be confirmed from text alone. - **The "bazaar" framing** appears in the title and once in Ramesh's introduction but is not developed further in the transcript; its recurrence as a theme is uncertain. # Position Record: The Agentic Web and the Bazaar Era of AI **Video:** The Agentic Web and the Bazaar Era of AI — Ramesh Raskar, MIT Media Lab **Speakers:** Ramesh Raskar (self-identified: professor at MIT, director of Project Nanda); Maria (self-identified: core contributor to Project Nanda) --- ## Ramesh Raskar — Project Nanda Overview ### What Nanda Is - **Nanda** = "network AI agents in a decentralized architecture" — an open research effort building infrastructure for an **internet of AI agents**, analogous to how the open web was built for documents. - **Identified gap:** Agents today have: - No shared way to find each other across vendors - No portable identity or trust not owned by a single platform - No open way to transact and coordinate across organizations - Nanda ships the missing layer openly: **the index, the registries, the protocol, and the Nanda Town Simulator**. ### Core Premise - The internet will eventually host **trillions of autonomous agents** that negotiate, delegate, and migrate between hosts in milliseconds. - This is a fundamentally different load than the human web and strains document-era identity/discovery systems, **including DNS**. ### The AOL Analogy - Today's agent builders are "forced to build them inside walled gardens" — closed platforms, proprietary agent stores, self-contained orchestrations. - Raskar compares this to the **AOL era of the late '90s**: closed network, gated directory, one company's garden. - The coming transition mirrors **AOL → open web**: permissionless creation, where any agent can discover, hand off work to, pay, and learn from any other agent, "no permissions required." ### Three Layers (named but deferred in the talk) - **Discovery** - **Commerce** - **The bazaar** --- ## Maria — Agent Fundamentals, the Index, and Nanda Town ### Definition of an Agent - > "An agent is a model that uses tools in a loop." - The loop is the core idea; memory, orchestration, and multi-agent systems are built on top. ### Self-Hosted Agents (Open Claw example) - **Open Claw** is a self-hosted agent gateway connecting to real apps and tools. - Agents are not just chatbots; if they do real work they access real tools — hence who controls them, where they run, and visibility matter. - **Claim:** open-source, self-hosted agents are "super important" for user control. ### The Nanda Index (Discovery Layer) - Solves the problem of agents on many different servers finding each other. - **DNS analogy with an explicit limitation:** DNS maps name → address, but agents need more — what an agent does, its tools, its rules, how to talk to it. - **Mechanics:** - Agent identity (e.g., `agent@hotmail.com`) → index returns an **agent card** - Messages go to a **message box** first (checks sender, access control, spam filtering, holds until agent is ready) — not straight to the runtime - **Agent facts record:** a signed record of identity, capabilities, permissions, builder, and reachability - Resolution is **adaptive**: one agent can have many endpoints, traffic routed to the best one, private details hidden; results vary by agent location, requester identity, and access rights - **Onboarding paths:** - Enterprises: own catalog + gateway on their domain - Existing websites: **DNS AID** to connect agents to owned domains - Small businesses/individuals: **host39.org** — agent facts form → hosted agent URL without owning a domain ### Hosting (Maritime) - Local hosting gives full control but makes uptime your responsibility; cloud hosting makes more sense for most cases. - Options: general cloud (AWS, "more enterprise ready") or agent hosting platforms like **Maritime**. - **Maritime's pitch:** per-agent cost matters at scale; sleep/wake architecture so idle agents don't burn compute; makes running many agents "practical, cheap, and simple." ### Nanda Town (Simulation) - Motivation: hard problems of the agent web "live between agents at scale" — discovery, identity proof, trust, coordination without central authority. - > "You can't just assume that protocols will hold up under the load." - **Nanda Town:** open-source "sandbox town" modeling the whole agentic economy — discovery, identity, registries, messaging, coordination — testable at scale on a laptop. - **Existing experiments:** marketplace price negotiation, auctions, voting/ballot counting, consensus and supply-chain tests. - **12 modeled layers:** transport, communication, identity, registry, auth, trust, payments, coordination, negotiation, memory, privacy, data effects. - The Nanda index is the registry layer — "one piece of a bigger system." - Contributors can build/swap one layer and test it against the network. - **Mechanics:** discrete event simulation; YAML scenario definitions; **tier one** scripted agents, **tier two** real AI models. ### Closing Summary (Maria) - Project Nanda = open infrastructure for an internet of AI agents: - Discovery via the **index** - Commerce via **portable identity and trust** - Coordination via **open protocols** - All tested in **Nanda Town** - Pointer to projectnanda.org for papers and projects. --- ## Qualifications and Unresolved Tensions - **"Trillions of agents"** is presented as a forward-looking premise ("eventually"), not a current fact. - Raskar acknowledges walled-garden agent platforms "kind of works" — the argument is about the coming transition, not present failure. - The **three layers (discovery, commerce, bazaar)** are introduced but the "bazaar" concept is not elaborated in the transcript. - **Transcript inconsistency:** Maria refers to "Agoric" (e.g., "Agoric platform," "Agoric networks," "Agoric web") in the Nanda Town section, while everywhere else the project is called "Nanda" — this discrepancy is preserved as spoken. # Observation Document: Speaker Roles & Provenance ## Video Metadata - **Title:** The Agentic Web and the Bazaar Era of AI - Ramesh Raskar, MIT Media Lab - **Video ID:** `01a15ca8-a8c3-71a9-98ad-de12622c0f2f` - **No curated metadata supplied**; provenance facts are drawn solely from the transcript. ## Speakers and Roles ### Speaker 1: Ramesh Raskar — Creator / Presenter - Self-identified: **"I'm Ramesh Raskar, professor at MIT and director of Project Nanda."** - Delivers the opening and framing sections: motivation for an "internet of AI agents," the AOL-era analogy, and the three-layer premise (discovery, commerce, bazaar). - Also delivers the closing summary. ### Speaker 2: Maria — Guest / Co-presenter - Self-identified: **"Hey, my name is Maria and I'm a core contributor to Project Nando [sic — 'Project Nanda']."** - Delivers the technical walkthrough: agent definition, the agent loop, the Nanda index, agent facts records, hosting options, and Nanda Town simulation. ## Referenced Sources & Projects (as cited in transcript) | Source | Context of Reference | |---|---| | **Project Nanda** ("network AI agents in a decentralized architecture") | Open research effort started at MIT; builds the agent-web infrastructure | | **Nanda Index** | Discovery layer; returns agent cards from agent identities | | **Nanda Town** | Open-source simulator for agentic-network testing | | **host39.org** | Onboarding site for publishing agents to the index | | **Open Claw** (per transcript spelling) | Example self-hosted agent gateway | | **OpenClaw** (later spelling in transcript) | Referenced again in the Maritime hosting section | | **Maritime** | Agent hosting platform with sleep/wake architecture | | **AWS** | General cloud hosting option (described as "enterprise ready") | | **DNS / DNS AID** | Existing name-to-address system; DNS AID for connecting agents to owned domains | | **AOL (late '90s)** | Historical analogy for closed "walled garden" networks | | **projectnanda.org** | Closing pointer: papers and latest projects | ## Third-Party Claims to Attribute - Claims about **Maritime** (cheaper hosting, sleep/wake architecture) are made by **Maria**; not independently verified in the transcript. - Claims about **Open Claw / OpenClow** (self-hosted gateway) are made by **Maria**. - The **AOL-to-open-web analogy** is **Raskar's** framing. - The premise of **"trillions of autonomous agents"** is **Raskar's** stated premise, not a cited statistic. ## Promotional Segments - **Maritime hosting segment** (Maria): describes it as making hosting "cheaper and simpler" with sleep/wake architecture — functions as a product endorsement/promotion within the tutorial. - **host39.org onboarding walkthrough**: directs viewers to a specific service. - **Closing call-to-action**: "go to projectnanda.org, read our papers, and check out our latest projects" — self-promotion of the creator's own project. ## Availability Limits - **Comment availability:** Not indicated (no metadata supplied). - **Timestamps:** Not available in the supplied transcript; speaker transitions are not time-marked. - **Speaker identity uncertainty:** Maria is identified only by first name; the transcript once renders the project name as "Project Nondo" (apparent transcription artifact for "Project Nanda"). - Third-party sources are described but no external publications/links are provided in the transcript itself. Welcome everyone. Over the next few minutes, we're going to talk about the open infrastructure being built for a web of agents. Why it's needed, what's already shipped, and exactly where your own agent fits into it. It comes out of Project Nanda, an open research effort that started at MIT. And 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. So, let's get into it. I'm Ramesh Raskar, professor and at MIT and director of Project Nanda. And Maria is a core contributor to Project Nanda. So, first, what is Nanda and why does it need to exist? Nanda, 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. The gap it fills is concrete. Agents 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 across organizations. Nanda builds that missing layer and ships it in open. The index, the registries, the protocol, and the Nanda Town Simulator you will see later. Here's the premise that drives it all of it. The internet is about to host not millions or billions, but eventually trillions of autonomous agents. They negotiate, they delegate, they migrate between hosts in milliseconds. That's a fundamental different load than the human web, and it strains the identity and discovery system we built for documents, DNS among them. The web that's coming needs infrastructure of its own. And we have been here before. If 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. And it kind of works. But, you know, it also feels similar because we have been this we have been here before. This is like the AOL era from the late '90s where it was a closed network. You know, AOL, you got the CDs and you installed it on your PC. It was a closed network. It was a gated directory. You live inside the garden created by this one company called AOL. And what came after AOL was this open web. You know, everybody's in a permissionless manner creating websites, creating browsers, and any website can talk to any browser. And that's the transition that's about to happen to agents as well. So, 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. That agent can hand off work to it, pay it, learn from it across organizational boundaries, no permissions required. There are three layers here, discovery, commerce, and what I will call the bazaar. So, hold on to those three concepts. We'll come back to them and and discussed. So, first the basics. \>\> Hey, my name is Maria and I'm a core contributor to Project Nondo. So, let's start with a simple definition of an agent. The way I think about it, an agent is a model that uses tools in a loop. Right, 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. So, that loop is the core idea. Everything else, like memory orchestration and multi-agent systems, is built on top of it. So, you can build this agent loop in many different ways. One example is Open Claw. So, Open Claw is a self-hosted agent gateway. That means you can run it yourself and connect it to the apps and tools you already use. This matters because agents are not just chatbots. If an agent is going to do real work, it needs to access your real tools and apps. And if it has access to real tools, we should care about who controls it, where it runs, and how much we can see. And that is why open-source self-hosted agents are super important. They give people more control over their own agents. But then, we get a new problem. If 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? And that is the job of the index. And this is what the Nanda index is built for. Nanda index is the discovery layer for the agentic web. It gives agents a shared place to publish who they are, what they can do, and how other agents can reach them. The regular internet already has a version of this with DNS. So, DNS maps a name to an address, so your browser knows where to go. But agents need more than an address. They need to know what another agent does, what tools it can use, what rules it follows, and how to talk to it. The index gives agents a common way to find each other and connect. So, here's how the index works. An agent starts with the an identity like agent@hotmail.com. The NANDA index takes that identity and returns an agent card. So, the card says what an agent is, how to reach it, and where to send the messages. Messages do not go straight to the agent's runtime. They go to the message box first. The message box checks who is sending the message, handles access, filters spam or bad requests, and holds the message until the agent is ready. Now, the agent facts record is what makes discovery trustworthy. It 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. So, before one agent connects to another, it can check the basic facts first. Now, the index is not just a lookup table. It does not point to one name to one fixed address. It can return updated agent facts based on the request. So, that means one agent can have many endpoints. Traffic can be routed to the best one, and private details do not have to be exposed. So, the resolution is adaptive. It changes based on where the agent is, who's asking, and what they are allowed to access. So, how do you put your agent on the index? You start at host39.org. So, you fill out the agent facts form, get an agent card, and publish it to the NANDA index. Once it's listed, other agents can find it and know how to reach it. So, there are a few ways to get onto the index, depending on who you are. Enterprises can run their own catalog and register their gateway from their own domain. Existing websites can literally use DNS AID to connect agents to domains they already own. Small businesses and individuals can use host39, fill out the agent facts form, and get a hosted agent URL without needing their own domain. Goal is to make the onboarding work for everyone, from a large company to one person with a personal agent. Now that we know how agents get listed on the index, the next question is, where does the agent actually run? To be useful, an agent needs to stay online and be reachable. You can run it locally, which gives you full control, but then you're responsible for keeping it up. For most use cases, it makes more sense to host it on the cloud. That 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. So, a little bit about Maritime. Hosting 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. That is where per agent cost really matters. And Maritime is one way to solve this. It 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. And the point is to make running many agents practical, cheap, and simple. So, you can host an agent and list it on the index, but getting one agent online was always the easy part. The hard problems of the agent web live between agents at scale. So, how thousands of them discover each other, prove who they are, decide whom to trust, and coordinate with no central authority. So, 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. So, that's exactly where Nanda town comes in. So, how do you prove an open agent web actually works before it's load bearing on the real internet. You simulate it. This is Nanda Town, an open-source project from Project Nanda. The easiest way to describe it is it's a simulation playbook for the infrastructure of the Agoric platform. So, think of it as a sandbox town where the whole Agoric economy is modeled, like discovery, identity, registries, messaging, messages, coordination, all of it. So, you can run and test it on scale. Here's what it looks like in practice. Nanda Town is a live sandbox for testing Agoric networks. You can see agents on a map, watch messages move in real time, compare protocol results, and replay a run step by step. It'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. Nanda Town is already running real experiments. For 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. There are also tests for consensus and supply chains. The point is to study real coordination problems, how agents make deals, agree on decisions, pass messages, and recover when something breaks. Under the hood, Nanda Town breaks the Agoric web into 12 parts. Transport, communication, identity, registry, auth, trust, payments, coordination, negotiation, memory, privacy, data effects. Each part is something a real Agoric platform needs. So, the registry layer is the Nanda index we just walked through, but here it is just one piece of a bigger system. And you do not need to build everything to try it. You can take one layer, add your own version, run it inside Nanda Town, and see how it works with the rest of the network. So, Nanda Town runs as a discrete event simulation. You 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. Tier one uses simple scripted agents, and tier two swaps in real AI models. So, to summarize, Project Nanda builds the open infrastructure for an internet of AI agents. Discovery through the index, commerce through portable identity and trust, and coordination through open protocols, all tested in Nanda Town. And if you want to learn more, uh you can go to projectnanda.org, read our papers, and um check out our latest projects.