Talk by Sunil Pai (Cloudflare) on "Code Mode," an approach where AI agents generate executable code (e.g., JavaScript) instead of using traditional JSON-based tool calling, enabling more efficient, stateful, and complex interactions with systems. Covers the limitations of tool calling at scale, why code generation is more efficient, a live demonstration, a proposed "Harness" software architecture, sandboxed execution with observability and security, long-running workflows, generative UI, and the resurgence of capability-based security. # Video Identity Observations: "Code Mode: Let the Code do the Talking — Sunil Pai, Cloudflare" ## Presenter Self-Description - **Speaker identity and employer**: Sunil Pai explicitly states, > "My name is Sunil Pai. Uh I work at Cloudflare. Uh I build agents over there uh for the agents SDK." - **Prior work**: Introduced by a host as creator of PartyKit: > "Our next presenter created PartyKit, the open-source tool for real-time multiplayer apps." - **Talk positioning**: He repeatedly frames the talk as exploratory, not vendor marketing: > "I'm trying very hard for this not to be a Cloudflare talk, but I think we are on the sponsor board" and > "this talk is a little weird, it's a little woo-woo. I'm not even sure where we're going." - **Claim of non-exclusivity**: > "We don't claim to have invented it, but this is a talk about the implications of something new that we're discovering." ## Intended Audience - **Developers who build AI applications**: Direct address: > "you guys have built uh AI applications, and tool calling gets weird at scale." - **Practicing programmers, especially UI/frontend**: > "everyone here is a programmer" and > "I'm a React programmer and I don't want to freak out the React people." - **Conference attendees at a developer-focused event**: References to a stage, ~800 audience members, and other talks ("my colleague Matt Carey, who's actually going to be speaking about this a little more tomorrow, you should watch his talk"). ## Claimed Scope of Topics - **Core concept**: "Code mode" — generating executable code instead of JSON tool-calling: > "Instead of doing this JSON back and forth thing, we asked the model to generate code, usually JavaScript, that we could run against an environment." - **Concrete technical claims**: Cloudflare's API surface (~2,600 endpoints) reduced from > "about 1.2 million tokens in your first call" to > "1,000 tokens" via a two-tool search/execute pattern (attributed to colleague Matt Carey). - **Architecture themes**: capability-based security, sandboxed execution with V8 isolates, observability, long-running workflows, and generative UI. - **Language-agnostic framing**: > "This isn't even a JavaScript talk. It can be in Python, it can be in WASM." ## Observable Presentation Patterns - **Live-demonstration style**: Includes a live demo with acknowledged risk: > "demos don't usually do me well on stage" — and the demo partially falters (pagination errors), which he narrates candidly. - **Humor and self-deprecation**: Jokes about Mac Mini prices, his mother's photo-management options, and Kenton letting the model lose at tic-tac-toe. - **Collegial attribution**: Credits coworkers by name (Matt Carey, Kenton Varda, "the creator of Cloudflare Workers") for the ideas presented, while joking > "he does the work, and I like taking credit for his work." - **Open-ended, idea-seeding tone**: Invites collaboration rather than prescribing: > "I want to like spread the idea to you and have you folks like integrated" and > "come talk to me about it at the pub." ## Notes on Evidence Limits - No curated metadata was supplied; all observations derive from title and transcript. - Statements about internal Cloudflare tools (dynamic workers, sandbox SDK, agents SDK) are **creator claims** about his employer's products, not independently demonstrated beyond the partial live demo. - No channel-wide identity claims (e.g., "this channel is about X") appear in the material; the evidence describes this single talk only. # Delivery Observations: Sunil Pai — "Code Mode: Let the Code do the Talking" ## Voice & Tone - **Casual, humorous conference style.** The talk opens with an immediate joke: *"Uh 20 minutes to the pub. Uh hi."* Humor recurs throughout, often self-deprecating, especially about the live demo failing: *"Love that this is happening to me on stage, because I did test it 10 times before coming on."* - **Heavy use of filler words.** "Uh" and "um" appear pervasively across the transcript (e.g., *"Uh I work at Cloudflare. Uh I build agents over there uh for the agents SDK"*), giving a conversational, unrehearsed feel. - **Self-aware asides.** He frequently flags his own digressions or risks: *"I'm trying very hard for this not to be a Cloudflare talk"*, later *"This is an anti-Cloudflare talk at this point."* ## Recurring Phrases & Verbal Habits - **"Uh the point being"** — used repeatedly to re-anchor after tangents: *"the point being that we were able to take a wide, super wide API surface and make it incredibly fast"*, *"Um the point being, uh we are able to shrink that down."* - **"By the way"** as a pivot device: *"By the way, you can actually see it"*, *"Uh by the way, it lost the game"*, *"This is something a bunch of companies are already doing really well, by the way."* - **"You know?"** and **"like"** as connective tissue: *"they don't hang out in the pub, they hang out in registries. They dream in types and syntax errors, you know?"* - **"To be clear"** for emphasis and disambiguation: *"To be clear, there is no tic-tac-toe code anywhere in this system."*, *"To be clear, your customers are still humans."* - **"What if"** framing for the speculative section: *"What if you could uh generate long-running workflows that run for days, months, years? What if each of those instances has some state..."* ## Teaching Habits & Structure - **Concrete numbers to anchor claims.** He quantifies the core benefit vividly: *"about 2,600 API endpoints... about 1.2 million tokens in your first call... down to 1,000 tokens... I think it's like 99.9% reduction."* - **Relatable analogies before abstractions.** The 200-photos example grounds the argument: *"you have 200 photos on your desktop. I need you to categorize and rename them. First thing you do is... you're going to open up an IDE."* - **Live demo as a teaching device** — deliberately interactive and openly risk-acknowledged: *"This is going to be scary. I actually have I have a live demo of this, and uh demos don't usually do me well on stage."* - **Direct audience address and callouts.** He references the room: *"everyone here is a programmer"*, and jokes demographically: *"there's a ghost in the shell reference here for anyone who's over the age of 40, you need ibuprofen."* - **Attribution of others' work.** He explicitly credits colleagues: *"my colleague Matt Carey, who's actually going to be speaking about this a little more tomorrow, you should watch his talk"* and the Kenton anecdote — while joking about taking credit: *"he does the work, and I like taking credit for his work."* ## Rhetorical & Argumentation Patterns - **Setup of a dichotomy then its collapse.** Programmers vs. his mother's options; then: *"LLMs are breaking this boundary."* Same move at the end: *"programmers... got code... everyone else got buttons and forms. That distinction is breaking."* - **Coining/sharing terminology.** He introduces vocabulary explicitly: *"The the phrase we've started using is it stopped generating a program and it instead inhabiting the state machine."* Also flags ambiguity: *"we're calling it a sandbox, which is again another completely overloaded term."* - **Speculation flagged as such.** He is candid about uncertainty in the ideas themselves: *"it's a little woo-woo. I'm not even sure where we're going and I want to like spread the idea to you."* This is an observable hedge about the talk's thesis, not a delivery flaw. - **Imperative takeaway lines.** Memorable closing directives: *"you need to let the code do the talking"*, *"you need to be thinking about what is the developer experience for these agents."* ## Isolated Moments (Not Established Patterns) - The password joke during the demo (*"if I'm not, I'll need all of you to close your eyes while I enter a password"*) is a one-off bit. - The *"llama poop"* trade example and the Mac Mini pricing rant are single digressions, not recurring themes. - Audience laughter and a snort are noted in the transcript, but only once each, so audience-response frequency can't be established. ## Uncertainty Notes - Timing/pace observations (e.g., whether filler-word density is consistent) can't be verified beyond transcript text; no audio or metadata was provided. - Some apparent disfluencies may be transcription artifacts (e.g., *"it instead inhabiting the state machine"*, *"parallelize parallelization"*), so quotes are given exactly as transcribed rather than interpreted. # Position Record: "Code Mode: Let the Code do the Talking" — Sunil Pai, Cloudflare ## Speaker **Sunil Pai** (introduced by an unnamed MC), Cloudflare, building agents for the agents SDK; creator of PartyKit. ## Core Thesis - Instead of JSON tool-calling between models and APIs, have the **model generate code** (usually JavaScript) that runs against a target environment in **one execution**. - Pai explicitly frames this as an emerging, not fully understood shift: *this talk is "a little weird, it's a little woo-woo. I'm not even sure where we're going."* ## Problem: Tool Calling at Scale - A few tools with short runs work fine, but hundreds of tools (Google services, Jira, wikis) fill context and **break down**; composition is awkward and the back-and-forth is slow. ## Claimed Benefits of Code Generation - Typed API, type checking, syntax errors as feedback - Models already trained on vast amounts of code - Single execution replaces multiple round trips - Native capabilities: **looping, state, sequencing, parallelization** ## Case Study: Cloudflare API "Search + Execute" - Attributed to colleague **Matt Carey** (speaking the next day); Cloudflare API surface is ~2,600 endpoints - Claim: exposing a tool per endpoint ≈ **1.2 million tokens** in the first call — "no way to create an MCP server for the entire Cloudflare API surface" - Solution: two tool calls, **search** and **execute**, both accepting code as a string; search receives the OpenAPI JSON spec - Claimed reduction to **~1,000 tokens** ("99.9% reduction") - DDoS example: an eight-round-trip MCP workflow becomes one generated code run "right next to the API surface" - **Live demo partially failed** — the pagination logic hit JavaScript errors on stage. Pai acknowledges: *"demos don't usually do me well on stage"* and jokes the listing may be working but not rendering. This failure remains part of the record; the token-reduction claims are asserted, not demonstrated. ## Kenton Varda Tic-Tac-Toe Example - Kenton (creator of Cloudflare Workers) asked a model to build a canvas app, then drew a tic-tac-toe board and asked the model to play - The model began generating a tic-tac-toe app; Kenton stopped it and instructed it to **inspect the system state** (an array of strokes) - The model recognized the board and played via strokes — **no tic-tac-toe code exists in the system** - Pai's framing: *"it stopped generating a program and it instead [is] inhabiting the state machine"* - Side note (unverified, reported from reasoning traces): the model **lost intentionally**, letting Kenton win — Pai flags this as "a whole other weird area of alignment" ## Proposed Architecture: The Harness - Everyone is building "harnesses" because coding agents are **general-purpose computing machines** - Side complaint: people running agents on Mac Minis is "the wrong machine" — a jab at the trend, with self-aware humor about buying one himself - Sandbox attributes (contrasted with containers): - Starts with **zero capabilities**; only executes code - Capabilities granted explicitly as APIs - Outgoing fetches controllable; **default recommendation: no outgoing fetches, only APIs** - Must be fast and have **absolute full observability** (e.g., auditing why the system made a $2.3M trade) - Implementation-agnostic: V8 isolates (Cloudflare's dynamic workers, ~10 years security hardening), WASM, or a custom JS interpreter — "that's not the main story" - Pai repeatedly disclaims Cloudflare ownership: *"someone else build something better if you think it's better"* ## Ambitious Extensions (Speculative) - **Long-running workflows** running days/months/years, each carrying state - **Generative UI**: fully custom interfaces per user instead of lowest-common-denominator design; context (cart, orders, frustrations) surfaced as **actions** rather than a blank chat box — though Pai concedes "blank chat box e-commerce might be a lot of fun" - Running the harness **closer to the user** (e.g., on iPhone) to mash up services — Pai jokes this makes it "an anti-Cloudflare talk" ## Broader Social Claim - Dichotomy historically acceptable: programmers write scripts; non-programmers (his mother) get $7/month apps with dubious daemons - **LLMs break this boundary**: every person has "a buddy that can spit out code" - Safe execution ("I say execute it safely here") is the open caveat he explicitly defers ## Agents as Users - "Your next billion users are these little robots" — but customers remain humans - Agents "hang out in registries," "dream in types and syntax errors" — prioritize **agent developer experience**: markdown docs, actionable errors, search discoverability - **Capability-based security** is the key idea to embed - Prediction: UI programmers will do well because they've been close to users ## Closing Claim - Programmers had "infinite power"; everyone else got buttons and forms. **"That distinction is breaking... let the code do the talking."** - Pai positions this as "a whole new area of research" and invites discussion. ## Qualifications & Tensions Preserved - Demo failure on stage vs. claimed reliability ("I did test it 10 times") - "I'm trying very hard for this not to be a Cloudflare talk" while demonstrating Cloudflare products - Confidence in the architecture vs. admitted uncertainty about where it leads ("woo-woo") - The model deliberately losing (alignment oddity) is reported but unresolved - All numeric claims (2,600 endpoints, 1.2M → 1,000 tokens, 99.9%) are speaker assertions from the talk # Speaker Attribution & Provenance Record ## Video **Title:** Code Mode: Let the Code do the Talking - Sunil Pai, Cloudflare **Video ID:** `01a15ca9-b242-7802-80cf-3ebdc1410c8d` ## Speakers & Roles | Speaker | Role | Evidence | |---|---|---| | **Sunil Pai** | **Main presenter / guest speaker** | Introduced by MC; introduced himself: creator of **PartyKit** (open-source real-time multiplayer tool), builds AI agents for the agents SDK at **Cloudflare** (day job) | | **Unnamed MC** | **Introducer** | Opens with: > "Our next presenter created PartyKit... Please join me in welcoming to the stage Sunil Pai." | | **Audience** | Unnamed (laughter, snorts, applause) | Bracketed crowd reactions only | ## Third-Party Individuals Referenced (Not Speakers) - **Matt Carey** — Sunil's colleague at Cloudflare; credited with the search/execute two-tool design for the Cloudflare API surface; speaking "a little more tomorrow" at the same event. - **Kenton** — described as creator of Cloudflare Workers; source of the tic-tac-toe / white-coding-environment anecdote. Sunil notes: > "he does the work, and I like taking credit for his work." ## Referenced Sources & Claims - **Cloudflare API surface**: ~2,600 API endpoints; ~1.2 million tokens if exposed as individual tools; claimed reduction to ~1,000 tokens via two code-accepting tools (search/execute) — per Matt Carey's design. - **Kenton's anecdote**: model (referred to as Opus in reasoning traces) played tic-tac-toe via canvas stroke state; claimed the model let Kenton win (per reasoning traces). - **Cloudflare products mentioned**: dynamic workers, V8 isolates, sandbox SDK (containers/VMs). - **Prior art acknowledgment**: Sunil states > "We don't claim to have invented it." ## Promotional Segments (Flagged) - **Sponsor mention**: > "I think we are on the sponsor board, so that's nice." - **Cloudflare product references** throughout (dynamic workers, sandbox SDK, agents SDK), framed by Sunil as unintended: > "I'm trying very hard for this not to be a Cloudflare talk." - **Cross-promotion**: recommendation to attend Matt Carey's talk the following day. - **Self-deprecating anti-promotion**: > "This is an anti-Cloudflare talk at this point... someone else build something better if you think it's better." ## Uncertainty & Availability Limits - **Live demo partially failed**: the on-stage worker-listing demo encountered rendering/pagination errors; speaker requested the audience assume success. - **No curated metadata** was supplied; all provenance facts derive from the transcript alone. - Comment availability: **not stated** in source material. \>\> Our next presenter created PartyKit, the open-source tool for real-time multiplayer apps. For his day job, he builds AI agents at Cloudflare. Please join me in welcoming to the stage Sunil Pai. Uh 20 minutes to the pub. Uh hi. Uh my name is Sunil Pai. Uh I work at Cloudflare. Uh I build agents over there uh for the agents SDK. I'm trying very hard for this not to be a Cloudflare talk, but I think we are on the sponsor board, so that's nice. Uh this is a talk about something we call code mode. Uh I've been wearing the hat uh and uh there's some prior art to it. We don't claim to have invented it, but this is a talk about the implications of something new that we we're discovering. So, um you guys have built uh AI applications, and tool calling gets weird at scale. When it's just a couple of tools and very short runs, it's fine. But the moment you start stuffing in uh your Google services, your Jira, your wiki, etc., and you have like hundreds hundreds of tools filling up the context, it starts breaking. Um and uh the composition is weird, and there's this back and forth that you have to do with uh the model that's really slow. Uh we decided to take a different act. Instead of doing this JSON back and forth thing, we asked the model to generate code, usually JavaScript, that we could run against an environment. Uh and some of the benefits seem a little obvious to us. Uh with code, you get a typed API, you can do type checking, there are syntax errors. Uh models are trained on gigabytes, if not terabytes, of data already in the training set. Uh and instead of doing this back and forth, you could write code that executes it all in one run, just one execution. So, uh so this is what I mean. Like there are uh fundamental capabilities of code. You're able to do looping, you're able to hold state, uh you're doing sequencing, parallelize parallelization, things that you would normally do with code anyway as an engineer. So, the first place we applied this uh my colleague Matt Carey, who's actually going to be speaking about this a little more tomorrow, you should watch his talk. Uh the Cloudflare API surface is about 2,600 API endpoints. If we exposed a tool for every single one of them, it's about 1.2 million tokens in your first call. Like it just blows. There's no way to create an MCP server for the entire Cloudflare API surface. And he had a very clever idea where he exposes just two tool calls. Uh search and execute. Both of these endpoints accept code as an input, literally a string of code. For search, the input to the function that you pass to it is the entire open API uh JSON spec. And once it does that, execute gives gives you a whole bunch of functions that you can call against the things that you called. And it reduced that 1.2, 1.5 million token thing down to 1,000 tokens. Kind of unheard of. I think it's like 99.9% reduction. Uh this is going to be scary. I actually have I have a live demo of this, and uh demos don't usually do me well on stage, but uh but the point being that we were able to take a wide, super wide API surface and make it incredibly fast. Uh the prompt itself can be uh fairly generic. So, I should have kicked up the font size on this one. The prompt here is as a customer, you come in and say, "We are getting DDoS'd. I want you to find every offending IP that's like attacking us and block them." In a moment of panic, when your website is going down, you don't have the time to do menu diving. Uh the Cloudflare dashboard is famously a little cumbersome to handle. Uh and you just want the thing done. And you can't even get an AE, it's like 3:00 in the morning. Uh with a regular MCP thing, and this isn't even talking about stuffing 1.2 million tokens. It would be about eight round trips to do each of those API calls. Instead, the model can generate this string of code, run it immediately right next to the API surface, and do it in one shot. And it's just running JavaScript. Just functions and um just things that you're exposing on the API surface. Okay, live demo. This is a demo of our mythical server. Uh I hope I'm logged in, because if I'm not, I'll need all of you to close your eyes while I enter a password. Let's say I just want to like list my workers. Oh, there it is. List my workers. I say send. Okay, okay, then there's no password required. Okay, fine, that's fine. Okay, I give it only read-only access for this demo. Uh do the thing. Yes, allow, sure, whatever. Ba ba ba ba ba. Nice. Okay, it comes back, and uh you'll see it'll start executing tool calls. I should be able to open this up. It has sent saying, "Hey, find me all API endpoints that just say the words list workers or something like that." Uh it then runs code, uh which Hey, yeah, it's like one single request for the API endpoint to get all the workers. Uh it must have received a whole bunch of these. It's actually going through JavaScript errors now. This is going to be fun to see if it actually succeeds. Yikes. Oh, is it trying to do it like per page? It's trying to paginate through the thing. Assume that this worked anyway, and I'll keep talking while it does this. Uh Love that this is happening to me on stage, because I did test it 10 times before coming on. Uh I need to pay for the Mythos uh model to make this work accurately. \>\> \[laughter\] \>\> Uh By the way, you can actually see it. It is actually like listing workers over here. It might just be having trouble uh rendering it over here. Um the point being, uh we are able to shrink that down. Now, if this was a talk about optimizing MCP servers, I would be done and dusted. I was like, "Hey, you should throw this, and trust me, it works when you're not staring at it and have 800 people looking at you on the stage." But it did give us an idea that there's something deeper going on here. The ability to like run this code and uh feels like there's a new way of interacting with systems, with LLMs. Uh here's what I think. Like everyone here is a programmer. And I give you a problem statement like you have 200 photos on your desktop. I need you to categorize and rename them. First thing you do is you look you're going to open up an IDE. You're going to write a little script. Maybe you're going to pass every image to a vision model now, because you get a nice caption for it. Uh rename it, and you're done and dusted. That is how you interact with systems. Uh my mother's not going to do this. Her options are to, well, call me up, or just that. There's going to be like lowest common denominator apps for photo management, and it's $7 a month. And for some reason, you have to install a daemon, which is stealing your crypto or some such stuff. Uh and there's been this dichotomy, and it's fine. Like until now, this has been an acceptable uh this has been an acceptable trade-off that non-technical people will have custom-made interfaces built for their needs and desires. LLMs are breaking this boundary. They every human being on the planet now has access to a buddy that can spit out code that can interact with systems. Uh it takes it takes a line like rename these files by date and location, and generates code, and can run it on your uh on whatever system you expose to it. Uh I say execute it safely here, and that's the bit that I do want to talk about in a minute. The other example I have, so this is Kenton. Kenton is the creator of Cloudflare Workers. Uh famously, I'm So, he does the work, and I like taking credit for his work. This is our relationship in the company. Uh so, he he had a thread a little while ago where he built he's built a little white coding environment for himself, because no one else does that in the world right now. So unique. Build your own little white coding thing. Uh the the thing he asked it to generate was a canvas, one of these TLDraw, Excalidraw style canvases. Uh and it did it it did a little canvas with little brushes and colors. And the first thing Kenton did was draw a tic-tac-toe board on it with a little X in the corner. This is the finished state, and I'll get to that in a second. He did that. And uh what he told the model then is, "I want you to play tic-tac-toe with me. The model, as you can guess, it started generating a tic-tac-toe app. Okay? Kenton stopped it immediately. He's like, "No. You have access to the entire state of the system. And the state of the system here is an array of strokes. You know, like just a whole bunch of points, grid line, grid line, X stroke, etc." He said, "Inspect that and play it with me." Uh immediately, the model started it output the state into its own context and it's like, "I recognize what this looks like. It looks like a tic-tac-toe board and I can see that you put an X in the top left. Let me draw a perfect circle in the middle of the app." To be clear, there is no tic-tac-toe code anywhere in this system. The the emergent behavior is that the model has like, "Sure, I now know how to interact with the system with a set of strokes." Uh also, it lost. Uh by the way, it lost the game and then when we saw the reasoning traces, we noticed that Opus let Kenton win. Which is a whole other weird area of alignment we're not talking about. Anyway, so this actually generated a lot of conversation internally and that's why like this talk is a little weird, it's a little woo-woo. I'm not even sure where we're going and I want to like spread the idea to you and have you folks like integrated. So, the the phrase we've started using is it stopped generating a program and it instead inhabiting the state machine. Uh there's a ghost in the shell reference here for anyone who's over the age of 40, you need ibuprofen, uh you should go back home. Uh but no, like it was a very strange thing to for us not to have a separate app generation stage that you then like interact with. That is entirely the part of the thing. So, what does this new software architecture look like? Uh everyone's building what they call a harness. Uh it's because over the last 3 to 6 months, everyone has realized that these coding agents are great general-purpose computing machines. It's why they're running Claude code or code No, they're running Pi on a Mac Mini, which is the wrong machine for this, by the way. You don't have to spend $400 for a thing that makes API calls. Uh it's been driving me mad. If you check all the second-hand prices of Mac Minis have like shot up. I got one before it, but I got it because I'm special that way. Uh you be everyone's building this harness and this architecture of the harness is not just that it can generate code, but it has a safe space to execute this code into which capabilities are uh exposed. Uh and there are some attributes to this sandbox. We're calling it a sandbox, which is again another completely overloaded term and I have friends in the industry, everyone's building a different kind of sandbox. Uh we have a sandbox SDK, which uses containers and VMs, but that's not even what I'm talking about right now. Uh there are some capabilities to it. Unlike a container, which comes with all sorts of features that you surround with security, you know, you do a bunch of things from the outside, you start with something that has no capabilities. The only thing it can do is execute code. It can't do fetches, there's no exposed APIs, no nothing. And then you grant capabilities to it explicitly. Uh we have something called dynamic workers. I told you I'm it's not really a Cloudflare code, someone else build something better if you think it's better, it's fine. Uh but this is what we use. We use V8 isolates because they start up really, really quickly and uh it's about 10 years of security hardening. Uh it's in our DNA, we care we care a lot about that. Anyway, so we you start exposing capabilities as APIs, A. And we also can control all outgoing fetches and any network connections. In fact, the default way we recommend you use this is no outgoing fetches, only APIs. It has to be fast and you need absolute full observability into it. You need to know why last Tuesday it made a trade for $2.3 million for I don't know, man, like llama poop or something, right? You need to go back to that code. Absolute observability on these systems. It can be V8 isolates like we use. Uh you could use, I don't know, a web web assembly, a custom JavaScript interpreter. Uh that's not the main story here. You just want something that's able to execute that you're able to expose capabilities to and run really quickly. From here, you can start getting really ambitious. The example that I showed you was a one-off, take some code, run it on an API, expand. Now, what if you could uh generate long-running workflows that run for days, months, years? Uh what if each of those instances has some state that it can carry with it uh through um through its lifetime? What if in this world of generative UI, you can start generating a perfect perfectly custom UIs for every single user that you have. Everyone who does e-commerce knows this problem. The more popular you get, the more UI becomes this bland thing that has to work for every single user. And then you bring in the ML people and they're like, "Oh, what if we change the color button this way if it's somebody else?" No. You can go absolutely custom. So, uh I I like the fact that I got Opus to generate generative UI for a slide where I'm making a point about generative UI and it still looks a little bit like Uh but the idea is everyone like e-com let me talk about that e-commerce. Like you have context about everything about the user, the things they like, the orders they have in their cart, the things that might be making them mad. You can surface these things as actions. The UI doesn't have to be a blank chat box. Though, honestly, blank chat box e-commerce might be a lot of fun. Uh here I have two different use cases. In the first one, it's uh I need to return these shoes and find something similar under $100. If the product engineers have not implemented this, how it's it's going to kind of suck in but you can generate something on the fly versus what is happening with my uh delayed order. Point being, we are now in a world where we can generate completely different programs backed by a system that you built on your back end for every single user. It's a new kind of software we're building. And this harness idea isn't just built into the product. A lot of people are finding power by running the harness closer to the user simply because then they get to start mashing up all their different services. This is an anti-Cloudflare talk at this point. I'm like, you should be running the software on your iPhone, like not so much on our servers. Please run it on our servers. Uh but you but there you start getting to stitch together different systems in this safe environment and you get to do it on a task-by-task basis. Um I put this in here because I'm a React programmer and I don't want to freak out the React people by saying no one really wants to build UI anymore. But really, it's a harkening back to rethinking everything that we have thought about UI and for this new age. I keep thinking about it as part of the tech tree we have not really explored for 30 years because eval wasn't around, but now we have a safe eval and we have these things that generate code for you. But you do need to be in a place where you understand that your next billion users are these little robots that are generating code for you. To be clear, your customers are still humans. The things interacting with your systems, uh if you really love your users, you need to find out where they hang out and they don't hang out in the pub, they hang out in registries. They dream in types and syntax errors, you know? \>\> \[snorts\] \>\> Uh you need to be thinking about what is the developer experience for these agents. This is something a bunch of companies are already doing really well, by the way, you know? Docs which are markdown, uh errors that let the agent know what to do next, uh discoverability via search. The big one that I do want to talk that I want you to embed in your head, I guess, is this idea of capability-based security. This isn't even a JavaScript talk. It can be in Python, it can be in WASM. Uh I hope it brings a resurgence of Lisp. It's how I kind of learned how like ASTs work, it kind of breaks your brain. Uh but the but the attributes are still very much the same. Events, sandboxing, capability-based security, embeddable so that it's really fast to start up and run ephemerally. Uh React programmers simply be Well, UI programmers simply because they have so much uh they've been so close to users, I suspect that they'll do particularly well here and that feels really good to me, by the way. I feel happy about it. So, to end, for the longest time, programmers like us, we got code, we had infinite power to interact with any system that we could and complain about it on Twitter because our documentation doesn't have the right CSS or something. JavaScript programmers super entitled, by the way. Uh everyone else got buttons and forms. That distinction is breaking. In a world like this, you need to let the code do the talking. The code is the thing that interacts with all your systems. Uh come talk to me about it at the pub. Like this is like it feels like it's opening up a whole new area of research for us. Uh and we have a lot of ideas and I get to finish my talk and the day with 6 seconds left. How good is that? Thank you very much. Appreciate it. \>\> \[applause and cheering\]