# The New Application Layer

Malte Ubl

AI Engineer Europe 2026

[Talk page](https://ai.engineer/talks/XKup1pj-34M-new-application-layer) · [Video](https://www.youtube.com/watch?v=XKup1pj-34M)

## The New Application Layer

Agents make more software worth building, change who uses it, and demand infrastructure that separates orchestration from execution.

### Connect an agent, then give it somewhere to work

[1:02](https://www.youtube.com/watch?v=XKup1pj-34M&t=62s)

An agent needs a way to reach people and an environment in which to work. Malte Ubl opens with two projects from his own development stack that address those needs. [Chat SDK](https://vercel.com/blog/chat-sdk-brings-agents-to-your-users) connects agents to messaging applications such as Slack, Telegram, and WhatsApp. [just-bash](https://github.com/vercel-labs/just-bash/blob/main/packages/just-bash/README.md) is a Bash interpreter written in TypeScript: a way to offer agents a sandbox-like working environment through a tool they already favor. These are small, concrete pieces of a larger shift—software increasingly built by agents, used by agents, and implemented as agents.

![The speaker beside a black slide displaying the Chat SDK and JUST BASH logos.](https://ai.engineer/api/talk-essay-media/c9b754c3bc6d432015959a4888a9041cd5c4bd93020d3576ea5988398bd75703)

Chat SDK and JUST BASH introduce the opening tool stack.

### A profession changing on both sides

[1:37](https://www.youtube.com/watch?v=XKup1pj-34M&t=97s)

The community around that work has a history. Ubl helped run JSConf EU in Berlin, ending its ten-year run in summer 2019. The timing proved fortunate: COVID would have prevented the next edition anyway. But the pandemic also interrupted the hoped-for handover to another generation of organizers. More than half a decade later, the gathering at [AI Engineer Europe](https://ai.engineer/europe/2026) represents a return, organized around a different discipline. Ubl sees AI engineering as web development’s successor as a mainstream field shaping the next decade. **The disruption affects both how engineers build—with AI—and what they build—AI applications and agents.**

That raises an immediate career question: is there still a place for engineers, especially the next generation? Ubl’s answer starts with a deliberately self-deprecating comparison. The introduction’s 25 years of engineering experience is, he says, a substantial understatement; none of that experience would make him good at creating TikTok videos. People who grow up inside a medium develop instincts that experienced outsiders may lack. Junior engineers formed in an AI-first working environment may similarly become better at the new discipline. His optimism for everyone else rests on a second idea: agents are a new kind of software, expanding the work that software can economically do.

![Black-and-white image of an arm wearing a large rectangular screen showing a man's face and social-video controls.](https://ai.engineer/api/talk-essay-media/628ddeb39dc52281f3e3e361d081d1a3f6f6072b6646863c73789b2ad8e536a8)

A wrist-mounted screen illustrates the social-video example.

### The software that was too expensive to write

[4:09](https://www.youtube.com/watch?v=XKup1pj-34M&t=249s)

Imagine a circle containing all the software that should exist. Traditional development fills only part of it. A useful automation might require someone to discover scattered business knowledge, work through exceptions, and encode the result in a growing collection of conditional statements. The benefit can be real while the implementation remains too expensive to justify. Agents make some of that previously uneconomic work affordable. Ubl pictures them filling the rest of the circle, while allowing that we will discover still more software worth creating. The important change is the expansion of feasible automation.

The same change affects the decision to make or buy software. Ubl observes more companies choosing to build what they previously would have purchased as SaaS. Silicon Valley’s talk of a SaaS apocalypse overstates his own conclusion—he expects SaaS companies to survive—but more companies making more software can mean more engineering work even when each task takes less time.

**Cheaper software production tests how elastic demand for software really is.** If lower costs unlock enough new projects, total engineering demand can grow alongside productivity. Ubl says he sees that growth happening. He acknowledges that the expansion could eventually follow an S-curve and level off, but sees no sign of that limit yet, especially while agent capabilities continue improving. This is an economic thesis about new demand, rather than a claim that existing jobs and tasks remain unchanged.

### Start with work that benefits from continuous service

[6:47](https://www.youtube.com/watch?v=XKup1pj-34M&t=407s)

Building this application layer does not require starting with the most ambitious autonomous agent imaginable. Coding agents work so well that it is tempting to assume their success transfers directly to every other domain. Ubl cautions that it often does not. Meanwhile, simpler opportunities can fit existing operations without the long, failure-prone effort of changing an entire business process. Ubl suggests these opportunities could save companies millions or billions of dollars, without presenting a measured savings case here.

Customer support is an obvious starting point: a business already has support agents and can ask which portions of their work software agents can handle. Sierra and Decagon exemplify the emerging agent-as-a-service category. The more general opportunity is **continuous availability**. Look for a job where moving from nine-to-five coverage to round-the-clock service would change the value delivered. An agent’s ability to keep working while people sleep can matter even before it changes the underlying task.

### Compress research, retain the human decision

[8:33](https://www.youtube.com/watch?v=XKup1pj-34M&t=513s)

A second archetype applies to a common business workflow:

1. A business event arrives.
2. Someone researches the relevant facts.
3. A human makes a decision.

An agent can take over the middle step while leaving the beginning and end intact. Ubl’s illustrative case reduces a task from 30 minutes to five minutes across 100,000 repetitions per year. The attraction is not just speed: the organization can retain its existing decision process. Ubl argues that keeping the human decision also preserves the risk profile. In implementation terms, that decision boundary must remain explicit; producing research should not itself authorize an action.

At Vercel, clicking Contact Sales starts this kind of workflow. Ubl reports hearing that approximately 75% of Contact Sales submissions actually need support and are handed to the support team. For a sales lead, the agent checks LinkedIn, searches for the company, investigates its size, and routes the result to the appropriate person. A human eventually checks that it makes sense. Ubl estimates that this preparation previously took a person perhaps 15 minutes per lead.

Abuse reports have the same structure, with a more consequential decision at the end. The agent investigates whether a website is abusive and what responses might be appropriate; the professional retains the final decision. A TypeScript interface can make that boundary visible by returning a review packet rather than performing enforcement:

```typescript
type AbuseReport = {
  id: string;
  website: string;
  allegation: string;
};

type Findings = {
  evidence: Array<{ source: string; observation: string }>;
  possibleResponses: string[];
};

type ReviewPacket = {
  report: AbuseReport;
  findings: Findings;
  status: "pending-human-decision";
};

async function prepareAbuseReview(
  report: AbuseReport,
  investigate: (report: AbuseReport) => Promise<Findings>,
): Promise<ReviewPacket> {
  const findings = await investigate(report);
  return {
    report,
    findings,
    status: "pending-human-decision",
  };
}
```

The function preserves the original report, adds research, and leaves the outcome pending. It contains no website takedown or other enforcement operation. That separation captures the useful scope of this archetype: automate preparation without silently expanding into decision authority.

### Make existing information usable

[10:29](https://www.youtube.com/watch?v=XKup1pj-34M&t=629s)

A company can possess information without being able to use it in practice. An issue tracker may be stale even though the relevant update appeared in Slack or in a Granola meeting recording. The missing work is collecting that evidence and putting it where people expect to find it. An agent can research those sources, help update the tracker, and prepare the status information a manager would otherwise have to request again.

![Slide with overlapping circles labeled “Compresses the research,” “Always running,” and “Surfaces hidden information,” beside a heading about practical non-coding agents.](https://ai.engineer/api/talk-essay-media/6d005a1ee3e39d2e9ade16b801cc21e67aec207bd2f1ee142c505ac7190b9484)

Agent archetypes overlap: compress research, run continuously, and surface hidden information.

These archetypes can overlap. Continuous operation determines when the work happens; research compression determines which part of a process becomes faster; surfacing information determines which existing evidence becomes useful.

| Archetype | Useful change |
| --- | --- |
| Always running | Work continues outside staffed hours |
| Compress research | Preparation accelerates before a human decision |
| Surface hidden information | Scattered updates reach the place they are needed |

An agent preparing status updates can do all three: watch for new information, collect it, and make it available before someone asks.

### Ask which work people hate

[11:36](https://www.youtube.com/watch?v=XKup1pj-34M&t=696s)

A direct way to find the next useful agent is to ask employees: “What do you hate most about your job?” Vercel’s in-house support agent provides Ubl’s example. Ubl reports a 90% deflection rate for Vercel’s support agent, meaning it helps users in real time without sending them onward. He supplies no reporting window or evaluation method for that figure.

The reported organizational benefit is a sharp improvement in support-team job satisfaction. Routine questions, such as a rejected credit card, no longer dominate the queue. Staff can spend more time investigating interesting cases and helping people who need deeper assistance. Ubl gives no numerical satisfaction measure, but the mechanism is clear: removing repetitive work changes the composition of the human job. Eliminating that toil is itself a worthwhile engineering goal.

### Design for agents using the software

[12:37](https://www.youtube.com/watch?v=XKup1pj-34M&t=757s)

Agents are also becoming the users of ordinary software. Developer tools are experiencing this transition early. Ubl reports that AI agents accounted for over 60% of page views on vercel.com during the preceding seven days. That observation is specific to Vercel’s site and that reporting window; he does not describe the traffic-classification method. Alongside it, he sees platform usage moving from dashboard clicks toward APIs and CLIs.

This changes feature review. A proposed screen is no longer enough: where is the CLI, how can the feature be automated, and how will an agent use it? As UI creation becomes cheaper, the programmatic interface becomes central to whether the product is usable by its growing machine audience. **Automation access belongs in the feature’s design, not at the end of its implementation.**

### Run the software without coupling it to the harness

[13:42](https://www.youtube.com/watch?v=XKup1pj-34M&t=822s)

Agent-written software also changes what developers expect from infrastructure. Someone who did not write the code may have fewer preferences about how it runs in production. They still expect it to run, with an experience approaching the simplicity of prompting the agent to create it. Infrastructure providers therefore have to accommodate software mostly written by agents, while agent applications themselves need new execution facilities. Sandboxes are one response. Ubl returns to just-bash as an experiment in giving an agent a computer-like environment without giving it a full computer.

The security implications remind Ubl of 1999: an industry deploying powerful systems before it has learned how to secure them reliably. His concrete architectural objection is that many popular agent harnesses combine the environment running the harness with the environment executing the code it generates. **Separate the harness from generated-code execution.** The component coordinating an agent and the environment running its generated programs should be distinct.

![Black slide reading “Software is for agents now” above “We need agentic (security) infrastructure,” with the speaker at left.](https://ai.engineer/api/talk-essay-media/7d3a46a3ec18bac1aeed9e257251b7a2c60ba3c79475a0a2dc394b728d02ad6f)

“We need agentic (security) infrastructure.”

Ubl points to an unnamed Anthropic agent product he had seen the previous day as an example that has this separation. The larger lesson is architectural openness: these problems are solvable, but today’s harness designs are not the final form. General agents using sandboxes already represent one paradigm shift. Further shifts will require engineers to reconsider assumptions about how agents run and what their infrastructure must isolate.

### Build a stable layer above changing models

[15:59](https://www.youtube.com/watch?v=XKup1pj-34M&t=959s)

The application layer can persist while models trade places. One model may lead today and another tomorrow; engineers still need to build the stable software above them. That creates room for AI innovation outside model labs. Ubl makes a regional argument here: Europe can lead in AI engineering without needing to lead model development.

His examples start with Vercel’s [AI SDK](https://vercel.com/blog/ai-sdk-5). Ubl reports that AI SDK receives over ten million downloads per week. He identifies Lars Grammel, working from Berlin, as its lead. He then points to [Pi](https://github.com/earendil-works/pi), the coding agent made in Austria, with Mario scheduled to speak the following day, and [OpenClaw](https://github.com/openclaw/openclaw), with Peter due onstage an hour later. These projects illustrate his claim that consequential AI engineering can happen independently of where frontier models are developed. His expectation that Europe will not play a major role on the model side is a forecast; his central point is that it need not do so to contribute at the application layer.

### Where the value of agents accumulates

[17:11](https://www.youtube.com/watch?v=XKup1pj-34M&t=1031s)

There are two possible futures for that layer. In the first, a model lab wins decisively. AI remains expensive, the winning company captures the value of agent technology, and application engineers effectively become its forward-deployed workforce. Ubl names OpenAI, Anthropic, and Google as possible winners in that scenario.

He considers the other future more likely: competition commoditizes the model layer. He praises **Claude** and **Codex**, predicts Google will catch up, and assigns Google an important economic role. In his assessment, Google has the cheapest infrastructure; he expects that cost advantage to encourage lower prices and sustain pressure on competitors.

| Future | Model economics | Position of application engineers |
| --- | --- | --- |
| Model-lab dominance | AI stays expensive | Value accrues to the winning lab |
| Model commoditization | Competition pushes prices down | Applications retain room to create value |

These are competing scenarios, not established outcomes. Ubl favors the second because inexpensive, competing models leave more opportunity for the applications built above them.

In that world, agents create the business value, and engineers retain the freedom to discover what comes next. OpenClaw is his example of an invention emerging from the application layer. The next paradigm of AI engineering can emerge there too: from people building useful agents, rather than only from the companies training their models.

## Resources

- [Chat SDK brings agents to your users](https://vercel.com/blog/chat-sdk-brings-agents-to-your-users) — How Chat SDK abstracts messaging APIs so agents can operate across chat platforms.
- [just-bash](https://github.com/vercel-labs/just-bash/blob/main/packages/just-bash/README.md) — A TypeScript virtual Bash environment with an in-memory filesystem, installation examples and security documentation.
- [Decoupling the agent harness from execution](https://www.anthropic.com/engineering/managed-agents) — Anthropic explains how Managed Agents separates durable sessions, agent harnesses and execution sandboxes.
- [Vercel's earlier AI support case study](https://vercel.com/blog/transforming-customer-support-with-ai-how-vercel-decreased-tickets) — A 2024 account of integrating AI into support workflows, reporting a 31% reduction in human-handled tickets.
- [AI SDK 5](https://vercel.com/blog/ai-sdk-5) — The 2025 release explains unified provider access, typed chat interfaces and agent-loop controls.
- [Pi agent toolkit](https://github.com/earendil-works/pi) — Source repository for Pi's agent toolkit and coding-agent command-line interface.
- [OpenClaw](https://github.com/openclaw/openclaw) — The open-source personal AI assistant project discussed in the keynote.

## Slides

### AI Engineer Europe

[0:00](https://www.youtube.com/watch?v=XKup1pj-34M&t=0s) [AI Engineer Europe logo centered on a black background.](https://assets.ai.engineer/sites/aiecode2025/sha256/24e440bf5dfd4ecccd4fffdc9ef9243be07fff10b16e87c6b80fbc594828e3ea.webp)

AI Engineer
EUROPE

### Presenting sponsor

Google DeepMind is the presenting sponsor.

[0:05](https://www.youtube.com/watch?v=XKup1pj-34M&t=5s) [Google DeepMind named as presenting sponsor.](https://assets.ai.engineer/sites/aiecode2025/sha256/9a43b90b83b6eca77e82f00387f025461fb2cd84d99ed7f55e0f7cc2bbdddc7d.webp)

PRESENTING SPONSOR

Google DeepMind

### Platinum sponsors

Braintrust, WorkOS, and OpenAI are platinum sponsors.

[0:10](https://www.youtube.com/watch?v=XKup1pj-34M&t=10s) [Braintrust, WorkOS, and OpenAI named as platinum sponsors.](https://assets.ai.engineer/sites/aiecode2025/sha256/124a6d271f12dad563d0b3b5b9b9a89bea7230878f0d771617e634c110bc0e77.webp)

PLATINUM SPONSORS

Braintrust
WorkOS
OpenAI

### Chat SDK and Just Bash

Malte Ubl introduces Chat SDK for connecting agents to chat apps and just-bash, a TypeScript Bash interpreter that provides a sandbox-like environment for agents.

[1:17](https://www.youtube.com/watch?v=XKup1pj-34M&t=77s) [Slide displaying the Chat SDK and Just Bash logos beside the speaker.](https://assets.ai.engineer/sites/aiecode2025/sha256/1e57fbb6342ea3559e72ad12d0395698a04a0c794cb9865faf1611079fd2cfc1.webp)

Chat SDK
JUST BASH

THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

AI Engineer EUROPE
PRESENTED BY
Google DeepMind

### From pages to agents

The slide traces a progression from sites to apps to agents as Ubl describes AI engineering as a successor to web development.

[2:36](https://www.youtube.com/watch?v=XKup1pj-34M&t=156s) [Diagram showing a progression from sites to apps to agents.](https://assets.ai.engineer/sites/aiecode2025/sha256/764bf9fd6e1f26e546b2252c2e9d94549711ab2593f6825f54e83edc9e78c1fa.webp)

From pages to agents

SITES → APPS → AGENTS

Interface text:
AI Engineer EUROPE
PRESENTED BY Google DeepMind
THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

### The disruption

Malte Ubl describes two disruptions: engineers build with AI agents, and they also build AI agents.

[3:01](https://www.youtube.com/watch?v=XKup1pj-34M&t=181s) [Two panels contrast how we build with AI agents and what we build: AI agents.](https://assets.ai.engineer/sites/aiecode2025/sha256/bf18063ad0f4db69d5761de22547f9bb2bd2813b2e6b1adb76071fc449572545.webp)

THE DISRUPTION

How we build
WITH AI AGENTS

What we build
AI AGENTS

Interface text:
AI Engineer EUROPE
PRESENTED BY
Google DeepMind
THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

### Software that would be nice to have

The circle represents software people would like to have. Ubl argues that agents could make some software that was too expensive to build traditionally economically viable.

[4:09](https://www.youtube.com/watch?v=XKup1pj-34M&t=249s) [A circle represents software that would be nice to have.](https://assets.ai.engineer/sites/aiecode2025/sha256/fd04de9e8ee5bc90b90c44b04a984fc98d9798db75dc0e6757b2e849acf9d09b.webp)

— SOFTWARE THAT WOULD BE NICE TO HAVE

Interface text:
AI Engineer EUROPE
PRESENTED BY
Google DeepMind
THE NEW APPLICATION LAYER
Braintrust
WorkOS
OpenAI

### The economics of software and agents

The speaker argues that agents make more desirable software economically viable. Their outline overlaps much more of the circle than traditional software’s, reducing the effort needed to build it.

[4:35](https://www.youtube.com/watch?v=XKup1pj-34M&t=275s) [A dashed ellipse overlaps a circle, comparing economically viable traditional software with software that would be nice to have.](https://assets.ai.engineer/sites/aiecode2025/sha256/4ff2575ad4c77c57c91bf1e2f2e788e7d6680eb2981233a7a8902ec038fec0b9.webp)

SOFTWARE THAT WOULD BE NICE TO HAVE
ECONOMICALLY VIABLE TRADITIONAL SOFTWARE

Interface text:
AI Engineer EUROPE
PRESENTED BY
Google DeepMind
THE NEW APPLICATION LAYER
Braintrust
WorkOS
OpenAI

[4:59](https://www.youtube.com/watch?v=XKup1pj-34M&t=299s) [Overlapping outlines compare desirable software, economically viable traditional software, and economically viable agents.](https://assets.ai.engineer/sites/aiecode2025/sha256/262b78cebd74dee653ec2e810752561b431eda9069f66eb7bcd08023914e342b.webp)

SOFTWARE THAT WOULD BE NICE TO HAVE
ECONOMICALLY VIABLE TRADITIONAL SOFTWARE
ECONOMICALLY VIABLE AGENTS

Interface text:
AI Engineer
Engineering the future of AI

### Buy or make

Malte Ubl says more companies are choosing to make their own software when weighing it against buying a SaaS product.

[5:21](https://www.youtube.com/watch?v=XKup1pj-34M&t=321s) [Two outlined buttons labeled Buy and Make.](https://assets.ai.engineer/sites/aiecode2025/sha256/bf311c2a3627396149c003f2601a4727078635a7908aa147892f732ed73bad5b.webp)

BUY
MAKE

AI Engineer
Engineering the future of AI
Modal

### Job postings for software engineers

Software engineer postings rebound sharply in late 2025, while overall postings remain near their lows. Ubl’s thesis is that cheaper software leads companies to make more of it, increasing demand for engineers.

[6:07](https://www.youtube.com/watch?v=XKup1pj-34M&t=367s) [Line chart compares Indeed software engineer job postings with overall postings from January 2024 to January 2026.](https://assets.ai.engineer/sites/aiecode2025/sha256/e7a5fc491e124e72a5888c0ff2423afc787d51dd74bb06678f429da96a9c2fc7.webp)

Job postings for software engineers are rapidly rising

Indeed Job Postings: Software Engineers (LHS) + 21dma
Overall Indeed Job Postings (RHS) + 21dma

Left Y-axis ticks: 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75
Right Y-axis ticks: 100, 105, 110, 115, 120, 125
Timeline: JAN 24, MAR 24, MAY 24, JUL 24, SEP 24, NOV 24, JAN 25, MAR 25, MAY 25, JUL 25, SEP 25, NOV 25, JAN 26

Interface text:
AI Engineer EUROPE
PRESENTED BY Google DeepMind
THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

### Build the next application layer

Malte Ubl calls on AI engineers to build the next application layer by building agents.

[6:45](https://www.youtube.com/watch?v=XKup1pj-34M&t=405s) [Text reads, “Let’s build the next application layer.”](https://assets.ai.engineer/sites/aiecode2025/sha256/e828ba3505a423926eb81b75c0c7e6c6c8506b8fb818c37241f79948a2280255.webp)

Let’s build the next application layer

THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

AI Engineer EUROPE
PRESENTED BY
Google DeepMind

### Archetypes of practical non-coding agents

Ubl highlights agents that run continuously or compress research. In the latter pattern, an agent gathers information after a business event, while a human still makes the decision.

[6:57](https://www.youtube.com/watch?v=XKup1pj-34M&t=417s) [An agent archetypes heading sits beside an empty outlined circle.](https://assets.ai.engineer/sites/aiecode2025/sha256/adb6ed83a2ca63054899ac431768d00fb5a8004a9804fae87d2a185353813cde.webp)

Archetypes of practical non-coding of agents

AI Engineer EUROPE
PRESENTED BY
Google DeepMind

THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

[8:55](https://www.youtube.com/watch?v=XKup1pj-34M&t=535s) [A diagram shows overlapping circles labeled “always running” and “compresses the research” inside a larger circle.](https://assets.ai.engineer/sites/aiecode2025/sha256/4ee89012f7f31a2caf038061c64f22a45132ebc439226ca493552ac14a975f17.webp)

Archetypes of practical non-coding of agents

ALWAYS RUNNING
COMPRESSES THE RESEARCH

AI Engineer EUROPE
PRESENTED BY
Google DeepMind

THE NEW APPLICATION LAYER
Braintrust
WorkOS
OpenAI

### Software is for agents now

Malte Ubl says software itself must serve agents as users. On Vercel’s own web properties, he says humans are now in the minority.

[12:44](https://www.youtube.com/watch?v=XKup1pj-34M&t=764s) [A black slide reads “Software is for agents now.”](https://assets.ai.engineer/sites/aiecode2025/sha256/3680ac0457d217d711601a9879daaf30a7e00e324a024f7b9b8356250a6aa112.webp)

Software is for agents now

THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

AI Engineer EUROPE
PRESENTED BY Google DeepMind

[12:59](https://www.youtube.com/watch?v=XKup1pj-34M&t=779s) [Slide stating that software is for agents now.](https://assets.ai.engineer/sites/aiecode2025/sha256/718e5550b6af9fa9dca2cdc3b519815af3a7dc65911678cee4dc622da6deaa6e.webp)

Software is for agents now

THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

AI Engineer EUROPE
PRESENTED BY Google DeepMind

### 60% of Vercel.com page views are agents

Malte Ubl says over 60% of Vercel.com page views in the last seven days came from AI agents. He also sees platform usage shifting from the dashboard to the API and CLI.

[13:06](https://www.youtube.com/watch?v=XKup1pj-34M&t=786s) [Slide stating that 60% of Vercel.com page views are agents.](https://assets.ai.engineer/sites/aiecode2025/sha256/e2c2815a6f5c59068c4b6fb9348c16c53fd7697f176f0379f0d2bf067a10d158.webp)

SOFTWARE IS FOR AGENTS NOW

60%
of Vercel.com page views are agents

THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

AI Engineer EUROPE
PRESENTED BY
Google DeepMind

### Agents should run the infrastructure

Malte Ubl argues that infrastructure must shift as agents write more software: people expect the resulting code to run in production after they prompt an agent to do the work.

[13:41](https://www.youtube.com/watch?v=XKup1pj-34M&t=821s) [Slide stating that agents should run the infrastructure.](https://assets.ai.engineer/sites/aiecode2025/sha256/4b1181e0e8b496ac09eaefd3ac867ce28d2f1bc532c559e88c17f6c7d46ea2e6.webp)

SOFTWARE IS FOR AGENTS NOW

Agents should run the infrastructure

Interface text:
AI Engineer EUROPE
PRESENTED BY Google DeepMind
THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel
neo4j
OpenAI
SENTRY
Modal

[13:47](https://www.youtube.com/watch?v=XKup1pj-34M&t=827s) [Slide states that agents should run the infrastructure.](https://assets.ai.engineer/sites/aiecode2025/sha256/082e0f56b9f651661ed6a83f17efde086af82f84756cb25e8278888b493bce37.webp)

SOFTWARE IS FOR AGENTS NOW

Agents should run the infrastructure

Interface text:
AI Engineer EUROPE
PRESENTED BY Google DeepMind
THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel
Modal

### We need agentic security infrastructure

Malte Ubl says agent applications need infrastructure such as sandboxes. He warns that popular agent harnesses run alongside the code they generate, creating a security concern.

[14:10](https://www.youtube.com/watch?v=XKup1pj-34M&t=850s) [Slide states that agentic security infrastructure is needed.](https://assets.ai.engineer/sites/aiecode2025/sha256/60f28e54c21d3deeba9c19804339fae48088fdc7200f548fe8c69ab1f7839fa7.webp)

SOFTWARE IS FOR AGENTS NOW

We need agentic (security) infrastructure

Interface text:
THE NEW APPLICATION LAYER
AI Engineer EUROPE
PRESENTED BY
Google DeepMind
Braintrust
WorkOS
OpenAI
Modal

### The new application layer can thrive independently of models

Ubl argues that AI engineers can build a stable application layer as the best model changes, and can drive AI innovation without working at a model lab.

[15:56](https://www.youtube.com/watch?v=XKup1pj-34M&t=956s) [Slide states that the new application layer can thrive independently of models.](https://assets.ai.engineer/sites/aiecode2025/sha256/531f0624f753c115a201e4c0fe7ccd094088d1bccbb10a5dbe608a1da283d5da.webp)

The new application layer can thrive independently of models

THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

AI Engineer EUROPE
PRESENTED BY
Google DeepMind

### Europe is the leader in AI engineering innovation

Ubl cites Vercel’s AI SDK, led by Berlin-based Lars Grammel, and Austria’s Pi coding agent as examples of European AI engineering innovation outside model labs.

[16:29](https://www.youtube.com/watch?v=XKup1pj-34M&t=989s) [Slide asserts that Europe leads in AI engineering innovation, beside a dotted map.](https://assets.ai.engineer/sites/aiecode2025/sha256/ecdf134235f0702164460ce83251b407f191a988e11be8fb25a6903b2dfcf35e.webp)

Europe is the leader in AI Engineering innovation

THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

AI Engineer EUROPE
PRESENTED BY
Google DeepMind

### The futures ahead: The labs win

Ubl argues that if model labs win, AI stays expensive and agent value accrues to the winner. He considers that future unlikely as model companies commoditize.

[17:14](https://www.youtube.com/watch?v=XKup1pj-34M&t=1034s) [A slide contrasts two futures. Future A says “The labs win” over a halftone image; Future B is empty.](https://assets.ai.engineer/sites/aiecode2025/sha256/fe27b45dd66538e963157567edb9386ed0d59d74ae5215aa344c5ba153c83527.webp)

THE FUTURES AHEAD

FUTURE A
The labs win

FUTURE B

Interface text:
AI Engineer EUROPE
PRESENTED BY Google DeepMind
THE NEW APPLICATION LAYER
MALTE UBL / CTO
Vercel

## Full transcript

[0:00](https://www.youtube.com/watch?v=XKup1pj-34M&t=0s) [upbeat music] Our first speaker draws on over 25 years of software engineering experience.

[0:22](https://www.youtube.com/watch?v=XKup1pj-34M&t=22s) From his time at Google and now Vercel, he will explore what it means to build infrastructure and applications in a world where agents are both the builders and users of software.

[0:35](https://www.youtube.com/watch?v=XKup1pj-34M&t=35s) Please join me in welcoming to the stage the CTO of Vercel, Malte Ubl. [upbeat music] [audience applauding]

[0:50](https://www.youtube.com/watch?v=XKup1pj-34M&t=50s) Good morning, everyone. This is awesome. I'm so glad to be here. Welcome to the first ever AI engineer conference in Europe. Um, my name is Malte, and I'm the CTO of Vercel.

[1:02](https://www.youtube.com/watch?v=XKup1pj-34M&t=62s) Now, this is not... You know, usually I give technical talks, but I thought because I'm currently going first, that I need to give a proper keynote. But I did want to feature what I call my vibe coding, uh, stack.

[1:14](https://www.youtube.com/watch?v=XKup1pj-34M&t=74s) Uh, I've been hacking on a thing called Chat SDK, which is a way to hook your agents to whatever, like Slack, Telegram, WhatsApp, chat app, you know, you like.

[1:24](https://www.youtube.com/watch?v=XKup1pj-34M&t=84s) And I've been working on just-bash, which is a Bash interpreter written in TypeScript that gives you something like a sandbox with zip nanosecond startup time, um, for your agents because they love Bash.

[1:37](https://www.youtube.com/watch?v=XKup1pj-34M&t=97s) All right. One thing I wanted to mention is that the reason why I'm so excited to be here is that I used to run a little conference in Berlin called JSConf EU.

[1:47](https://www.youtube.com/watch?v=XKup1pj-34M&t=107s) And I feel I... Once in my life, I had completely impeccable timing because it was the summer of 2019, and we decided after 10 years it was enough, and we went out with a bang.

[1:59](https://www.youtube.com/watch?v=XKup1pj-34M&t=119s) And the reason why this was such great timing was that there just wouldn't have been a conference one year later because of COVID. But also, when we decided that we would, you know, step away, we were hoping that someone else would take the reins, and again, that did not happen because of COVID.

[2:17](https://www.youtube.com/watch?v=XKup1pj-34M&t=137s) So it's now been more than half a decade, and I'm very excited that things are finally starting up again.

[2:23](https://www.youtube.com/watch?v=XKup1pj-34M&t=143s) But it was also clear that it wasn't going to be like a web development conference that would really bring the tech community in Europe back together in 2026, right?

[2:32](https://www.youtube.com/watch?v=XKup1pj-34M&t=152s) In many ways, I think AI engineering is the legitimate successor to web development as a really mainstream discipline of engineering that will shape the next decade of software development as, you know, software eng- engineering itself faces an unprecedented disruption.

[2:50](https://www.youtube.com/watch?v=XKup1pj-34M&t=170s) So you're definitely in the right place today, and it's more important than ever to come together as a community and reflect on both our profession as software engineers and AI engineers.

[3:01](https://www.youtube.com/watch?v=XKup1pj-34M&t=181s) And that's because we're facing a disruption of both how we build, which is with AI, and what we build, which is AI and agents.

[3:11](https://www.youtube.com/watch?v=XKup1pj-34M&t=191s) And of course, disruption can sometimes lead to anxiety. In fact, I really actually very often get asked, "Hey, Malte, is there still a place for engineers in the future?

[3:22](https://www.youtube.com/watch?v=XKup1pj-34M&t=202s) And what about that next generation of engineers?"

[3:26](https://www.youtube.com/watch?v=XKup1pj-34M&t=206s) And I couldn't be more convinced that the answer is yes.

[3:30](https://www.youtube.com/watch?v=XKup1pj-34M&t=210s) I often give this example of like envision me doing a TikTok video. They mentioned in the intro that I have 25 years of experience, which is actually substantial understatement.

[3:39](https://www.youtube.com/watch?v=XKup1pj-34M&t=219s) And so I would not be good at a TikTok video. I should not be recording TikTok videos because I didn't grow up with this, right? And in a very similar way, the next generation of junior engineers are just gonna be so much better at this discipline because they get molded in the AI world, just like all of

[3:55](https://www.youtube.com/watch?v=XKup1pj-34M&t=235s) you are. But it's not only the kids that are gonna be all right. We'll all be fine, and this is why. One of our main thesis is that agents are a new kind of software.

[4:09](https://www.youtube.com/watch?v=XKup1pj-34M&t=249s) Because there was always all this stuff we wanted to automate, but not all of it was economically viable to do with traditional software, but it is with agents. And what that means is there will be just so much more software in the future.

[4:23](https://www.youtube.com/watch?v=XKup1pj-34M&t=263s) Indulge me with a Venn diagram. Um, maybe the circle should be bigger because the circle represents all software that should exist. Imagine all software that should exist. The problem was that we couldn't write all of it because it was too expensive using traditional methods.

[4:42](https://www.youtube.com/watch?v=XKup1pj-34M&t=282s) Like you can envision like all these things where like you have all this if statements, you have all this like knowledge about the c- the business. Like you have to figure it all out.

[4:49](https://www.youtube.com/watch?v=XKup1pj-34M&t=289s) You have to hard code into the application. So much of this software you just would never write because it would be obviously, uh, too expensive.

[4:58](https://www.youtube.com/watch?v=XKup1pj-34M&t=298s) But now with agents, that part of software becomes economically viable. I can build it now, um, with not that much e- much effort, right? And that means that with AI agents, essentially all the software, maybe not all of it, we'll find more in the future, but like that circle gets filled out, right?

[5:17](https://www.youtube.com/watch?v=XKup1pj-34M&t=317s) All of this stuff that should be automatable is automatable. There, there's gonna be so much more software out there.

[5:24](https://www.youtube.com/watch?v=XKup1pj-34M&t=324s) And in a similar line, more and more companies when they ask that question whether they should buy some software like a SaaS product or make some software themselves, they're answering that with the make side, right?

[5:38](https://www.youtube.com/watch?v=XKup1pj-34M&t=338s) Over in Silicon Valley where I live today, we are talking a lot about the SaaS apocalypse, I think is what it's called, right? People like make their own stuff, and they don't buy the SaaS software anymore.

[5:51](https://www.youtube.com/watch?v=XKup1pj-34M&t=351s) I actually think the SaaS companies will be all right. You know, don't worry about them. But as engineers ourselves,

[5:59](https://www.youtube.com/watch?v=XKup1pj-34M&t=359s) more companies making more software, again, leads to us having more work even if it's faster.

[6:06](https://www.youtube.com/watch?v=XKup1pj-34M&t=366s) And in fact, the way I've been kind of framing this for a while is that we are speed running what's really an experiment in economics of how elastic the software market is.

[6:17](https://www.youtube.com/watch?v=XKup1pj-34M&t=377s) The thesis being that the cheaper it is to make software, the more software we're going to make, right?

[6:24](https://www.youtube.com/watch?v=XKup1pj-34M&t=384s) And as a consequence, what's actually happening is the demand for software engineers is going up. Now, we don't know, like, you know, like, there's gonna be an S-curve, you know?

[6:36](https://www.youtube.com/watch?v=XKup1pj-34M&t=396s) But there's no signs of, of, of us reaching the S-curve. In fact, because we're getting better at agents, et cetera, there's so much leeway in the future, um, I think we'll be all right.

[6:47](https://www.youtube.com/watch?v=XKup1pj-34M&t=407s) So as AI engineers, it's our job to build that next application layer.

[6:53](https://www.youtube.com/watch?v=XKup1pj-34M&t=413s) And of course, what that actually means is building agents, right?

[6:58](https://www.youtube.com/watch?v=XKup1pj-34M&t=418s) I wanted to spend some time talking to-- talking about archetypes of agents that I'm seeing actually being built today, actually being effective, actually something you can do today without, you know, having to make major changes.

[7:10](https://www.youtube.com/watch?v=XKup1pj-34M&t=430s) I think we're all a little bit drunk on the coding agents because they're so great, right? They work so well, and it seems so obvious that you can translate that to all other domains, and, and sometimes these things don't go so well, right?

[7:23](https://www.youtube.com/watch?v=XKup1pj-34M&t=443s) But the thing is that we don't really have to be doing the most advanced agents you could possibly imagine. That there's just so much low-hanging fruit to be, to be, be done where you can really, really help companies save them millions or billions of dollars without actually, you know, making these massive changes of processes that in practice

[7:41](https://www.youtube.com/watch?v=XKup1pj-34M&t=461s) will always take a long time, fail often, et cetera.

[7:45](https://www.youtube.com/watch?v=XKup1pj-34M&t=465s) So this is what I'm actually seeing in the wild. The first part, when you think about what agents you can build, is people think, "Well, agents? That rings a bell.

[7:54](https://www.youtube.com/watch?v=XKup1pj-34M&t=474s) I have a team of support agents. Maybe I can automate part of that," right? And that's also where kind of the first generation of what we call s- agent as a service, you can make that acronym in, in your head, um, startups are, are shipping, right?

[8:10](https://www.youtube.com/watch?v=XKup1pj-34M&t=490s) Like, you know, the, the Sierras and Decagons of this world. Um, but more generally, I think it's worth asking yourself, in your business,

[8:18](https://www.youtube.com/watch?v=XKup1pj-34M&t=498s) is there a job where it would be quite transformative if, if that, you know, we went from a nine-to-five thing 'cause, you know, people need to sleep, and I can actually do twenty-four/seven because agents don't need to sleep, right?

[8:30](https://www.youtube.com/watch?v=XKup1pj-34M&t=510s) And I think there's, there's many places for that.

[8:33](https://www.youtube.com/watch?v=XKup1pj-34M&t=513s) The next one is probably actually even more important. Um, I call it, call it compress the research. Because every business has a certain type of business process that in a very abstract fashion has the following shape.

[8:46](https://www.youtube.com/watch?v=XKup1pj-34M&t=526s) There's some business event, and you have to do some research, and then you make a human decision, right? And you can just build an agent that does the research phase automatically, and that's all you do.

[8:59](https://www.youtube.com/watch?v=XKup1pj-34M&t=539s) That's all you ship, right? And the important part why this is, like, such an easy thing to ship is because the process is still the same. There's still that business event, there's another research,

[9:09](https://www.youtube.com/watch?v=XKup1pj-34M&t=549s) and there's a human decision. The research goes faster, and, you know, maybe it goes from something that took a human thirty minutes. Now they can do the same thing in five minutes, and if you run that process a hundred thousand times a year, you just save the company a wh-whole lot of money.

[9:23](https://www.youtube.com/watch?v=XKup1pj-34M&t=563s) But you didn't increase the risk profile, and you didn't have to change the process.

[9:27](https://www.youtube.com/watch?v=XKup1pj-34M&t=567s) At Vercel, we actually have, like, at least two agents of this shape. When you go to vercel.com and you hit the Contact Sales button,

[9:35](https://www.youtube.com/watch?v=XKup1pj-34M&t=575s) that message actually goes to an agent, right? And I hear about seventy-five percent of the time that agent says, "Well, actually, they just wanted support," and hand it over to the, the support team.

[9:46](https://www.youtube.com/watch?v=XKup1pj-34M&t=586s) But then in the other case, it will go, "Oh, that's interesting. Um, let me check out their LinkedIn. Let me google the company. Let me figure out the-- how large they are.

[9:54](https://www.youtube.com/watch?v=XKup1pj-34M&t=594s) Let me route it to the right person," right? And then there's a human eventually taking a look that it makes sense, but that obviously was something that took maybe a person fifteen minutes before, and now they don't have to do it anymore.

[10:07](https://www.youtube.com/watch?v=XKup1pj-34M&t=607s) And another example is exact- the same process. If you sent us a re- abuse report,

[10:13](https://www.youtube.com/watch?v=XKup1pj-34M&t=613s) again, there's an agent taking a look. Is that website abusive? What, what should we do, right? Still, obviously, the decision in the end should be done by, by the actual professional, but they don't have to, like, do all this research themself anymore.

[10:29](https://www.youtube.com/watch?v=XKup1pj-34M&t=629s) Next is what I think is probably the most magical thing that you can do in any company today,

[10:36](https://www.youtube.com/watch?v=XKup1pj-34M&t=636s) which is to surface information that already exists.

[10:40](https://www.youtube.com/watch?v=XKup1pj-34M&t=640s) It's extremely common that there's information somewhere in the company, right? But for all intents and purposes, you cannot practically use it.

[10:52](https://www.youtube.com/watch?v=XKup1pj-34M&t=652s) Take, for example, everyone, y- you all engineers, you have issue trackers, right? So

[10:59](https://www.youtube.com/watch?v=XKup1pj-34M&t=659s) is it up to date? Probably not all the time, right? Could it be up to date? Like, does the information exist? Did you Slack it? Did you have a Granola recording that technically contains the information that could update your issue tracker?

[11:12](https://www.youtube.com/watch?v=XKup1pj-34M&t=672s) Yes, right? Like, probably yes. And so you can, you know, build an agent that does this for your company, right? Whenever you have, like, a manager saying, "Well, give me a last-- list of updates," right?

[11:24](https://www.youtube.com/watch?v=XKup1pj-34M&t=684s) Why don't they already have the updates, right? Why doesn't an agent have already kind of done that research already? Um, so again, this just makes-- takes advantage of existing information, which is so powerful.

[11:36](https://www.youtube.com/watch?v=XKup1pj-34M&t=696s) And finally, for the last big category, um,

[11:40](https://www.youtube.com/watch?v=XKup1pj-34M&t=700s) there's a magical question that you can do to figure out agents you should build in your company, which is to ask folks, "What do you hate most about your job?"

[11:50](https://www.youtube.com/watch?v=XKup1pj-34M&t=710s) And I actually have a case study about this in-- at Vercel. So we actually did build our own in-house support agent, and it has what's called a ninety percent deflection rate.

[11:58](https://www.youtube.com/watch?v=XKup1pj-34M&t=718s) So ninety percent of the time it just helps the person in real time rather than going down somewhere else. And what happened? This job satisfaction rate on our support team exploded.

[12:10](https://www.youtube.com/watch?v=XKup1pj-34M&t=730s) Why? Because they no longer have to do the boring stuff, right? "Oh, my credit card got rejected," blah, blah, blah, right? Now they get to actually go and figure out these actually interesting cases, actually help people who really need help, rather than doing all the toil, right?

[12:25](https://www.youtube.com/watch?v=XKup1pj-34M&t=745s) So that's, like, I think eliminating boring work is a very noble mission that we should all kind of strive to do for the companies that we work for.

[12:37](https://www.youtube.com/watch?v=XKup1pj-34M&t=757s) Cool. So clearly that new application layer are agents, but we also have to shift, uh... we have to consider another shift, that the software itself is going to be used by agents now, right?

[12:49](https://www.youtube.com/watch?v=XKup1pj-34M&t=769s) And, you know, I work in software development, developer tools, et cetera, and I think we're kind of ahead of the game here, speed running that transformation. Um, what I will share, though, is that, you know, on our own web properties, humans are actually now in the minority.

[13:08](https://www.youtube.com/watch?v=XKup1pj-34M&t=788s) So in the last seven days, and we have not shared this before, over sixty percent of page views on vercel.com were AI agents. In a similar way, we're seeing the way you use our platform going from people clicking around in the dashboard to, uh, usage shifting to our API and CLI.

[13:26](https://www.youtube.com/watch?v=XKup1pj-34M&t=806s) So whenever I now, you know, have someone proposing a feature to me and they show me, like, a UI, I'm like, "Guys, what's the CLI? Like, how do you-- How do I automate this?

[13:35](https://www.youtube.com/watch?v=XKup1pj-34M&t=815s) How do, how does an agent use this?" You know? You know, UI is now something that's so cheap.

[13:42](https://www.youtube.com/watch?v=XKup1pj-34M&t=822s) The other thing that we're observing is that kind of the relationship changes between software development, software developers and infrastructure, right? If I didn't write the code myself, I also don't have maybe as strong feelings about how that stuff runs in production, right?

[13:57](https://www.youtube.com/watch?v=XKup1pj-34M&t=837s) And so for a company like us, it's really important that we shift how we deploy infrastructure to a model where most of the software was written by agents and, you know, has to just run, and people, like, expect it to run just like they prompted the agent to do the work.

[14:13](https://www.youtube.com/watch?v=XKup1pj-34M&t=853s) And finally, and nobody here obviously is surprised about this, the applications themselves are, you know, they're agents now, and that requires us to have different infrastructure available, right? Everyone's now shipping sandboxes.

[14:26](https://www.youtube.com/watch?v=XKup1pj-34M&t=866s) I think it's almost a meme. Um, I was mentioning earlier that I created this thing called just-bash, and I'm really interested in kind of this innovation of how you can give an agent a computer maybe without giving them a computer.

[14:38](https://www.youtube.com/watch?v=XKup1pj-34M&t=878s) There's lots of interesting stuff there in the market, um, and I'm, I'm sure this conference is gonna have lots of stuff there as well.

[14:46](https://www.youtube.com/watch?v=XKup1pj-34M&t=886s) And then also more broadly, again, it was mentioned, [laughs] I've been here for, for a while, like, we're, we're, like, marching head-on into a security nightmare. It almost feels like a little bit like nineteen ninety-nine, where really everything can be hacked, right?

[14:58](https://www.youtube.com/watch?v=XKup1pj-34M&t=898s) And we just didn't know how to make something secure. Um, I think we'll have a rude awakening, but what that really means is that we have to be open-minded for, for how to change things.

[15:10](https://www.youtube.com/watch?v=XKup1pj-34M&t=910s) Uh, I will give one example. Um, I think almost all currently popular agent harnesses have fundamentally the wrong architecture, and that is that they combine where the harness runs with where the code that it runs, uh, that it generates runs, right?

[15:25](https://www.youtube.com/watch?v=XKup1pj-34M&t=925s) Um, as of actually yesterday, I, I did see that Anthropic dis- agrees with that thesis 'cause they, they, um, on their new agent product, they do have that separation, and it's really, really key.

[15:36](https://www.youtube.com/watch?v=XKup1pj-34M&t=936s) And that's really just also a point that, that these are all solvable problems, but my ma- main message today is that we are still in the very early innings, and we have to be prepared to be open-minded about, like, paradigm shift happening in the future, right?

[15:48](https://www.youtube.com/watch?v=XKup1pj-34M&t=948s) We just had the paradigm shift of agents being kind of these, like, very general sandbox-using things. In the future, we will see more of those paradigm shifts.

[15:59](https://www.youtube.com/watch?v=XKup1pj-34M&t=959s) Cool. Um, last point I wanted to make is that this new application layer that we're building can thrive independent of the models, right? 'Cause sometimes model X is better, sometimes model Y is better, but we are, as AI engineers, building a stable layer on top.

[16:17](https://www.youtube.com/watch?v=XKup1pj-34M&t=977s) And one of the very interesting consequences is that we don't have to work at a model lab to m- drive AI innovation. In fact, and I think this is almost like a narrative violation, Europe is the leader in AI engineering innovation, right?

[16:33](https://www.youtube.com/watch?v=XKup1pj-34M&t=993s) Um, our own AI SDK, which Vercel makes, it gets now over ten million downloads a week, and it's led by Lars Grammel, who lives in Berlin, right? Um, he's working on this.

[16:44](https://www.youtube.com/watch?v=XKup1pj-34M&t=1004s) Then there's obviously Pi, the coding agent, um, made in Austria. You'll be hearing from Mario about it tomorrow. And of course, probably some of you have heard of it, there's a little thing called OpenClaw, um, and Peter will be on stage here in an hour.

[16:58](https://www.youtube.com/watch?v=XKup1pj-34M&t=1018s) And so it appears to be that Europe, against all odds, is taking actually a leadership role in AI engineering. But we also have to be realistic, right? Like, Europe isn't gonna play a major role on the model side, but I don't think it needs to.

[17:11](https://www.youtube.com/watch?v=XKup1pj-34M&t=1031s) In fact, I do see kind of two big futures ahead of us.

[17:15](https://www.youtube.com/watch?v=XKup1pj-34M&t=1035s) One is where the big model labs win. In that world, AI will stay very expensive. All the value of all that cool agent tech will accrue to that company, and we won't really be AI engineers anymore, right?

[17:31](https://www.youtube.com/watch?v=XKup1pj-34M&t=1051s) We'll be, like, forward-deployed engineers who- whoever the winner is, if it's OpenAI, Anthropic, or Google.

[17:39](https://www.youtube.com/watch?v=XKup1pj-34M&t=1059s) But I don't think that's very likely, and I think what's actually going on is that the opposite is happening. The model companies are commoditizing. Claude is amazing. Codex is amazing.

[17:49](https://www.youtube.com/watch?v=XKup1pj-34M&t=1069s) Google will catch up. And importantly- [audience laughing] I'll give them props now because I think Google's playing an amazing role here because they have the cheapest infrastructure on the, on the cost side.

[18:01](https://www.youtube.com/watch?v=XKup1pj-34M&t=1081s) And so in that commoditized world, they will always decide to make it cheaper, right? And that will keep the price for where it should be, which is very low, and that's the outcome that we want, right?

[18:12](https://www.youtube.com/watch?v=XKup1pj-34M&t=1092s) Because in that world, we, the AI engineers, are the powerful ones. Our agents are the ones that actually create the business value,

[18:19](https://www.youtube.com/watch?v=XKup1pj-34M&t=1099s) and it's the application layer where the real innovation happens, right? This is where OpenClaw gets invented, and that's where the next paradigm of AI engineering gets discovered. And that's really all I wanted to leave you with today.

[18:32](https://www.youtube.com/watch?v=XKup1pj-34M&t=1112s) Thank you very much. [audience applauding] [upbeat music]
