Presenters

• swyx

Source

• AI Engineer Europe 2026

AI Engineers: Building the Future, One Agent at a Time 🤖✨

The world of work is rapidly evolving, and at the forefront of this transformation are AI engineers. Swyx, co-founder of the AI Engineer conference, shares his journey and insights on how AI is not just a tool for coding, but a powerful force for augmenting human productivity and reimagining business operations.

The Evolution of AI Engineer Conferences 🚀

Swyx has been a consistent voice at the AI Engineer conference, tracking the seismic shifts in AI adoption.

• Year 1: Focused on the productivity gains from increased AI usage.
• Year 2: Advocated for more AI adoption, highlighting the dramatic cost reductions in AI technology (roughly 100x every 12-18 months).
• Year 3: Introduced the concept of tiny teams – teams generating millions in revenue with fewer employees. This emphasis stems from the belief that every company, regardless of size, can benefit from this lean operational model. Swyx himself leads AI Engineer with a tiny team of nine full-time people, generating over $9 million in revenue.

The “AGI Pill” Moment: Integrating AI into Workflow 💊

Swyx recounts a pivotal moment when he started seriously integrating coding agents into his workflow at Cognition. Initially driven by the fact that they were free, this experiment led to a profound realization about the true potential of AI.

From Static Stack to Dynamic Design 🎨➡️💻

The AI Engineer conference’s tech stack was traditionally non-AI: Figma, React, Supabase, Tidio, Google Sheets, and Sessionize. The integration of AI, specifically Devin, began subtly.

• The Figma Challenge: A contract designer presented a Figma page, expecting a lengthy turnaround time (weeks). Swyx integrated Devin into the workflow, hooking it up to Figma (with the help of a coworker, highlighting the usefulness of tools like Coworker for such tasks).
• The “Yak Shaving” Revolution: A significant benefit of agents like Devin is their ability to eliminate “yak shaving” – the tedious, dependency-tree crawling often associated with development. Fixing dependency issues, especially in Python, is a prime example of where agents excel.
• Pixel-Perfect Reality: The result was a perfectly functioning website, pixel-perfect to the original Figma design, delivered in a surprisingly short time. This success validated the power of AI agents, even for those initially skeptical of marketing claims.

The Explosion of Agent Usage and Human Augmentation 📈

Following the initial success, agent usage skyrocketed. The AI Engineer’s Slack channel saw a significant surge in activity, with 207 replies indicating widespread adoption.

• Asynchronous Collaboration: Swyx observed a fascinating dynamic: he would start a task, go to bed, and his designer in Indonesia would wake up and continue working with Devin. This designer, without explicit instruction, began prompting Devin with redlines on annotations, mimicking human communication.
• Empowering Non-Technical Teams: This seamless interaction is crucial for teams with non-technical members, fostering comfort and engagement with AI agents.
• Beyond the Expected: The team started tackling projects they previously wouldn’t have considered. An Easter egg on the website, a fun addition by Swyx, exemplifies this newfound creative freedom.
• The Fun Factor: A viral tweet about a design aesthetic inspired an impromptu project. Swyx initially intended it as a throwaway experiment to see what Devin could do. However, his designer, motivated by the enjoyment of the process and the elimination of waiting times for feedback or developer input, jumped in and began actively working on it.
• Unlocking Human Potential: The most significant revelation for Swyx was that AI agents were increasing productivity by making work more enjoyable for humans. This led to more animations, polish, and overall output from his team, work they simply hadn’t done before. The focus shifted from lines of code to augmented human output.

AI as a Data Management Powerhouse 📊

Swyx then explored how AI agents could tackle the complex data management challenges of running a conference.

• Conference as a Data Problem: Organizing a conference with 130 speakers, sponsors, and attendees involves significant data synchronization and management.
• Devin and Content Management: Instead of relying on traditional CMS tools like Sanity, Swyx committed the data to code and used Devin to manage it. This approach allowed for easier updates and synchronization.
• Automated Speaker Updates: When a speaker like Marta submits a change, Swyx simply forwards the email or screenshot to Devin, who handles the update. This allows a small team of nine people to manage a thousand-person conference, with plans to scale to 6,000 people in San Francisco without increasing team size.
• Agents for ETL and Beyond: AI agents are now being used for routine tasks like ETL (Extract, Transform, Load), syncing data from external vendors and ensuring a single source of truth.

Agents for Purchasing and Personal Automation 🛒

The application of AI extends even to personal automation and purchasing.

• The Lobster Acquisition: Inspired by a viral tweet, Swyx decided to place a lobster statue in front of the conference. He tasked Devin with researching where to buy one in London. Devin provided phone numbers, email addresses, and websites, leading to the successful acquisition of the lobster seen at the event.
• Serverless, On-Demand Knowledge Work: This demonstrates how agents with web access can handle tasks that would otherwise require an executive assistant or junior employee, making them available serverless and on-demand.

The Future is Agent-Centric 🌐

Swyx emphasizes that the trend is not limited to coding agents.

• Beyond Coding Containment: He foresees an explosion of purpose-fit knowledge management tools and agents, citing examples like wikis and tools adopting nano-claws (referencing OpenAI’s GPT Store and similar initiatives). This is identified as a top 3-5 trend for 2026.
• From Apple Notes to Notion: Swyx showcased how he used Town (likely a typo or a specific tool he used) to transform his Apple Notes into a nicely formatted Notion document with research on potential speakers for the World’s Fair.
• Replacing SAS Tools: The ultimate ambition for some, including Swyx, is to replace entire SAS tools with custom-built solutions powered by AI. He acknowledges the “psychosis” involved but stresses the importance of addressing employee concerns and systematically reducing them.

The Shifting User Landscape: Bots Over Humans 🤖👥

The final point highlights a fundamental shift in user behavior, as articulated by Malte in his opening keynote.

• Vercel’s Bot Majority: 60% of Vercel’s user base is now bots or agents, not humans. This means dashboards are becoming less relevant, while APIs, CLIs, and MCPs (Machine Control Panels) are gaining prominence.
• Agent Experience (AX): Edo and Liad’s talk on ETM suggests that custom UIs are diminishing, and the focus is shifting towards shipping UI to existing apps. This leads to the concept of Agent Experience (AX), where the primary user is changing.
• Embrace the Change: Swyx’s closing message is clear: agents for everything else are coming. He urges everyone to wake up, use them, and bring them home to work. For those not yet convinced, he humorously suggests prescribing an “AGI pill.”

The future of work is here, and it’s being built by AI engineers, augmented by intelligent agents, and driven by a relentless pursuit of innovation and efficiency. 🚀

Appendix