Browse · Organizations
Organization · Latent Space wiki

AIE Europe

AI Engineer’s London conference operation, combining technical programming on agents, models and infrastructure with open conference-data interfaces.

AIE Europe, or AI Engineer Europe, is a technical conference organised by Software 3.0, Inc. Its April 8–10, 2026 London edition brought the AI Engineer multitrack conference format to Europe, with workshops, keynotes and sessions covering agents, model development and AI infrastructure. The organiser described it as Europe’s first flagship AI Engineer event.[4][5]

2026 rollout and programming

The rollout was underway by February 19, 2026, when Latent Space announced that tickets were live and first-wave speakers had been announced. On March 5, it highlighted a Harness Engineering track, alongside the publication’s growing assessment that the systems surrounding models had practical value.[1][3]

The track addressed a concrete debate in agent engineering: how much value comes from a model’s capabilities, and how much comes from the software that connects tools, manages context and orchestrates execution. Latent Space illustrated the basic agent loop as executing a model’s tool calls, capturing results, appending them to context and calling the model again. The conference gave this systems work a dedicated place in its program.[3][4]

April 8 was the workshop day; April 9–10 were keynote and session days. Tracks covered personal agents, context engineering, harness engineering, evaluations and observability, voice and vision, coding agents, MCP, AI architecture, generative media, and GPUs and LLM infrastructure. In the debrief published April 23, swyx said he curated the tracks and identified OpenClaw, harness engineering and context engineering as leading interests, alongside recurring infrastructure and evaluation topics.[4][8]

The published schedule paired these themes with named contributors. OpenAI’s Ryan Lopopolo was listed for “Harness Engineering: How to Build Software When Humans Steer and Agents Execute” and an OpenAI Symphony and harness-engineering AMA moderated by independent researcher Vibhu Sapra. Google DeepMind research scientist Sander Dieleman was listed to discuss building generative image and video models at scale; Factory’s Matan Grinberg and Luke Alvoeiro were listed together for “Factory Missions - Multi-Agent Systems That Ship for Days.”[6]

Conference data for developers and agents

AI Engineer Europe also offered conference data as material for developers to build with. Its schedule published references to session and speaker JSON, llms.txt, iCal, embeddings and MCP, inviting attendees to create their own schedule applications. The event page advertised agent skills and the npx @aidotengineer/aie CLI alongside those interfaces.[6][4]

The advertised MCP server lets agents query conference data through list_conferences, search, fetch and get_schedule tools. The embedding offering was labelled Gemini Embedding 2, with MRL 128-dimensional vectors. These interfaces extended access beyond the interactive schedule to apps, agents and other tools; the organiser described the endpoints as free and open.[4]

Scale, sponsorship and operations

The organiser advertised 100+ speakers, 100+ sessions, 11 tracks and 23 workshops across three days, with 1,000+ engineers expected to attend. The attendance figure was a projection. Google DeepMind was labelled founding sponsor, while Together AI identified itself as a Gold Sponsor and advertised booth G1, plus presentations on context parallelism by Max Ryabinin and voice-agent engineering by Rishabh Bhargava.[4][7]

The April 23 debrief, recorded just after AIE Europe, placed London within a wider event operation that swyx described as running roughly one conference per quarter. He also discussed a possible operational change: replacing event and sponsor management SaaS with custom AI-built software. He said bringing his team along was a constraint and that he was still hesitating to proceed. Jacob Effron and the Unsupervised Learning production team hosted and edited the crossover episode.[8]

From the Latent Space archive

Original posts and separately reviewed explanations.

Sources

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8

Model source and teaching reviews passed. Human editorial review pending. This profile describes the sources as of 2026-09-17.

Latent Space

Join Latent Space.

Sign in with Google, or get the newsletter by email.

Loading Google sign-in…

or subscribe by email

Free. Unsubscribe any time.