AI.Engineer Conference NYC 11/19-11/21

reddit [#143] Created with Sketch.
James Farrelly
Site Author
1 min read

AI Engineer Code Summit Conference

The AI Engineer conference is very focused on the engineering of AI solutions.
Deep dives into modern vibe coding, guardrails, code quality, speed, and maintainability in the new era of AI-assisted coding development.

Agents Trust, Autonomy, and Parallelism

There is a lot of work coming from Google, Claude, and others that will enable “vibe” coding agents to start running better orchestrated smaller context windows to produce faster and more focused results.

  • Anthropic discussed context window cleaning
  • Google discussed agent parallelism and agents that are trusted to be more autonomous

Who Will Babysit the Agents?

Agents are getting better at managing their own context, dividing work into smaller and more focused elements, and having long-running parallel workflows to get tasks done.
However, this means that we will need better agent observability.

Agent Optimization

Agents will be running more autonomously. We need a way to understand what they are doing and monitor their performance over time.

AI Impact on Code and Engineering

  • At least four vendor displays and two stage presentations focused on the quality of code output by “vibe” coding solutions.
  • As we create more code quickly, we need a better way of understanding code quality in the context of our organization and doing agentic code review and agent testing.
  • AI tools work best with well-organized code bases and well-described data sources.
  • AI enhancements to coding practices need to account for the entire SDLC life cycle.
  • New engineering practices are emerging as we accelerate adoption of AI engineering techniques.

Best line heard: “What is good for AI Agents is also good for humans.”
This was in the context of organization and quality of code/data repositories that make it easier for AI — also make it easier for humans.

Relevant Presentations

Asaf Bord

https://www.youtube.com/watch?v=cMSprbJ95jg&t=18184s

Probably the most relevant talk for our work at Pfizer. Asaf Bord walked through a GenBI Transformation at Northwest Mutual.
The GenBI program has built an agent stack under governance and orchestration to translate business questions into accurate data-backed answers, emphasizing successful outcomes from enriching metadata.

Jake Nations – Netflix

https://www.youtube.com/watch?v=xmbSQz-PNMM&t=21879s

One of my favorite talks — revisiting Fred Brooks’ Mythical Man Month and the “There is no Silver Bullet” topic.
Very interesting: the tools are so much better now, but you still need to know your business, and AI can’t do the thinking.

Kitze

https://www.youtube.com/watch?v=xmbSQz-PNMM&t=11964s

A hilarious talk on the future of front end, vibe coding, and MCP.
Kitze also explored Cursor’s Composer One — definitely worth checking out if you’re following VIBE coding trends.

Dan Shipper – Every

https://www.youtube.com/watch?v=cMSprbJ95jg&t=28341s

AIE CODE 2025: AI Leadership ft Anthropic, OpenAI, McKinsey, Bloomberg, Google Deepmind, and Tenex
Dan Shipper highlighted that converting an organization’s tacit knowledge (bugs, plans) into an explicit library of prompts and sub-agents is the core strategy for building AI-native companies.
This enables “compounding engineering,” allowing a single engineer to manage complex production products, resulting in a 10x operational effectiveness difference when AI tools are adopted universally.

Kevin Hou – Product Lead Anti-Gravity

https://www.youtube.com/watch?v=xmbSQz-PNMM&t=29388s

Anti-gravity is excellent. I have been playing personally – it takes vibe coding to another level.

Anti-gravity is a new agent-first Integrated Development Environment (IDE) that leverages the multimodal capabilities of Gemini 3 Pro. The platform’s dynamic artifact primitive allows the agent to generate visual representations (like plans) instead of endless strings of tokens, and it enables human-in-the-loop interaction without interrupting the agent’s execution loop.

Martin Harrysson & Natasha Maniar – McKinsey - AI Native Operating Models

https://www.youtube.com/watch?v=cMSprbJ95jg&t=8363s AIE CODE 2025: AI Leadership ft Anthropic, OpenAI, McKinsey, Bloomberg, Google Deepmind, and Tenex

Focused on how AI agents are reshaping software delivery through changes to the people and operating models within large enterprises. Most enterprises are stuck in the past with 8-10 person teams.