Connecting Your Laptop to Present �Google Slides

If you plan to use speaker notes, please follow the steps below to ensure your laptop is set up correctly. This will allow you to view your notes while the audience sees only your slides.

Before Connecting

• Close all unnecessary apps & tabs
• Turn off notifications
• Disable Night Mode / TrueTone
• Don’t enter full-screen presentation mode yet

Using MAC

• Go to System Settings > Displays
• Click the “Arrangement” tab
• Make sure “Mirror Displays” is unchecked

At the podium

• Plug in the HDMI cable
• If needed, use your USB-C to HDMI adapter
• On Mac, click “Allow” if prompted

Using PC

• Press ⊞Win + P, then select “Extend”
• Or: Right-click on the Desktop > Display Settings
• Under Multiple Displays, select “Extend these displays”

USE “EXTENDED DISPLAY” MODE (NOT MIRRORED)

Connecting Your Laptop to Present �Google Slides

When Using PowerPoint

• Start slideshow: ⌘Cmd + Enter (Mac) or F5 (PC)
• Or: Click Slide Show > Play from Start
• If notes show on the wrong screen, click “Swap Displays”
• https://www.youtube.com/watch?v=gQ3D4m-5pww

Google Slides

• Use Google Chrome as your browser
• Click the dropdown next to ‘Slideshow’ > select Presentation Display Options
• Check “Presenter View” and “Full Screen”
• If prompted, select "Allow"
• Choose the external display (may have a strange name like PanasonicXYZ)
• Click Start Slideshow
• https://www.youtube.com/watch?v=-GT7WCvPcys

Rickey Zachary, Thoughtworks

The Next Developer Portal: Emerging Patterns for Building AI Native Developer Experiences

Rickey Zachary�Global Lead Platform Engineering�Thoughtworks

Title

Are AI tools missing the mark?

• Disconnected AI tools create silos - Teams independently select solutions without coordination, leading to integration gaps and overlapping functionality.

Tool Fragmentation

• Developers struggle with underutilized tools due to poor integration, steep learning curves, inconsistent adoption, and diminishing usage.

Limited Adoption Patterns

• Disjointed AI tools lead to inconsistent experiences, manual data transfers, and limited insight sharing across the SDLC.

Workflow Disruption

• Managing multiple disconnected tools creates challenges with standardization, visibility, governance, and cost efficiency across the enterprise.

Scaling Challenges

• Individual tool investments deliver isolated, task-specific value that's difficult to measure across teams and the entire SDLC.

ROI Limitations

We are seeing more and more organizations not seeing the benefits of their AI investments.

​

Many are just tools for developers to use.

Engineering platform

A technical product or platform which allows a product team member to function more effectively resulting into business agility with faster time to market

Traditional Dev Portal

​

Developer Interfaces

Visual Studio Code (IDE)

Backstage

(Dev Portal)

Agentic CLIs

Custom AI Interfacts

The new Model Context Protocol (MCP) allows LLMs direct access to tools, services, documentation, or any other internal information. This allows for action to be take including updating or adding data. Other benefits include:

​

Runtime integration framework: connects LLMs to live production systems

Unified API gateway: enables controlled access to operational capabilities

Fine-grained permissions model: secures AI action execution

Real-time data connectors: ensure access to current information

Observability hooks: track AI-initiated actions for oversight

Versioned tool interfaces maintain backward compatibility

Event-driven architecture: supports asynchronous AI workflows

Semantic caching: optimizes repeated information retrieval

Audit trails: ensure accountability for AI-driven decisions

Automated guardrails: prevent unintended system modifications

​

Context based LLM

UI Experience

Data Catalog

Service Catalog

CLI / SDK

Docs as Code

API

Domains

Code

Governance & Access Layer

Metrics

Knowledge Management Layer

Self Service Layer

Tools

Infra

Services

Through qualitative surveys, experience based feedback, quantitative data, find ways to identify waste and friction across the ecosystem… and remove it.

Developer

Code

Commit

IaC Application Blueprint

Feedback

Deploy

Developer

Once

Test

Environment

Testing

Old

Environment

New

Environment

End User

Ephemeral Environment

Switch environments on successful tests

• Goals
• Actions
• Domain Knowledge

Environment

Agent

Agent

• Goals
• Actions
• Domain Knowledge

Sensing

Effecting

Collaborating or Competing

Ⓐ Purposeful Action with Knowledge�An AI Agent utilizes domain knowledge to achieve specific goals

Ⓑ Environmental Awareness and Collaboration �An AI Agent is aware of its environment and is ready to collaborate with humans or other AI agents.

Ⓒ System Interaction

An AI Agent interacts with systems for function calling or serving.

Collaborating

Directing

Serving

• Systems
• Services
• UI

Ⓐ

Ⓑ

Ⓒ

What is an agent?

So… What is an agent made of ?

​

01

03

04

�Instructions

Prompts that define role of the agent with necessary guardrails, in natural language.�

LLM

Model to fit the purpose.

Tools

Window to the outside world.

​

02

​

​

Memory

Short term memory to retain and share context�Long term memory to reinforce learning from past actions and outcomes

​

​

Agent in the catalog

Generating custom kinds to support new agentic components in backstage.

​

New Kinds:

Agents

MCPs

LLMs

RAGs

​

Each new agentic component becomes a kind for metadata to be attached so that it can be orchestrated as a template or a new catalog item.

Old kinds can be augmented with new agentic capabilities.

For Engineers

1. Master the Fundamentals
2. Embrace AI as a Partner
3. Build Platform Thinking

For Technical Leaders

1. Create the Environment
2. Foster the Culture
3. Build the Coalition

For Product Leaders

1. Start with Problems
2. Design for AI
3. Iterate Rapidly

For Business Leaders

1. Invest in Foundations
2. Measure Right Things
3. Communicate Vision