#artificial intelligence

I have been running an AI second brain as a developer for the past year, and it saves me at least 20 hours every week. It knows my repos and the conventions in each one, dispatches coding work to run in the background, drafts my email replies, and pings me on Slack when something needs my attention. In this talk I will show you how to build one in TypeScript.

We will walk through the four parts that make it work. The memory is just markdown files that the agent writes to and curates itself. Hybrid search runs over that memory so the agent pulls the right context at the right time. A scheduled loop makes it proactive, reviewing what happened yesterday, promoting what mattered into long term memory, and surfacing what needs me today. And an integration layer connects all of it to my repositories, my inbox, and the rest of the platforms I actually work in.

You will leave with the full architecture and a clear path to building your own.

Wes discusses the evolving ways of interacting with software, from UI to voice commands and tailored interfaces. He explores generative UI, explaining its three levels: basic, declarative, and fully generative. Discussions on finding a middle ground between known components and fully generative approaches. Discussing challenges with current MCP interfaces and the need for adaptive and reactive interfaces, using smart home examples to illustrate preferences for UI interactions and personalized automation. Exploring the versatility of WebMCP in switching between manual and agent-controlled interactions for enhanced user experiences. Discussing staying hopeful in front-end development, embracing new technologies, and the concept of dynamic UI for user-driven component selection. Discussing the importance of preserving the openness of the web, advocating against centralized control by a few companies, and the risks associated with MCPs in user consent and security.

The Talk introduces APIs, Codemode, and MCP, highlighting the role of APIs in empowering agents and AI's access to external resources. CloudFlare's evolution with MCP involves remote tool sharing and addressing challenges like large context windows. Challenges in Cloudflare's API tool access include progressive tool disclosure and splitting MCP servers. The discussion covers tool search, project management API creation, CodeMode concept, and self-documenting tools for secure tool execution. Proposed solutions include Code Mode, SDK generation, efficient control flow, token efficiency, and secure code execution. Key topics include dynamic worker loaders, Cloudflare API integration, Visitor Counter implementation, and exploring MCP server functionality.

JavaScript developers face security risks in npm packages with self-replicating worms compromising packages. Supply chain attacks target prolific maintainers leading to unauthorized code injections. Vulnerabilities in package maintainers and implicit trust contribute to successful attacks. Secure package installation practices vary between npm and Deno. AI-generated code introduces new security risks with embedded secrets. Secure AI code execution involves sandboxing for isolation and secure handling of API keys.

The Talk delves into the impact of LLMs on engineering workflows, emphasizing the shift towards AI for code writing efficiency. It highlights the importance of problem-solving abilities and product development in hiring practices. LLMs enhance productivity for developers, designers, and product managers, enabling faster iterations and collaboration. Efficient product workflows are achieved through simultaneous UI and code work using Builder.io. Task delegation is streamlined with async agents and Builder bots, optimizing project editing and code management. Integration with tools like VS Code and Figma streamlines design and code editing, enhancing collaboration and workflow automation.

Software engineering is evolving with AI and VIBE coding reshaping work, emphasizing collaboration and embracing AI. The future roadmap includes transitioning from augmented to AI-first and eventually AI-native developer experiences. AI integration in coding practices shapes a collaborative future, with tools evolving for startups and enterprises. AI tools aid in design, coding, and testing, offering varied assistance. Context relevance, spec-driven development, human review, and AI implementation challenges are key focus areas. AI boosts productivity but faces verification challenges, necessitating human oversight. The impact of AI on code reviews, talent development, and problem-solving evolution in coding practices is significant.

The Talk delves into motion control with multimodal AI, exploring TensorFlow.js models for gesture recognition and enhancing user interactions. It discusses leveraging LLMs for gesture-based interaction, investigating Gemiini for gesture recognition, and controlling light states with Gemini functions. The conversation includes webcam-based gesture recognition, custom gesture databases, and the future of personalized AI assistance with acoustic recognition.

AI is transforming software engineering by using agents to help with coding. Agents can autonomously complete tasks and make decisions based on data. Collaborative AI and automation are opening new possibilities in code generation. Bolt is a powerful tool for troubleshooting, bug fixing, and authentication. Code generation tools like Copilot and Cursor provide support for selecting models and codebase awareness. Cline is a useful extension for website inspection and testing. Guidelines for coding with agents include defining requirements, choosing the right model, and frequent testing. Clear and concise instructions are crucial in AI-generated code. Experienced engineers are still necessary in understanding architecture and problem-solving. Energy consumption insights and sustainability are discussed in the Talk.

Today we will discuss React server components with AI and how to build a better search experience using them. We will learn how to make a Next.js app AI-enabled using the Vercel AI SDK. The Vercel AI SDK's streamUI function with the GPT 4.0 model will be used to make suggestions interactive. We will explore the use of history and conversation in AI and how to continue the conversation and read the result. The concept of generative UI with the vector database will be introduced, along with querying the database for movies. We will process user queries and return movies based on them. The power of React server components in enhancing UI will be demonstrated. In summary, the Talk covers vector embeddings, natural language search, and generative UI.

AI design challenges include bias, safety, and security. Trust and transparency are important in AI. Design principles for AI include user control, fighting bias, and promoting good decision-making. AI can enable the design process and investors expect to see it included in products. AI empowers individuals to create and share ideas, but managing expectations is crucial.

In this Talk, the speaker demonstrates how to create an AI chat bot that can answer questions based on information it was never trained on. They build a basic RAG pipeline in just five minutes using live coding. The speaker also shows how to create embeddings and a vector database, set up a vector search index and endpoint, and modify the chat route to enhance the chat bot's capabilities. The program is run and tested, and the Talk concludes with an invitation to join a workshop for more information.

AI in game development has evolved rapidly, with generative AI being a focus. However, game developers like Romero Games have concerns about ethics and prefer using AI to automate processes and make creative work easier. AI has been used in games for decades, from path-finding AI to decision trees. Procedural world building and advanced AI technology are pushing the boundaries of FPS games. Different teams within a company have different approaches to the use of AI, depending on their specific needs and requirements.

Web AI in JavaScript allows for running machine learning models client-side in a web browser, offering advantages such as privacy, offline capabilities, low latency, and cost savings. Various AI models can be used for tasks like background blur, text toxicity detection, 3D data extraction, face mesh recognition, hand tracking, pose detection, and body segmentation. JavaScript libraries like MediaPipe LLM inference API and Visual Blocks facilitate the use of AI models. Web AI is in its early stages but has the potential to revolutionize web experiences and improve accessibility.

Codium is a generative AI assistant for software development that offers code explanation, test generation, and collaboration features. It can generate tests for a GraphQL API in VS Code, improve code coverage, and even document tests. Codium allows analyzing specific code lines, generating tests based on existing ones, and answering code-related questions. It can also provide suggestions for code improvement, help with code refactoring, and assist with writing commit messages.

Join Christoph from Nakazawa Tech in building the AI for Athena Crisis, a game where the AI performs actions just like a player. Learn about the importance of abstractions, primitives, and search algorithms in building an AI for a video game. Explore the architecture of Athena Crisis, which uses immutable persistent data structures and optimistic updates. Discover how to implement AI behaviors and create a class for the AI. Find out how to analyze units, assign weights, and prioritize actions based on the game state. Consider the next steps in building the AI and explore the possibility of building an AI for a real-time strategy game.

This Talk discusses building a voice-activated AI assistant using web APIs and JavaScript. It covers using the Web Speech API for speech recognition and the speech synthesis API for text to speech. The speaker demonstrates how to communicate with the Open AI API and handle the response. The Talk also explores enabling speech recognition and addressing the user. The speaker concludes by mentioning the possibility of creating a product out of the project and using Tauri for native desktop-like experiences.

This talk explores the use of AI in web development, including tools like GitHub Copilot and Fig for CLI commands. AI can generate boilerplate code, provide context-aware solutions, and generate dummy data. It can also assist with CSS selectors and regexes, and be integrated into applications. AI is used to enhance the podcast experience by transcribing episodes and providing JSON data. The talk also discusses formatting AI output, crafting requests, and analyzing embeddings for similarity.

What does it take to build a React UI that talks to an AI agent, streams responses in three languages, and runs on both WhatsApp and the web — in production? In this case study, I'll share hard-won lessons from shipping a multilingual disaster risk awareness platform for Sri Lanka, covering real-time streaming UIs, auth-aware chat components, and the React patterns that make AI agents feel alive rather than laggy and broken.

AI, React, Vue - what do all of these things have in common? They are tools meant to enhance the development experience - we must remember that. Join me in going over best practices on using AI to LEARN and not to do you your job for you. I’ll take you through the steps of how I used AI as a React developer to learn Vue - did I actually learn? Join this talk to find out!

What happens when AI tools meet the realities of a 130-year-old organisation with legacy systems, complex processes, and high expectations for reliability? In this talk, I’ll share practical lessons from introducing AI-assisted software development in a traditional environment: where it genuinely accelerates delivery, where it fails, and how teams can use it responsibly. Expect real examples, workflows, and honest insights on using AI beyond the hype.

AI is writing a growing share of my production code, and it fails in ways my old testing habits were never built to catch. This is the honest field report: what breaks, what I test for now, and the checks that actually catch it before it ships.

Telling your AI coding agent "please don't commit directly to main" is about as effective as asking your cat not to knock things off the table. LLMs are non-deterministic by nature - a polite prompt instruction is not a security boundary.
In this session, you'll learn how to enforce what you can't simply ask for. We'll dive into lifecycle hooks that intercept tool calls before they execute, blocking unauthorized Git commits, file deletions, or rogue shell commands at the architectural level. We'll also explore how to run coding agents autonomously and safely inside sandboxes - so your agent can do real work without being a liability.

Whether you're building agentic workflows or just trying to stop your AI colleague from pushing directly to production: this talk gives you the controls that prompts never will.

AI agents have gotten remarkably good at thinking and remarkably bad at showing. Most agentic products still respond with paragraphs and markdown tables.

This talk covers generative UI, letting the agent produce the interface best suited to the moment. We'll look at the emerging protocol layer, including Google's A2 UI, and walk through a working example. You'll leave knowing where generative UI fits in your stack and how to ship a first version without rewriting your front end. 

When an enterprise JavaScript monorepo or merge queue experiences flaky test runs or deployment pipeline failures, developers lose hours debugging large logs across multi-package builds. I will give the audience a practical blueprint for an edge-triggered AI incident responder that acts as an ephemeral agent inside a self-hosted CI job. You will learn how my team built a stateless system using Claude Code that automatically spins up when merge train health thresholds dip, conducting "differential diagnosis" against historical test logs to catch elusive semantic conflicts that local passing branches missed. We’ll cover critical production guardrails: two-layer deduplication to prevent alert spam, runner timeout architectures, and minimizing trust boundaries using existing CI secrets to prevent modern prompt-injection attacks.

WebLLM brings local AI inference to the browser using WebGPU, WebAssembly, MLC-LLM, and Apache TVM. As WebLLM has grown to support more capable models and applications, many of the important problems now sit at the boundaries around model execution: loading large artifacts efficiently; coordinating work across JavaScript, WebAssembly, and WebGPU; supporting different model architectures and device capabilities; connecting model outputs to application logic; and preserving state within an ephemeral browser environment. This talk will cover recent work in WebLLM across these boundaries. We will cover cross-layer profiling and runtime optimizations that go beyond model kernels to reduce memory use, synchronization, and token-generation overhead across the full execution path. We will also discuss explicit compiler-runtime interfaces and model manifests, which define how compiled models expose their capabilities and requirements and enable a common path for multimodal and hybrid models. At the application boundary, WebLLM integrates XGrammar and structural tags to constrain generated outputs into structured data and tool calls. Finally, because a page reload or worker termination can otherwise discard an ongoing generation, WebLLM extends browser-local persistence from cached model artifacts to active inference, using OPFS-backed token journals and asynchronous KV-cache checkpoints to restore and continue interrupted generation.

AI-generated code speeds up development—but it also accelerates security mistakes. When developers trust generated output without proper review, vulnerabilities slip into production silently. This talk explores real security pitfalls introduced by AI-assisted coding, explains why traditional review habits break down, and shows how attackers exploit these blind spots. You’ll learn how to use AI productively without surrendering security responsibility.

State management has traditionally focused on synchronizing UI interactions, frontend stores, and backend persistence. But AI agents introduce a new type of actor into modern applications: autonomous workflows that can read, execute, and modify application state alongside human users.

In this session, we explore how frontend state management concepts can evolve into shared full-stack architectures where both humans and AI agents collaborate on the same data model. Using Angular-based application examples, we will examine sandboxed agent execution with forked state snapshots, deterministic merge strategies, concurrent updates between users and agents, and observable agent actions.

Rather than treating AI as an external API call, this talk presents agents as first-class participants in application state transitions. We will extend familiar concepts such as stores, signals, reducers, and event streams beyond the frontend into backend orchestration systems that keep human and agent state consistent.

Attendees will leave with practical architectural patterns for building collaborative human + AI applications using ideas FE developers already know.

What if your React app could understand users without sending their data anywhere?Modern browsers can now run meaningful machine learning locally. This talk shows how to use in-browser ML to build React interfaces that personalize, adapt, and respond in real time, while staying fast, private, and predictable in production.

Generative UI is currently one of the most explosive frontiers in frontend engineering. Modern UI frameworks like React are moving from static components to streaming, AI-generated micro-frontends.Andrey's talk explicitly references Google's A2UI protocol, which standardizes how agents emit declarative layout schemas that are safely interpreted and rendered by trusted client components. React component architecture, server actions, and client-side design patterns are exactly where this protocol must be implemented. Senior React developers are looking for ways to move past boring, messy chat-bubble boxes and build structural, design-compliant generative components.

We have been supporting builders of AI products with over a dozen of features in TanStack AI. Today I want to show you how most AI workflows work under the hood, what it takes to run a harness, what is important after the initial implementation of a chat experience and how to grow your functionalities and scale to the needs of your users, and most importantly, how TanStack AI enables you to do all of this with ease and joy.

Agentic development feels like magic to some teams and overhyped to others. Harnesses like Claude Code or Codex have been built around the local loop, the way things have always been done: a single developer changing code before opening a PR. Inside real organizations, agents hit walls: repo boundaries that demand human attention, a lack of episodic memory across sessions and unresolved questions around trust and permissions.Fixing this requires a higher level of abstraction: the *meta-harness*. Just as meta-frameworks like Next.js or Remix let React stay simple while giving developers what they need to ship, a meta-harness keeps the agent CLI focused on its job while bringing the components teams need to be effective: cross-repo orchestration, testing, and change tracking for greater autonomy,session memory that understands decisions and extracts tribal knowledge,collaboration features for easier handoffs, reviews, shareable sessions andsecurity primitives like sandboxes, remote execution, and clearly scoped agent permissions.This talk argues that the meta-harness - a layer above any single lab or agent - is the missing piece for integrating agents into how we build software across the stack.

State management has traditionally focused on synchronizing user interactions, frontend state, and backend persistence. But AI agents introduce a new type of actor into modern applications: autonomous workflows that can read, execute, and modify application state alongside human users.In this session, we explore how familiar frontend state management concepts can evolve into shared full-stack architectures where both humans and AI agents collaborate on the same data model. Using examples from modern React applications, we will examine sandboxed agent execution with forked state snapshots, deterministic merge strategies, concurrent updates between users and agents, and techniques for making agent actions observable and debuggable.Rather than treating AI as an external API call, this talk presents agents as first-class participants in application state transitions. We will extend familiar concepts such as state stores, reducers, event streams, and optimistic updates beyond the frontend into backend orchestration systems that keep human and agent state consistent.Attendees will leave with practical architectural patterns for building collaborative human + AI applications using ideas they already know from modern frontend development.

How do you know an AI system will refuse the thing it must always refuse? Today the answer is testing: ask it a thousand ways and hope the thousand-and-first isn't different. The policy in this talk has 55 variables a caller controls: that's 36 quadrillion situations! It also has a test suite of 33 unit tests, all green. This talk is about deciding the question instead of sampling it. A safety rule like "never do X without checking Y" becomes a logic circuit, and a circuit becomes arithmetic: equations whose only solutions are the situations the rule allows. Asking whether the rule can be broken turns into asking whether that system has any solution. Live: one rule proved unbreakable, with a short record of why that you can check yourself without trusting the program that found it; and two that break, each repaired and re-proved on the spot. This proves the rail, not the model.

AI assistants now write a growing share of the code your team ships, but most review processes and quality gates were designed for human authors. Reviewing more code faster isn't a strategy. In this lightning talk, Albert Franquesa introduces Control, Evidence, Trust, a simple model for leading teams through this shift: enforce your standards before code is written, measure what AI actually contributes and what it gets wrong, and use that evidence to decide where agents can safely do more and where humans must stay in the loop. Trust in AI code isn't a feeling. It's earned with control and proven with evidence.

How do you deliver with AI and still sleep at night? By engineering trust into your delivery system.
In this talk, I'll share an end-to-end AI engineering workflow we built and ran on a production codebase, redefining roles, responsibilities, quality gates, and human decision points across the entire delivery lifecycle.
You'll leave with a practical framework for engineering AI workflows where reliability is designed in, trust is earned, and success becomes the default outcome, not a fortunate accident.

AI makes individual engineers faster - and quietly more isolated.
As coding agents become everyone's default pair, the collaboration that keeps an org healthy erodes: pairing fades, context gets trapped in private chat histories, knowledge stops diffusing, teams drift into silos.
Velocity dashboards look great, so nothing seems wrong - until onboarding slows and bus-factor spikes.
This talk names the failure mode, explains why standard metrics miss it, and shows what leaders can do to keep teams connected as AI adoption accelerates.

Teams often start agent adoption by adding tools: MCP servers, API wrappers, chat interfaces in developer portals. But once coding agents move from local help into real delivery workflows, they usually fail earlier than the tool call. They read stale ownership, conflicting runbooks, missing deploy history, and policies that were written for humans.

This talk introduces company memory as the missing layer for production coding agents: trustworthy, permission-aware engineering context that agents can read, cite, and act on safely. We’ll look at what belongs in that memory, how to expose it through capability registries and machine-readable tool contracts, how to use on-behalf-of identity instead of shared AI service accounts, and how to grow trust from read → recommend → act.

The goal is simple: before giving agents more tools, make sure they can trust what they read.

Frontend teams are increasingly using AI to speed up UI development by generating code directly from designs. While the results can look impressive, they often fall short in real production environments: incorrect components, missing accessibility, and broken design-system rules are common failure modes. In this talk, I’ll present a real-world Figma-to-Code case study that treats AI as part of the frontend infrastructure rather than a code generator. By teaching AI how a design system works and orchestrating a structured workflow, we moved from “cool demos” to production-ready UI code - an approach we now rely on in real product development, saving significant time while improving consistency, quality, and user experience.

AI is giving developers unprecedented autonomy. Every engineer can now create their own workflow, combining coding agents, prompts, knowledge bases, and tools in different ways. While this can dramatically increase individual productivity, it introduces a new challenge for engineering leaders: maintaining consistency across teams.

In this talk, we'll explore how AI is creating new forms of organizational drift, why team alignment matters more than ever, and practical approaches for building shared AI practices without limiting innovation.

Presentation on building a customer support bot with Langraff, Amazon Bedrock as base. LangChain Python library for LLMs. RAG - Retrieval Augmented Generation for accessing current information. LangChain using Bedrock and Amazon Knowledge Base. LangRaph classifies queries intelligently. Using LandGraph on top of LangChain for multiple scenarios with human intervention. Connected to Amazon Bedrock for knowledge base. No guessing with large language models; rely on a connected knowledge base for accurate responses. Chat GPT creation involves memory for context; tools like Land Graph aid in this process. Land Graph usage for flow creation; Demo with Kiro agent ID and new file uploads. Setting up AWS MCP server with agent skills; Using AWS agent toolkit for best practices in service setup; Flow for intent routing and order refund eligibility. Handling Order Refund Requests Using Agent Tools and Lookup; Case Sensitivity in Order Number Search; Validating Order Eligibility for Refund. Uploading Knowledge Base to AWS MCP Server; Verification of Email Address for Refund; Human in the Loop for Refund Approval. Human in the Loop Feature of LaneGraph; Uploading Documents to Bedrock; Utilizing Vectors for Knowledgebase Queries. Utilizing AWS Agent Toolkit for Workflow Best Practices; Connecting to KnowledgeBases for Querying; Building App with Mastering LaneGraph. Generating Design and Requirements for App Development; Utilizing Agent Skills in Spec-Driven Workflow; Ensuring Accuracy with Mastering Langraff Skill. Detailing Setup for Coding Assistant with AWS Agent Toolkit; Requesting Feedback and Interaction; Importance of Langraff in Graph-based Solutions. Enhancing Chatbot Prompt Suggestions Dynamically based on Conversation Context; Utilizing In-Memory Thread for Follow-up Questions in Langraff; Swapping Question Prompts with Tags for User-friendly Experience. Utilizing Different Models for Chatbot Tasks; Implementing Security Measures; Continuous Knowledge Base Updates for Accuracy.

Exploring AI engineering and agent-assisted development through Claude's Code on GitHub. Analysis of Cloud Code commits, repository scale, and language distribution. The impact of agents on coding in various languages like Python, TypeScript, Rust, Dart, and Zig, facilitating easier coding in traditionally more complex languages. The significance of stars on repositories in Cloud Code projects and the observation of growth patterns post-Christmas season. The impact of Christmas on exponential growth in commits per day, the current doubling time trend, and a focused analysis on the top 100 popular repositories in Cloud Code projects. Highlighting key differences in commits between AI and non-AI repositories, focusing on feature contributions, bug fixes, and UI changes. Exploring diverse use cases of Cloud Code, highlighting how Clickhouse utilizes extensive Cloud Code for optimization, CI testing, and infrastructure, fostering engineers' focus on more impactful tasks like feature development and bug fixes, along with the emerging trend of Cloud Skill repositories. The growing trend of active projects utilizing AI contributions, specifically Cloud Code's significant impact on real, critical software projects beyond AI-focused repositories. Exploring models and coding agents that self-identify like Cloud Code, particularly the aggressive nature of Cloud in generating code, as observed with GitHub Copilot and other models like Mistral. Projects adopting AI development observe an increase in GitHub issues due to code production, leading to functional issues. The popularity growth of Clickhouse may be linked to active use of Claude to accelerate development, not directly boosting popularity. Metrics tracking revealed exponential growth in the use of short codes, focusing on valuable contributions to critical code with continued usage.

Bernie emphasizes using AI to enhance rendering performance, with a focus on debugging and the evolution of computing speed standards. Rendering optimization challenges persist despite advancements in multi-threaded development. Debugging CSS and rendering performance complexities require tools like profilers for JavaScript efficiency analysis. AI automation can improve performance by understanding rendering engine intricacies. AI aids in identifying and resolving performance bugs, investigating rendering issues, and optimizing icon rendering. Tailwind's impact on performance and leveraging AI for CI checks showcase AI's role in troubleshooting and long-term learning in software development.

Daniel, a consultant from Germany, discusses AI-generated code and challenges in rapid code shipping. The Pharoes AI case study highlights code reviews and challenges in agile development. Incidence per pull request and code churn issues are discussed in software development. AI-assisted coding and quality gates are introduced. Developing with Angular and TypeScript emphasizes the importance of basics, AI sub-agents, and pull request review services. Utilizing hooks for deterministic code handling and enhancing code handling with various hooks are crucial. Transitioning to behavior-driven development, utilizing tools like PlayWrite and Jira integration, and evaluating code review strategies are key points. Optimizing code review practices, smart integration of AI and hooks, advanced code review models, ownership in code reviews, software metrics, and quality gates challenges are also covered.

The speaker delves into loop engineering's role in automating decisions and enhancing productivity in AI development workflows. Loop engineering is highlighted as a tool to build and automate tasks, even for creating programming languages like Kerst, with a focus on rapid project development. The importance of repetition in loops, automation criteria, and the suitability of loop engineering for full-blown apps are emphasized. Discussions include read-only tests, verifier and backlog agents, and the necessity of a clear definition of done for effective loop implementation.

Daniel Avila introduces Cloud Code Desktop, a new platform simplifying Cloud Code usage. Explore features like Cowork, connectors, and plugins. Manage sessions, customize environments, and interact with Cloud efficiently. Understand skills, context, and execution in Cloud Code. Develop comprehensive skills with precise control measures. Optimize skill execution, design interfaces, and implement skills in MCPs. Enhance MCPs with Cloud Skills and advanced installation processes in Cloud Code Desktop.

Nawal discusses using LLMs for migrating component libraries at Zalando, highlighting challenges and complexities. Discussing a recent migration problem at Zalando involving unification of component libraries and challenges in automating migration process. Encountering limitations with codemods, leading to exploration of LLMs for automated migration processes. Conducting iterative experiments to enhance LLM accuracy in component migration. Building a CLI tool for component migration. Using AIDAR for automating git commits & maintaining LLM accuracy. Recommendation for team closeness to code, successful complex transformations by LLM, challenges with hallucination and ignoring instructions. Lessons learned include field addition, non-deterministic outputs, slow responses, project constraints, and tool evaluation. Evaluating tool enhancements, PROMPT best practices, core approach, improved tools, LLMs for simple migrations, and tool limitations with multiple file changes.

Developers and QA engineer Wendy discuss the challenges in development and testing, using John's experience as an example. John encounters challenges with expired coupons affecting the checkout process and sprint deadlines. Developers and testers have complementary perspectives in software development; introducing the concept of QAI for enhanced quality assurance and collaboration. The QAI introduces a new perspective to code development, focusing on thorough alignment and error scanning for enhanced quality assurance. The importance of code alignment in development, the consequences of silent failures, and the necessity of handling potential hidden failures with a security-focused approach. AI builds fast and trusts everything, teaches suspicion, identifies trust gaps, and addresses accessibility failures for a secure code environment. Every team has patterns to anticipate, leveraging skills for immediate help, and focusing on effectiveness and protection in AI development. John and I collaborate using QAI to enhance code effectiveness, prioritize code alignment, fix contradictions, and address hidden failures effectively. Ship clean code timely, find bugs early to save time and trust. AI and QAI enhance speed and safety in development. Wendy emphasizes the importance of detecting bugs early and the impact on user trust and deadlines.

Daniel Avila discussing templates and components for Cloud Code, emphasizing the importance of configuration settings and skills integration in the development process. Cloud Code settings, commands, sub-agents, and skills integration play a crucial role in developing a complete template for AI agents. Skills injection, documentation, development template, Cloud Code components sharing, harness engineering, logs review, process improvement, configuration levels. Configuration levels, settings.json, hooks, verification, deterministic projects, templates, status lines, cloud.md, best practices. Explaining skill invocation control, model usage, and skill specialization for message crafting within commits. Emphasizing the importance of managing the context window effectively for skill injection and understanding the interaction with various agents and tools.

Voice interaction is evolving rapidly with AI agents moving beyond chat windows. Achieving natural conversations with voice AI presents significant challenges, especially in meeting real-time interaction expectations. Advancements in real-time conversation technology and infrastructure are improving latency gaps in voice apps. Optimizing latency at each stage of voice agent processing is crucial for efficient development. Semantic turn detection and efficient function calling are key to optimizing user experience in Voice AI. Workflow management, cost management, signal recognition, and observability are vital aspects in voice AI development.

Eric Hanchett discusses spec-driven development, emphasizing structured code creation with specifications before writing any code. The importance of context in guiding large language models and the evolving nature of software development is highlighted. Kiro, an AI IDE coding assistant, offers a comprehensive solution for improved spectrum development processes. Kiro's focus on vibe and spec mode caters to the need for working on larger and more complex projects. Spec-driven development in Kiro is beneficial for in-depth features and structured project building. The discussion also includes property-based tests, MVP creation, and the Model Context Protocol (MCP) integration within the spec-driven development flow.

Kitze introduces the purpose of conferences as networking, meeting people, and teaching. The speaker discusses their businesses Sizzy and Benji, emphasizing tools for developers and a life OS solution. They highlight Zero to Shift, Tinker Club, and SuperMac. React's evolution and frontend development advancements are summarized. Vibe coding and management practices are discussed, focusing on quick changes and embracing code nuances. Vibe engineering involves leveraging agents and committing often. Effective prompting, voice dictation, and vibe coding challenges are addressed. AI's impact on programming jobs and the importance of innovation are highlighted. Role-based frameworks like React Cowboys and Fibre engineering's code consistency are emphasized.

Gawe discusses React frameworks, tools, and agents, emphasizing speed and collaboration in building Z, a Rust-based code editor. Z differentiates itself with fast performance, real-time collaboration, and support for diverse platforms. The talk covers new frontend tools, code completion, and linting, highlighting Z's features for React developers. Z offers an open platform, custom theme creation, and deep integration with agents for improved workflows. The founders' journey from Adam to Rust showcases Zed's commitment to performance, open-source UI, and support features like MCP servers and global skill recognition.

Angel from ImageKit demonstrates the use of MCPs to improve image and video performance on React websites. The Chrome Developer Tools MCP helps identify and address performance issues by analyzing image loading speeds. MCPs automate issue resolution on websites, leveraging Chrome developer tools for analysis and Image Kit for optimizing images efficiently.

Starting with recent comments criticizing the stance on AI and coding, highlighting the divide between senior and junior developers' perspectives. Reflecting on the impact of AI on teaching code and content creation, transitioning to AI-driven teaching methods. Challenges in transitioning to AI-driven teaching focused on critical thinking over syntax. Discussion on scaling AI tools, Bloom's Taxonomy, and the shift to creation-focused development. Importance of critical thinking in software engineering and nurturing junior talents in the evolving industry. AI's role as a teaching tool, concerns about job replacement, and the need for continuous skill growth. The significance of maintaining critique ability in AI output evaluation.

The speaker discusses the evolution of AI agent workflow in mobile development, emphasizing the importance of feedback loops and verification. They highlight the challenges of human involvement as agents dominate code production. Agent Device simplifies mobile automation by focusing on accessibility trees and text-based actions. The tool enhances app verification, debugging capabilities, and provides advanced features like network issue identification. Agent Device is versatile for CI setup, feedback loops, and visual regression detection. It supports real devices and offers capabilities for migrations and automation. Additionally, the speaker compares flakiness and TV platform support, showcasing performance improvements over traditional tools.

The Talk delves into the exploration of Agents on the Canvas, integrating LLMs with visual elements and spatial relations. It covers the development of the Agent Starter Kit for navigating Canvas tasks and introduces fairies as collaborative agents. The discussion includes managing tasks, utilizing Cloud Code for project collaboration, creating interactive UIs, and customizing UI components. The evolution of agentic canvas code, agent interactions, user actions, token usage, and data handling with LLMs are highlighted, emphasizing experimentation and optimization for SDK integration.

The speaker delves into code editor development, emphasizing collaboration between humans and AI. Discussions revolve around adapting to technological changes while facing persistent cognitive limitations. Balancing cognitive load in software development is crucial for optimal task completion. Focus and alignment in software development are essential for effective problem-solving. Addressing challenges of team alignment in AI-driven environments is crucial to avoid creating legacy code bases. Pair programming enhances collaboration, accountability, and learning within development teams. Valuing collaboration, trust, and autonomy fosters speed and efficiency in software development. Leadership strategies focus on promoting autonomy, mastery, and purpose while addressing burnout. AI impact on productivity and collaborative coding practices are reflected upon, emphasizing the benefits of pair programming. Effective onboarding and encouraging pair programming adoption contribute to better problem-solving and team collaboration.

Shifra, founding developer relations engineer at Render. Render is the cloud for builders. How to relate to users who are not people? Company growth challenges with AI recommendations affecting signups. The challenges of AI recommendations in contrast to traditional SEO. Impact on team operations and product development. Need for a strategic shift towards agent-centered developer experience. Developing interface design for agents, content portfolio importance, and human gate validation. The evolving role of agents in product consumption and the necessity for a fundamental shift in development focus. Facing challenges head-on, emphasizing agentic experience, and prioritizing system self-correction for productive agent and human interactions at Render.

Richard Rodenkemper, senior software engineer at Sentry, discusses ensuring quality with AI. GitHub data shows exponential growth in coding. Concerns arise about the reliability of coding agents versus human engineers. Impact of AI and Cloud on code production and app quality is questioned. Challenges in code reliability despite increased production are highlighted. AI as a quality tool in software development. Importance of reliability for product success highlighted. AI's strengths in handling data and searching code base discussed. Examples of AI usage in code reviews and quality assurance at Sentry shared. AI efficiency in endpoint deprecation and system updates highlighted. AI's assistance in migrating design systems and reducing notifications using Cloud Code emphasized.

Zalim Tuzlak explains features of building 10 stack AI, enhancing capabilities with middleware and tool code mode, efficient AI tool script writing, challenges and breakthroughs in AI development, building scalable AI tools with activity-based tree shaking, early AI concepts and code quality strategies, empowering rapid AI iteration, TensTech AI beta release with Angular support, AI model comparison and evaluation, AI SDK features enhancement, quality assurance processes for TanStack AI docs, personal stress and decision fatigue in AI development, challenges in middleware development with AI assistance, skill creation process within the team, and TanStack AI's end goal and selling point.

Today's Talk delves into the Ralf Wiggum method for autonomous AI agents, emphasizing persistent looping based on specifications in AI coding. It addresses context management challenges in AI chats, promotes optimized coding workflows with feedback loops and engineer decision-making. Automation of specifications and task implementation is highlighted, alongside interactive conversation, prototyping, and managing code implementation. The importance of code quality enhancement, exploring new approaches with AI coding agents, and ensuring smart application of AI tools is discussed. Utilizing impactful skills and convincing leadership with results from the Ralph method are key points, along with enhancing task efficiency and overcoming tool limitations for efficient work.

The talk by Paolo Ricciuti introduces Svelte framework basics, syntax evolution in Svelte 5, challenges with LLMs in Svelte, Model Context Protocol (MCP) for LLMs, utilizing MCP for code verification, enhancements with Svelte MCP CLI, Svelte benchmark creation, integration of agent skills in node modules, Svelte adoption and DX differentiation, and Svelte autofixer and MCP linter.

We see how Monday used AI to break down a decade-old client-side monolith into micro-frontends, ensuring a complex software with long sessions and open tabs remained functional. The AI tool, Morphix, autonomously split prompts and validated tasks for consistent execution. The team focused on safe code migration, automated PR merging, and collaboration between AI and humans. They also optimized PR reviews, leveraged MorphX for monorepo restructuring, and managed AI costs effectively during the extraction process.

Ido Salomon, creator of MCPUI, discusses the need for interactive and branded user interfaces in chats and the adoption of MCPApps by companies. The development of an open protocol for interactive apps over MCP in partnership with OpenAI and Anthropic. Community-driven development of MCP apps focusing on transmitting UI over MCP for fully interactive applications. Real-time generation of interactive UI by models in MCP apps for enhanced user interaction. Personalized assembly of UI components by AI assistants based on user preferences. Evolution of MCP apps standards for standardized communication and enhanced interaction, leading to universal adoption. Innovative uses of MCP UI and apps in various creative and productivity applications with a focus on security and OAuth implementation. MCP apps impact on user engagement through unique HTML rendering and integration with Google Web MCP for sending parts of websites to chats.

Sasha discusses AI for web development, based on web dev AI survey results. AI's pervasive presence highlighted through personal experiences and industry trends. Increase in AI usage reflected in code generation and refactoring practices, impacting workflows and costs. Top pain points of AI tool usage include hallucinations, inaccuracies, and rising financial costs. Concerns about AI impact on job security, gender divide in environmental impact perception, and Moore's Law implications on rising AI costs. Speculation on faster, cheaper AI future, self-evolving code, and digital immune system. Software evolution towards probabilistic systems, personalized codebases, and digital DNA. AI potentially controlling digital lives, rendering software tools obsolete, and prompting critical thinking on future implications. Insights from AI survey results showcasing AI usage trends based on experience levels. Surprising impact of AI usage and philosophical views on AI terminology. Implications of AI advancement on coding skills, adaptation of software skills for the AI era, and concerns over AI-generated tech debt and social impacts. Speculation on AI tokens as global currency, ethical considerations, and impact on coding skills. Contemplation of AI future consequences and preparation for potential outcomes.

Discussion on effective thinking in the age of augmented tooling, user interface development, AI's impact on software engineering, and the importance of saving time. Importance of identifying high cognitive load tasks and introducing a framework for tech leads to guide teams with AI. Leveraging existing knowledge in the era of AI, transformative learning framework, and recognizing one's identity in the craft. Aang's transformative decision, reimagining roles for AI engineers, effective communication with AI, and revival of liberal arts in tech education. Startups' agility, clarity in logic, and continuous adaptation in engineering. Constant evolution, adapting to changing technology, and problem-solving focus in tech roles. Loop engineering concept, autonomous problem-solving loops, importance of defining problems, utilizing data effectively, and adapting to changes in technology. AI-driven code review, agent loop framework, embracing being wrong for learning, and example of tool Dev 3000. AI-driven automation for browser control and improvement, focusing on cumulative layout shift optimization. Introduction to Agent Browser for browser automation and layout shift optimization. Agent Browser for continuous improvement through verification, repetition, and learning. Balancing busyness and sustainability in work and life, challenges in work-life balance for creative technology jobs, and measuring creative project success. Challenges in code reviews, automation, managing agent loop costs, and optimization.

Ishai, CTO at LearnrB, discusses the evolution and impact of AI code reviews in software development. The transition from experimentation to maturity with autonomous AI agents in code reviews is highlighted. Challenges in AI-driven code creation and review processes, as well as the benefits of AI code reviews for productivity and efficiency, are discussed. Common mistakes and challenges in AI code review adoption, along with the importance of adapting processes for AI advancements, are emphasized. The importance of enhancing AI code review processes, measuring their impact, and leveraging AI for productivity metrics is also explored.

Shifra guides the migration of a Repl.it app to Render with emphasis on simplicity and mock data. Ensure Render deployment requirements are met, including specifying root directory, build and start commands, and correct port. Configure GitHub repository setup, connect to Git, and create a repository. Confirm code in the repository and deploy on Render, showcasing a sample landing page.

Welcome to spec-driven development talk by Daniel Sogel. AI agents struggle with benchmarks and real-world tasks. Developers spend extra time fixing code generated by AI agents. Security issues in AI-generated code, limitations with context management, multi-file coordination, silent failures, architecture decisions, and enterprise context challenges addressed by Spectre from development. AI agents removing safety checks, generating code with fake data, silent failures worse than crashes, spectrum development benefits like complete context upfront, structured reasoning, validation checkpoints, and autonomous implementation without manual code review. Act with AI agents like pair programmers, combine spectrum, behavior-driven, and test-driven development. Spec-driven development adoption in the industry, using Keyro IDE by AWS and popular tools like SpecKit and OpenSpec for different project types. Define tech stack, API structure in planning, task phase implementation, handover to AI agents, and choosing the right spec-driven development tools.

Carly discusses memory, context in LLMs, and managing information for effective results in agentic systems. LLMs hallucinate due to knowledge gaps, biases, overfitting, and incentives. Importance of context in minimizing hallucinations and focus on memory storage for decision-making. Short term memory acts like RAM, while long term memory stores knowledge and experiences. Efficient memory storage using disk, file systems, and semantic memory. Emphasis on managing message context, retrieval, and concatenation for LLM input. Attention to context length, information evaluation, and memory optimization for accurate responses.

Lugari Karasen from Red Hat AI discusses the technical aspects of large language models, including data collection, tokenization, and neural network internals. Model training involves converting human language to mathematical representations and adjusting parameters iteratively in a complex environment. Inference engines like VLLM aid in deploying models for rapid data processing. Optimizing model size for efficiency without sacrificing accuracy is crucial, with quantization reducing model size while maintaining precision. Local deployment offers privacy and control, and smaller purpose-driven models can enhance workflows and experimentation.

Introducing different AI interaction styles like chat interface vs. generating interactive interfaces for learning. Evolution of AI interfaces from text-based apps to tailored UIs based on user needs. Generative UI applications creating visual interfaces with AI. AI-driven UI component selection ensuring safety and flexibility. Challenges of true code generation for React or HTML. Balancing AI development approaches for innovation and reliability. Foundational aspects of AI SDK and generative UI with React components. Enhancing learning with voice-enabled interaction and expanding AI applications with voice UI. Voice-enabled workflow automation for various tasks. Utilizing production-ready AI technology for adaptive interfaces.

Hello everyone, React Native AI, AI SDK from Brazil, connect AI SDK with React Native, Szymon Rybczak, senior React Native developer, React Native AI, run AI on device, powerful devices, on-device AI on mobile, iPhone 15 Pro performance. Devices are powerful. Inspired by Guillermo Rauch's tweet on Google Chrome's Gemini model. Researched topic, learned about providers in AISDK, create and plug providers for various APIs. Excitement for WWDC, experimenting with Apple Foundation Model in React Native, impressive performance. Apple investment in React Native, code sharing success with Vercel SDK and React Native AI Apple package, extensive features and community support. Integration with ASDK, expanding providers for on-device models, Android models support, real-life privacy-focused applications. Simulator compatibility on macOS 26 and Pixel 9 Pro, upcoming Android models, LLM offline training trade-offs, privacy and efficiency benefits. Image and video generation potential, Device model parameter size comparison, Apple's optimized on-device model performance, Implications for other AI apps like Gemini and Jacky P.T. On-device model implementation for offline scenarios, Support for multiple runtimes, Memory consumption comparison, Whisper.RN and LLM.RN functionality, Device compatibility verification via APIs.

Sharing insights on scaling AI coding assistance & adopting AI native dev tools. Discussing the need for a shift from viewing adoption as a tech problem to a change management challenge. Exploring a tactical framework with three key phases: onboard, adopt, succeed. Emphasizing the importance of onboarding developers, ensuring access to licenses, and setting clear policies to avoid 'shadow AI'. Highlighting the necessity of human infrastructure for successful AI dev tool implementation, including AI advocates, communities of practice, and executive sponsorship. Adoption strategies require a focus on interconnected metrics and all pillars simultaneously. Leadership checklist for post-conference actions involves measuring metrics, distinguishing between strategic and tactical points, and staying connected through platforms like WhatsApp and LinkedIn.

Nico introducing himself and his work in machine learning and AI. Exploring the essence of Kit and its human-like features, and the use of Transform.js for machine learning in the browser. Understanding the significance of model parameters and the availability of open source models like Apertus and GEMMA3. Creating a fine-tuned version of the AI pipeline using prompt engineering with a new system prompt for improved behavior and responses. Utilizing Reddit to extract scenes, transforming screenplays into structured JSON, and creating an LLM summarizer for semantic similarity comparisons. Introduction to Agentic Rack System and Autonomous Task Completion by AA Agents. Utilizing TensorFlow and Transformer JS for Various Tasks and Models. Running LLMs in the browser is not practical due to large model sizes. Device performance matters in browser-based AI processing.