Table of contents:Please find the summarized notes from the DevFest2025 Amman event hosted by GDG Amman, featuring speakers from around the globe and packed with valuable insights. I’ve condensed the key points from the sessions I attended, and hopefully with the help of AI summarizing my notes while trying to understand the concepts worked out well!
For more Information:
Key Terms to Know Before You Dive In
Generative AI focuses on producing new content such as text, images, audio, or code by learning patterns from large datasets. It mainly responds to user prompts instead of independently taking action or making decisions. Its strength is creativity, content generation, and assisting with information or ideas.
AI agent systems are sophisticated programs that can take a single, complex request (a high-level goal) and autonomously break it down into multiple smaller, manageable steps. They then use different specialized tools and processes to execute each step sequentially or simultaneously, acting much like a highly organized team of specialists. These agents can monitor their progress, correct their own errors, and ultimately combine their work to deliver a complete, finished solution to the user, going far beyond a simple single-query answer. They represent a major step toward autonomous task completion in software.
Firebase is a set of cloud services and tools from Google that helps developers quickly build, manage, and scale mobile and web applications. It provides features like real-time databases, authentication, hosting, and analytics, allowing developers to focus on the frontend user experience rather than building complex backend infrastructure.
Vibe coding is a casual term for coding using AI mainly, while listening to music or in an environment that puts the developer in a focused, relaxed, and productive mental state (the "vibe"). It emphasizes flow state and personal comfort over strict development rigidity.
Human-in-the-Loop (HITL) in the context of advanced AI like Gemini's Deep Search means a human actively provides feedback, corrections, or validation during the AI's continuous workflow.
Architecting Production-Grade AI Solutions on Google Cloud
Google Antigravity is presented as an advanced, AI-driven development environment designed to help developers by allowing intelligent agents to take over complex, multi-step tasks during software creation. Built around Gemini 3, particularly Gemini 3 Pro, the platform relies on agents receiving specific instructions, rules, and priorities to guide their work.
Google Antigravity fundamentally changes the developer's job from manually writing every line of code to architecting the entire system. Instead of focusing on syntax, developers define high-level goals and project rules. They use the platform to dispatch smart AI Agents (powered by Gemini 3) to execute multiple complex tasks simultaneously, such as writing a new feature, fixing bugs, and running tests. This move from being a coder to being an orchestrator allows the human to focus on big-picture design and validation, making development faster and more strategic.
So, the workflow is, Local IDE (You) → Cloud (Gemini Brain) → Local IDE (Agent Action).
A lead engineer needs to launch a new invoicing service and refactor five legacy APIs before the deadline. Using Antigravity, they launch three agents: one to build the new invoicing service, a second to rewrite the legacy APIs according to new compliance rules, and a third to stress-test the entire system. The engineer simply reviews the final, verified results (Artifacts) from all three agents working in parallel.
Another note to talk about is, Cloud Elasticity is the ability of a cloud system to automatically expand or shrink its computing resources (like servers, CPU, or memory), It quickly scales out when demand unexpectedly spikes (e.g., during a sale) and then scales in to release those resources when demand drops.
For more information:
Google Antigravity Google AntigravityBuild Your First MVP Using Firebase Studio
Sometimes, it makes sense to develop a small application as an MVP to explore a simple business idea from an application perspective. In such cases, it can be acceptable to skip an extensive planning phase and instead focus on testing a small vision to see how it performs. However, it is crucial to validate the MVP from a business perspective to understand its potential market use, overall value, and how it compares to competitors. Selecting a small group or community to provide realistic feedback helps refine the concept and ensures the insights are meaningful.
Tools can be used to track ideas, notes, and feedback, ideally combining them into a system that organizes and analyzes the input. The process begins with understanding the technology you want to use, evaluating the business side based on early feedback, and creating a small representation of the idea to gather insights. This approach allows a lightweight MVP to demonstrate the concept and its relevance to the market before diving deeper into a full technical implementation. For example, a user application built as a simple MVP might lack advanced features, but its purpose is to showcase the overall idea and correlate it with business potential.
We can view the build from the user’s perspective as the outcome generated by the prompt we provided, showing the results of our input. From the system’s perspective, we can see the same underlying code and make direct edits to it to modify behavior or outcomes.
For more information:
The GitOps Maturity Model
CI/CD stands for Continuous Integration and Continuous Delivery, a methodology that enables rapid and automated software development and deployment. Continuous Integration focuses on merging code and updates into the main branch frequently, followed by automatic testing and validation, while Continuous Delivery ensures that the validated code is deployed to an environment similar to production, but not live, before final deployment, Delivery means it will be directly on the live production phase.
This approach supports faster, more reliable delivery, easier rollbacks, and smaller, manageable code changes, aligning with the principle of if it hurts, do it more, meaning frequent smaller updates reduce pain compared to large, infrequent releases. At a large scale, such as in enterprise environments with both on-premises and cloud infrastructure serving many users, CI/CD becomes more complex. Challenges like environment drift can occur, where multiple clusters run different versions of the same application, each with its own visibility, making it harder to track changes consistently and maintain system reliability.
GitOps is a way to manage systems using Git as the single source of truth, where all infrastructure and application settings are stored and tracked, allowing easy rollbacks and full audit trails. When developers update code or configurations in Git, automated pipelines or GitOps operators apply these changes directly to the system, such as updating a Kubernetes cluster without manual steps. A continuous control loop monitors the live system, and if any service crashes or manual changes occur, it automatically restores the desired state from Git, keeping the system always in sync with the source of truth.
So, CI merges and tests small chunks of code. CD deploys those changes quickly instead of waiting for big monthly releases. both mainly focus on the application layer. GitOps fills this gap by keeping infrastructure definitions in Git as YAML, and a GitOps controller automatically builds, updates, and fixes the infrastructure. Since everything is version-controlled, you can easily roll back, switch environments, and guarantee that what’s running in production always matches what’s in Git.
For more Information:
Building a Fast Website for every single visitor
The idea of responsive images in HTML is to make images automatically adjust to different screen sizes and resolutions, so they look good on desktops, tablets, and mobile devices. Instead of one fixed-size image, the browser can choose the most appropriate image based on the device’s screen width or pixel density. This helps improve page load speed, saves bandwidth, and enhances user experience.
In HTML, this is usually done using the srcset and sizes attributes in the <img> tag, which allow specifying multiple image sources for different screen conditions. The browser then selects the best option automatically.
<img src="small.jpg"
srcset="small.jpg 480w, medium.jpg 1024w, large.jpg 1600w"
sizes="(max-width: 600px) 480px, (max-width: 1200px) 1024px, 1600px"
alt="Responsive example">This code tells the browser to use small.jpg for screens up to 480px wide, medium.jpg up to 1024px, and large.jpg for larger screens.
AI can interact with responsive images by automatically generating or optimizing multiple image versions for different screen sizes and resolutions. Some AI tools can analyze an image and create appropriately sized or compressed variants, improving load speed and visual quality. Additionally, AI can suggest the best image formats, cropping, or enhancements based on device or user context.
Rendering engines are the core components of web browsers that interpret HTML, CSS, and JavaScript to display web pages visually. Chrome uses the Blink engine, Firefox uses Gecko, and Safari uses WebKit. Each engine parses the code, applies styles, executes scripts, and paints the content on the screen, ensuring websites appear correctly. They also handle layout calculations, rendering optimizations, and some browser-specific features. Essentially, rendering engines are what turn the raw code of a webpage into the interactive, visual experience users see.
On iOS, all browsers, regardless of brand, run on the Safari engine (WebKit) due to Apple’s restrictions. This means that even if you open a browser like Chrome or Firefox on iOS, it still uses WebKit under the hood. On Android, browsers like Chrome use Blink, so the same website or app may render differently compared to iOS. As a result, content can behave or appear differently between Android and iOS, even when using the same application.
WebP is an image format developed by Google that provides high-quality images with smaller file sizes compared to JPEG or PNG. Its compression reduces file size significantly while preserving most details, though extreme compression can slightly affect image clarity.
A browserlist is a configuration that defines which web browsers and versions your website or project should support. It helps tools like CSS preprocessors or JavaScript compilers automatically add necessary prefixes or polyfills for compatibility, Browserlist gets browser usage data from sites like Can I Use and StatCounter. It uses this info to decide which browsers are popular enough to support, based on rules like "> 1%" or "last 2 versions".
"browserslist": [
"last 2 versions",
"> 1%",
"not dead",
"iOS >= 12",
"Chrome >= 60"
]This tells tools to support the last 2 versions of all browsers, browsers with more than 1% market share, exclude outdated (“dead”) browsers, and specifically include iOS 12+ and Chrome 60+.
Web applications can detect a user’s poor connection and adjust content delivery to ensure accessibility. Browsers provide network inf3ormation through APIs like the Network Information API, and servers or clients can use this data to serve lower-quality videos or compressed images, balancing quality with usability. While the Network tab in DevTools shows connection info, real-time adjustments are made programmatically. This approach ensures content is still delivered even on slow connections, trading resolution or detail for reliability.
For more Information:
See ya soon
These are the four sessions I attended and wrote notes for. I couldn’t watch the remaining sessions because I was busy with video recording and photo capturing with my GDGoc BAU team, as I’m currently the mentor for that chapter.