Agentic Coding for Mobile Developers

Software engineering has changed dramatically. In just a few months, we’ve gone from AI-assisted code completion to fully autonomous coding agents that can build entire applications from simple prompts.

*Or can they?*

In this talk, I am going to explain what coding agents are, how they work, and how you can use them most efficiently for your work.

You will learn

– How coding agents work (inner / outer loop)
– The role of AGENTS.md
– What skills are, and why they are essential for coding agents
– The importance of writing a clear specification
– How to use all of this in your own projects

Throughout the talk, I will demonstrate the key concepts by vibe-engineering a food-ordering app, backed by Firebase. You will walk away with a concrete recipe for using agentic coding in your own projects.

Transcript

1. 2026 @peterfriese.dev Agentic Coding Peter Friese, Developer Relations Engineer, Google for Mobile Developers Follow me on Bluesky
2. When was the last time you wrote assembly code?
3. Will we stop using high-level programming languages to build software in the next 5 years? Peter Friese
4. LLMs are the new operating systems, and we program them in English Andrej Karpathy
5. 1950s 1970s 2024 1996 2022 2019 Punch cards Typing code Code completion Agentic coding Tab completion Chat + copy & paste Intellisense VisualBasic 5.0 2008 StackOverflow “loop” Turn around time: hours - days Real-time editing Tabnine: 2018/19 Copilot: 2021 Cursor Tab: 2024
6. How do coding agents work
7. /ˈprɒmpt/
8. /ˈprɒmpt/ 2/ An instruction given to an artificial intelligence by a human using natural language rather than computer language 1/ A sign on a computer screen that shows that the computer is ready to receive your instructions An instruction given to an artificial intelligence by a human using natural language rather than computer language Source: https://dictionary.cambridge.org/dictionary/english/prompt
10. Zero-shot prompting Simplest type of prompt Provide only description of task Zero-shot = “no examples” One-shot prompting Provide example(s)
11. System prompting Set overall context and purpose Big picture Contextual prompting Provide specific details Background information Role prompting Assign specific character / identity Helps model to be consistent
12. Source: https://www.linkedin.com/posts/the-anatomy-of-gait-prompt-goal-define-share-7325433603348164608-Nr0m/
13. Source: https://x.com/rubenhassid/status/2027991271252320693
14. - Peter Friese Coding agents are like Tiramisu - just not as delicious!
15. System Instructions ! AGENTS.md ! Skills ! User prompt !
16. AGENTS.md A simple, open format for guiding coding agents, used by over 60k open-source projects. Think of AGENTS.md as a README for agents: a dedicated, predictable place to provide the context and instructions to help Al coding agents work on your project. Explore Examples © View on GitHub # AGENTS.md ## Setup commands - Install deps: phpm install" - Start dev server:'phpm dev" - Run tests: 'pnpm testi ## Code style - TypeScript strict mode - Single quotes, no semicolons - Use functional patterns where possible
17. Agent Skills Overview Specification Client Showcase For skill creators Quickstart Best practices Optimizing descriptions Evaluating skills Using scripts For client implementors Adding skills support Q Search... * Ask Assistant • agentskills/agentskills ₺ 21,205 Agent Skills Overview A standardized way to give Al agents new capabilities and expertise. • Copy page = On this page What are Agent Skills? Why Agent Skills? How do Agent Skills work? Where can I use Agent Skills? Open development Get started with Agent Skills What are Agent Skills? Agent Skills are a lightweight, open format for extending Al agent capabilities with specialized knowledge and workflows. At its core, a skill is a folder containing a SKILL. md file. This file includes metadata ( name and description, at minimum) and instructions that tell an agent how to perform a specific task. Skills can also bundle scripts, reference materials, templates, and other resources. my-skill/ SKILL.md scripts/ references/ assets/ # Required: metadata + instructions Optional: executable code Optional: documentation #Optional: templates, resources # Any additional files or directories
18. User Application Prompt Function call Response Model Prompt and Function declaration Function identifier and parameters Request API Response Response from API Response
19. Inner loop ! Execute Call tools (Compile, run tests, shell commands) ! Observe Capture tool output and terminal feedback Evaluate Assess progress against current objective ! Strategic plan Goal decomposition & task tracking ! User task “Implement a Gelato ordering app” ! Task complete Code committed / PR opened Milestone achieved Continue working
 on current task
20. Demo!
21. Gelato GO
22. Live demo Try it our! Order gelato: https://bit.ly/gelato-go
23. Source: https://x.com/shmidtqq/status/2068704187492221405?s=46
24. thanks.
25. Resources Firebase livestreams Slides for this talk https://www.youtube.com/@Firebase/streams https://speakerdeck.com/peterfriese Firebase blog https://firebase.blog/ Livestream https://www.youtube.com/@PeterFriese/streams
26. One more thing…
27. (Free)
28. fin.