The future of software development – the amplified developer

© 2025, Amazon Web Services, Inc. or its affiliates. All
rights reserved. Amazon Confidential and Trademark. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark. The future of software development The augmented developer Ricardo Sueiras (vi/vim) B U I L D W I T H O U T L I M I T S Developer Advocate, AWS

rights reserved. Amazon Confidential and Trademark. 3 A picture of Sir Isaac Newton standing on the shoulders of giants. pixel art (8 bit) with a blue colour theme.

rights reserved. Amazon Confidential and Trademark. 4 Developer Velocity Assembly Language Machine Code Levels of abstraction Punch Cards Compiled Languages Interpreted Languages

rights reserved. Amazon Confidential and Trademark. 5 Autocompletion Tools Frameworks (code generation) Domain Specific Languages IDE Developer Velocity Sophistication of tooling Version Control Proﬁlers Compilers Linters

rights reserved. Amazon Confidential and Trademark. 6 Each one helps us to do a better job and to write better code

rights reserved. Amazon Confidential and Trademark. 7 Search Engines Community knowledge bases Pre-internet Supporting infrastructure and systems Developer Velocity

rights reserved. Amazon Confidential and Trademark. 8 You are (so) wrong!

rights reserved. Amazon Confidential and Trademark. Search Engines Community knowledge bases Pre-internet Supporting infrastructure and systems Developer Velocity Generative AI

rights reserved. Amazon Confidential and Trademark. 10 You are absolutely right!

rights reserved. Amazon Confidential and Trademark. 11 Binary Code 00100110 10010011 … 11000011 00111100 Assembly Language LDA #20 ADC #50 STA RESULT Assembler (Tool) High-Level Language int A, B, RESULT; A = 20; B = 50; RESULT = A + B; Compiler (Tool) Ai-Powered Coding Assistant “Start with integers A and B, add them and store as RESULT.” AI-Assistant / LLM (Tools)

© 2025, Amazon Web Services, Inc. or its aﬃliates. All
rights reserved. Amazon Conﬁdential and Trademark. 12 AI Coding Assistants provide a GUI on top of large language models, that optimize use for developers across the software development life cycle (SDLC)

rights reserved. Amazon Conﬁdential and Trademark. 13

rights reserved. Amazon Conﬁdential and Trademark. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark. The craft of the developer - reimagined 14

rights reserved. Amazon Confidential and Trademark. 15 Emerging patterns Exploration Learning and Research Developer flow

rights reserved. Amazon Confidential and Trademark. 16 Exploration • Reducing friction • Doing the unexpected • Rapid prototyping

rights reserved. Amazon Confidential and Trademark. 17 Learning Research • Learn new concepts • Critic and reviewer • How does this work?

rights reserved. Amazon Confidential and Trademark. 18 Developer flow • Staying in context • Automation • Debugging and stack trace analysis • Speedy onboarding • Refactoring code

rights reserved. Amazon Confidential and Trademark. 19 Task Complexity Project Maturity Low High Greenfield Brownﬁeld +10-15% Complex tasks require deeper human insight +0-10% Constrained by outdated code and intricate dependencies +35-40% Repetitive and well deﬁned tasks +15-20% Legacy projects still benefit from simpler tasks

rights reserved. Amazon Confidential and Trademark. 21 “When an AI-powered coding assistant proposes some code, the developer must be able to read it, understand it, and make a decision as to its applicability and correctness.” Jeff Barr “My golden rule for production- quality AI-assisted programming is that I won't commit any code to my repository if I couldn't explain exactly what it does to somebody else” Simon Willison

rights reserved. Amazon Confidential and Trademark. 22 • Develop skills in how to break down complexity • Do one thing • Build incrementally, make small changes

rights reserved. Amazon Confidential and Trademark. 23 By breaking big goals into discrete tasks, Taskmaster prevents the AI from wandering off-course, letting it tackle projects piecewise.

rights reserved. Amazon Confidential and Trademark. 24 “The hottest new programming language is English” Andreij Karpathy

rights reserved. Amazon Confidential and Trademark. Generating code is the third most popular application

rights reserved. Amazon Confidential and Trademark. 26 Top Use Cases Stack trace and debugging Automating development tasks Refactoring code

rights reserved. Amazon Confidential and Trademark. 28 Daily habits Start with intent Context is EVERYTHING Perfect your Prompts Be frugal with Context Build your community 1 2 3 4 5 6

rights reserved. Amazon Confidential and Trademark. 29 no one ever mastered a new tool by just using it once

rights reserved. Amazon Confidential and Trademark. 30 70:20:10

rights reserved. Amazon Confidential and Trademark. 31 Find out what works for you

rights reserved. Amazon Confidential and Trademark. 32 Start with intent

rights reserved. Amazon Confidential and Trademark. 33 The Calm Coding Philosophy Code not with stress, but with structure. Prompt not with noise, but with intent. Build not just fast — but with flow. https://thecloudranger.github.io/tao-of-calm-coding

rights reserved. Amazon Confidential and Trademark. 34 Context + Prompt = Output

rights reserved. Amazon Confidential and Trademark. 35 Ensure that your prompts are detailed and unambiguous

rights reserved. Amazon Confidential and Trademark. 36 > How do I deploy this container on AWS? > I need to deploy a containerized Node.js e-commerce application that handles 50,000 daily users with peak loads during promotional events. Requirements: - High availability across multiple regions - MongoDB for persistence - Auto-scaling capabilities Please provide: 1. AWS architecture diagram 2. List of required services with configurations 3. Security best practices 4. Operational monitoring recommendations Avoid Better

rights reserved. Amazon Confidential and Trademark. 37 Refine and iterate your Prompts

rights reserved. Amazon Confidential and Trademark. 38 Meta prompting is your friend

rights reserved. Amazon Confidential and Trademark. 39 Prompt > I want to use the following prompt to get you to write me some code. Can you review the prompt and suggest areas where it should be improved. Here is the prompt: Build a simple Fact Checking application. - Generate a web application that can be used in a browser - Users will need to register with an email address to login - When Users login, a Dashboard will be displayed that provides a simple explanation of what the application does, and displays any available Facts that have been created - From the home Dashboard, Users will be able to click on any existing Facts to fact check. They will also be able to Create a new Fact or Create a new Category - When Users are viewing Facts, they will have the ability to click on two buttons - Fact or Fake - When Users are viewing Facts, they can also provide supporting info - Provide a simple web design that can be updated easily using CSS

rights reserved. Amazon Confidential and Trademark. 40 Provide (good) examples

rights reserved. Amazon Confidential and Trademark. 41 Good • What are your expected Output? • Specify how you want the information presented • Provide example input and output data formats • Provide sample code snippets or examples you want it to follow Avoid • Be specific in what you want as inputs and outputs, don't generalise

rights reserved. Amazon Confidential and Trademark. 42 Use precise language when crafting prompts

rights reserved. Amazon Confidential and Trademark. 43

rights reserved. Amazon Confidential and Trademark. 44 Context is everything! Provide the right context

rights reserved. Amazon Confidential and Trademark. 45 Create Personas to tailor the output of your AI coding assistant

rights reserved. Amazon Confidential and Trademark. 46 ├── personas │ ├── backend-engineer │ │ └── Agent.md │ ├── database-architect │ │ └── Agent.md │ ├── fullstack-engineer │ │ └── Agent.md │ ├── product-manager │ │ └── Agent.md │ ├── software-architect │ │ └── Agent.md │ ├── software-engineer │ │ └── Agent.md │ ├── software-testing-engineer │ │ └── Agent.md │ └── user-interface-developer │ └── Agent.md Your Role: You are an experienced database architect. Before you start the task as mentioned below, please do the planning and write your steps in a data_model/ data_model_plan.md file with checkboxes against each step in the plan. If any step needs my clarification, please add it to the step to interact with me and get my confirmation. Do not make critical decisions on your own. Once you produce the plan, ask for my review and approval. After my approval, you can go ahead to execute the same plan one step at a time. Once you finish each step, mark the checkboxes as done in the plan. database-architect

rights reserved. Amazon Confidential and Trademark. 47 ├── personas │ ├── backend-engineer │ │ └── Agent.md │ ├── database-architect │ │ └── Agent.md │ ├── fullstack-engineer │ │ └── Agent.md │ ├── product-manager │ │ └── Agent.md │ ├── software-architect │ │ └── Agent.md │ ├── software-engineer │ │ └── Agent.md │ ├── software-testing-engineer │ │ └── Agent.md │ └── user-interface-developer │ └── Agent.md Your Role: You are an expert testing strategist responsible for comprehensively testing a full-stack application. Your goal is to verify that all components of the system work together seamlessly. Before you start the task as mentioned below, please do the planning and write your steps in the testing/testing_plan.md file with checkboxes against each step in the plan. If any step needs my clarification, please add it to the step to interact with me and get my confirmation. Do not make critical decisions on your own. Once you produce the plan, ask for my review and approval. After my approval, you can go ahead to execute the same plan one step at a time. Once you finish each step, mark the checkboxes as done in the plan. Software Testing Engineer

rights reserved. Amazon Confidential and Trademark. 48 What makes good Context?

rights reserved. Amazon Confidential and Trademark. 49 Data Model API definition Architectural diagrams Screen Mockups Example Code Snippets Specifications

rights reserved. Amazon Confidential and Trademark. I’ll just throw all my docs, design docs, requirements, test cases, and source code into the context and everything will be fine……

rights reserved. Amazon Confidential and Trademark. Stanford University

rights reserved. Amazon Confidential and Trademark. 52 Be frugal with what you add to Context

rights reserved. Amazon Confidential and Trademark. 53 Accelerate learning by building developer community

rights reserved. Amazon Confidential and Trademark. 54 #modelcontextprotocolinterest #amazon-builder-genai-power-users

rights reserved. Amazon Confidential and Trademark. 55 https://www.promptz.dev/ Community created projects that are helping organisations share, discover, and re-use prompts that their developers are ﬁnding useful

rights reserved. Amazon Confidential and Trademark. 56 AWS User Group Leeds, 7th October

rights reserved. Amazon Confidential and Trademark.

rights reserved. Amazon Confidential and Trademark. Secure code….?

rights reserved. Amazon Confidential and Trademark. Code Hallucinations

rights reserved. Amazon Confidential and Trademark. Trust but verify

rights reserved. Amazon Confidential and Trademark. Review code snippet Review logic Good practices Verify error handling Use Explain

rights reserved. Amazon Confidential and Trademark. beware the generated dependency ﬁle

rights reserved. Amazon Confidential and Trademark. Small code bases

rights reserved. Amazon Confidential and Trademark. Stanford University

rights reserved. Amazon Confidential and Trademark. Echo chamber

rights reserved. Amazon Confidential and Trademark. One of the impacts of AI coding tools is how they will make “technologies preferred by LLMs” dominant even more.

rights reserved. Amazon Confidential and Trademark. The end of proprietary tech?

rights reserved. Amazon Confidential and Trademark. Update developer workflows

rights reserved. Amazon Confidential and Trademark. Do not rely on AI coding tools to undo changes

rights reserved. Amazon Confidential and Trademark. Code Reviews …. the new bottleneck

rights reserved. Amazon Confidential and Trademark. https://github.com/BloopAI/vibe-kanban

rights reserved. Amazon Confidential and Trademark. 76 76 AI Fluency Software Developer knowledge Low High Low High Over reliance on AI – the “Over-AI’d” The “augmented developer” The “Unready” Prof. Douglas Schmidt William & Mary The “under confident”

rights reserved. Amazon Confidential and Trademark. 77 77 AI Fluency Software Developer knowledge Low High Low High Over reliance on AI – the “Over-AI’d” The “Unready” Prof. Douglas Schmidt William & Mary The “under confident” The “augmented developer”

rights reserved. Amazon Conﬁdential and Trademark. 78 78 AI Fluency Software Developer knowledge Low High Low High Over reliance on AI – the “Over-AI’d” The “Unready” The “under confident” Prof. Douglas Schmidt William & Mary The “augmented developer”

rights reserved. Amazon Confidential and Trademark. 79 79 AI Fluency Software Developer knowledge Low High Low High Over reliance on AI – the “Over-AI’d” The “Unready” Prof. Douglas Schmidt William & Mary The “under confident” The “augmented developer”

rights reserved. Amazon Confidential and Trademark. 80 80 AI Fluency Software Developer knowledge Low High Low High Over reliance on AI – the “Over-AI’d” The “Unready” Prof. Douglas Schmidt William & Mary The “under conﬁdent” The “augmented developer”

rights reserved. Amazon Confidential and Trademark. 83 Helping developers write code faster 2023 Completing development tasks end-to-end with human in the loop Generating larger pieces of code and answering questions Unprecedented pace of change AUTO-COMPLETE 2024 ASSISTANTS 2025 AGENTS

rights reserved. Amazon Conﬁdential and Trademark. Vibe Coding: Prompt and Pray AI unleashed, enabling experimentation and the development of prototypes • Rapid, conversational code generation (CHOP) • Iterative, back and forth • Ephemeral • Point in time prompts • Transient context

rights reserved. Amazon Conﬁdential and Trademark. AI coding tools excel at small tasks but can fail with complex projects Scaling AI development Existing tools make it difficult to collaborate with and manage agents Limited control Challenges with AI development

rights reserved. Amazon Conﬁdential and Trademark. 86 Challenges with AI development Getting a project from proof-of- concept to production while maintaining quality control becomes increasingly difficult Code quality

rights reserved. Amazon Conﬁdential and Trademark. 87 Good practices emerged Emergence of good practices to help keep AI on a short leash Developers learned how to manually break down large problems into smaller units and build incrementally Break down large problems Speciﬁcity and Clarity Context and Prompt engineering Precision and clarity are key in directing AI coding assistants to generate good outputs Providing the right context is key to producing consistency and control

rights reserved. Amazon Confidential and Trademark. 88 “You don’t program by chatting. You program by writing documents.”

rights reserved. Amazon Confidential and Trademark. 89 A written specification aligns humans

rights reserved. Amazon Confidential and Trademark. 90 90 Spec Driven Development Core philosophy Invest time to understand what you are trying to build Clarity before Code Iterative refinement Code via docs Iterate and capture evolution of what you are trying to build From ephemeral chat to persistent documents that can be shared with your stakeholders

rights reserved. Amazon Conﬁdential and Trademark. 91 91 Spec Driven Development Separate the design from the build of a project Create clear requirements and design specifications. Define the vision Make architectural decisions Implement with context Choose technologies, patterns, and approaches upfront. Use AI to generate code that fulfills your documented specifications. 1 2 3

rights reserved. Amazon Confidential and Trademark. 92 The Vibe • Rapid, conversational code generation (CHOP) • Iterative, back and forth • Ephemeral • Point in time prompts • Transient context Spec driven • Focus on upfront planning and intent • Break down requests into discrete tasks • Steering documents ground agentic outputs • Lineage between intent and code Prompts to chase implementations Preserve intent, improve visibility

rights reserved. Amazon Conﬁdential and Trademark. Lifecyle: Spec-driven development AI under control: Specification driven lifecycle REFINE and UPDATE ISSUES FOUND Test/QA Implementation Intent Spec-driven development Context Steering documents Requirements Generate requirements Design Create initial Design Tasks Deﬁne tasks needed Maintenance INFLUENCES ITERATE ITERATE OPINIONATED WORKFLOW Deployment UPDATE or NEW

rights reserved. Amazon Conﬁdential and Trademark. 94 “Kiro (https://kiro.dev/) is the first IDE to take specs seriously and not see them as side reuseable prompts” Patrick Debois

rights reserved. Amazon Conﬁdential and Trademark. Inner loop has changed, the outer loop is catching up

rights reserved. Amazon Conﬁdential and Trademark. Inner loop Outer loop push Develop Build Test Integrate Test Release Deploy

rights reserved. Amazon Confidential and Trademark. Amazon Q CLI

rights reserved. Amazon Confidential and Trademark. © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark. 98 It has never been a better time to be a developer

rights reserved. Amazon Confidential and Trademark. 100 What about developers just starting out?

rights reserved. Amazon Confidential and Trademark. AI Coding Assistants are a gateway to software development. “AI is unleashing a generation of wildly creative builders beyond anything I’d have imagined, and they grow up ”knowing” they can build anything!” Thomas Wolf, HuggingFace

rights reserved. Amazon Confidential and Trademark. 105 https://tidyﬁrst.substack.com/p/90-of-my-skills-are-now-worth-0 “The value of 90% of my skills just dropped to $0. The leverage for the remaining 10% went up 1000x” Kent Beck

rights reserved. Amazon Confidential and Trademark. 106 Stuff that generative AI developer tools can do better than me Things that generative AI developer tools will help me do better Things that developers will always do best “Some tools of thought are so blunt that they are almost useless; others are so sharp that they are dangerous. But the wise man will have the use of both kinds.” ~ Gregory Bateson

rights reserved. Amazon Confidential and Trademark. 107 Augmented developers – the new normal?

rights reserved. Amazon Confidential and Trademark. 108 1 Create daily habits – find out what works for you! 2 Start with intent 3 Dive deep into Prompt and Context engineering 4 Start building your community today!

rights reserved. Amazon Confidential and Trademark. 109 “What I suggest, is that people experiment with them.” Martin Fowler

rights reserved. Amazon Confidential and Trademark. 110 AI can write code but it can’t build software https://bytesauna.com/post/coding-vs-software-engineering Matias Heikkilä

rights reserved. Amazon Confidential and Trademark. 111 https://s12d.com/builder-id AWS Builder Centre Start your community today!

rights reserved. Amazon Confidential and Trademark. Thank you! © 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark. https://www.linkedin.com/in/ricardosueiras 094459.bsky.social https://github.com/094459

The future of software development – the amplif...

The future of software development – the amplified developer

More Decks by Ricardo Sueiras

Featured

Transcript