Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Throw Away The Vibes: Context Engineering Is Al...

Throw Away The Vibes: Context Engineering Is All You Need

We've all been there: GitHub Copilot promises to be your coding companion, but instead feels more like that overeager intern who confidently writes brilliant code for the wrong problem. As AI-assisted development tools become ubiquitous, the industry narrative promises revolutionary productivity gains. Yet many practitioners find themselves playing an exhausting game of context whack-a-mole—constantly explaining, re-explaining, and fixing what their AI "partner" confidently got wrong.

Having spent considerable time wrestling with this gap between AI promises and reality, I've discovered that the real challenge isn't prompt engineering—it's context engineering. This talk explores why "vibe coding" is fundamentally broken and introduces the systematic discipline that separates magical AI experiences from expensive disappointments. Through the constraint-context matrix and real examples like the Breadcrumb Protocol, I'll demonstrate why human expertise in scaffolding, steering, and domain understanding isn't just relevant—it's the secret ingredient that makes AI actually work. You'll leave with practical strategies for engineering context systematically and a framework for building sustainable human-AI collaboration that leverages both your skills and the machine's capabilities.

Avatar for Dasith Wijesiriwardena

Dasith Wijesiriwardena

August 28, 2025
Tweet

Transcript

  1. Throw Away The Vibes: Context Engineering Is All You Need

    2nd Sep 2025 Dasith Wijesiriwardena @dasiths Melb .NET Meetup
  2. I am here because I have a love hate relationship

    with distributed systems. HELLO! I AM Dasith @dasiths dasith.me
  3. @dasiths “ ..Our philosophy is don't fix the code. Fix

    the context.. https://www.acquired.fm/episodes/how-is-ai-different-than-other-technology-waves-with-bret-taylor-and-clay-bavor
  4. @dasiths ▪ It doesn’t work, I don’t know why. ▪

    It works but I don’t know why. ▪ It works with my prompt. Growing Pains
  5. @dasiths ▪ It doesn’t work, I don’t know why. ▪

    It works but I don’t know why. ▪ It works with my prompt. Growing Pains
  6. @dasiths ▪ It doesn’t work, I don’t know why. ▪

    It works but I don’t know why. ▪ It works worked with my prompt. Growing Pains
  7. @dasiths ▪ It doesn’t work, I don’t know why. ▪

    It works but I don’t know why. ▪ It works worked with my prompt. Growing Pains
  8. • Poisoning • Distraction • Confusion •Clash How Long Context

    Fails @dasiths https://www.dbreunig.com/2025/06/26/how-to-fix-your-context.html
  9. Give your model the… > Right Context At the… >

    Right Time Let’s Fix Context @dasiths
  10. Give your model the… > Right Enough Context At the…

    > Right Time Let’s Fix Context @dasiths https://blog.langchain.com/multi-needle-in-a-haystack/
  11. Context Engineering involves creating systems that provide the correct information

    and tools, in the proper format, exactly when needed, so that a large language model (LLM) has everything required to complete a task. The What? @dasiths https://www.philschmid.de/context-engineering
  12. Context Engineering involves creating systems that provide the correct information

    and tools, in the proper format, exactly when needed, so that a large language model (LLM) has everything required to complete a task. The What? @dasiths https://www.philschmid.de/context-engineering
  13. Context Engineering involves creating systems that provide the correct information

    and tools, in the proper format, exactly when needed, so that a large language model (LLM) has everything required to complete a task. The What? @dasiths https://www.philschmid.de/context-engineering
  14. Context Engineering involves creating systems that provide the correct information

    and tools, in the proper format, exactly when needed, so that a large language model (LLM) has everything required to complete a task. The What? @dasiths https://www.philschmid.de/context-engineering https://x.com/karpathy/status/1937902205765607626
  15. vs Prompt Engineering @dasiths https://www.philschmid.de/context-engineering Aspect Prompt Engineering Context Engineering

    Core Definition Crafting the perfect set of instructions in a single text string Designing and building dynamic systems that provide the right information and tools, in the right format, at the right time Scope Single interaction optimization System-level design and architecture Approach Static template refinement Dynamic, adaptive system construction Focus "What to say" to the model "What context to provide" to the model Nature String-based System-based Timing Fixed at prompt creation Dynamic, created on-the-fly
  16. 1. Write Context 2. Select Context 3. Compress Context 4.

    Isolate Context Core Tactics @dasiths https://rlancemartin.github.io/2025/06/23/context_engineering/ https://www.dbreunig.com/2025/06/26/how-to-fix-your-context.html
  17. 1. Write Context 2. Select Context 3. Compress Context 4.

    Isolate Context @dasiths https://rlancemartin.github.io/2025/06/23/context_engineering/ https://www.dbreunig.com/2025/06/26/how-to-fix-your-context.html Core Tactics
  18. 1. Write Context 2. Select Context 3. Compress Context 4.

    Isolate Context @dasiths https://rlancemartin.github.io/2025/06/23/context_engineering/ https://www.dbreunig.com/2025/06/26/how-to-fix-your-context.html Core Tactics
  19. 1. Write Context 2. Select Context 3. Compress Context 4.

    Isolate Context @dasiths https://rlancemartin.github.io/2025/06/23/context_engineering/ https://www.dbreunig.com/2025/06/26/how-to-fix-your-context.html Core Tactics
  20. 1. Write Context 2. Select Context 3. Compress Context 4.

    Isolate Context @dasiths https://rlancemartin.github.io/2025/06/23/context_engineering/ https://www.dbreunig.com/2025/06/26/how-to-fix-your-context.html Core Tactics
  21. ▪ Optimize for KV Cache ▪ Prefer tool masking over

    removing ▪ File system as context ▪ Manipulate attention mechanism through recitation ▪ Error recovery as core capability ▪ Pattern mimicry problem @dasiths https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus Lessons from Manus.im
  22. ▪ Optimize for KV Cache ▪ Prefer tool masking over

    removing ▪ File system as context ▪ Manipulate attention mechanism through recitation ▪ Error recovery as core capability ▪ Pattern mimicry problem @dasiths https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus Lessons from Manus.im
  23. ▪ Optimize for KV Cache ▪ Prefer tool masking over

    removing ▪ File system as context ▪ Manipulate attention mechanism through recitation ▪ Error recovery as core capability ▪ Pattern mimicry problem @dasiths https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus Lessons from Manus.im
  24. ▪ Optimize for KV Cache ▪ Prefer tool masking over

    removing ▪ File system as context ▪ Manipulate attention mechanism through recitation ▪ Error recovery as core capability ▪ Pattern mimicry problem @dasiths https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus Lessons from Manus.im
  25. ▪ Optimize for KV Cache ▪ Prefer tool masking over

    removing ▪ File system as context ▪ Manipulate attention mechanism through recitation ▪ Error recovery as core capability ▪ Pattern mimicry problem @dasiths https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus Lessons from Manus.im
  26. ▪ Optimize for KV Cache ▪ Prefer tool masking over

    removing ▪ File system as context ▪ Manipulate attention mechanism through recitation ▪ Error recovery as core capability ▪ Pattern mimicry problem @dasiths https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus Lessons from Manus.im
  27. ▪ Inconsistent Implementation – “The overeager intern” ▪ Knowledge Silos

    – “When the why is missing” ▪ Progress Fragmentation – “Worked with my prompt” @dasiths https://dasith.me/2025/04/02/vibe-coding-breadcrumbs/ The Misalignment Problem
  28. The cost of this misalignment grows as development continues. ▪

    Code reviews become more difficult ▪ Onboarding new team members takes longer ▪ The AI becomes less effective as a collaborator over time. @dasiths https://dasith.me/2025/04/02/vibe-coding-breadcrumbs/ The Misalignment Problem
  29. The solution lies in creating an external, persistent shared context

    that both humans and AI can access and update. It’s a structured workflow consisting of… - Structured Planning & Task Management - Centralized & Accessible Knowledge Context - Living Documentation & Shared Understanding @dasiths https://dasith.me/2025/04/02/vibe-coding-breadcrumbs/ Solution: Breadcrumb Protocol
  30. @dasiths “You’re going to see progress made. They’re going to

    get better. But I think that there’s an upper limit with the way that autoregressive transformers work that we just won’t get past.” - Microsoft Azure CTO Mark Russinovich Upper Limit?
  31. @dasiths “As we take charge of bridging the gap, the

    machine claims the praise.” > We Often Overestimate Machines and Undervalue Ourselves Using large language models (LLMs) for programming requires many supporting skills. However, these skills are not clearly defined and depend a lot on the situation, so right now, the best way to improve is through lots of practice and learning from mistakes. https://ferd.ca/the-gap-through-which-we-praise-the-machine.html
  32. ▪ Start with one context tactic and measure results ▪

    Build systematic practices around context management ▪ Recognize context engineering as core technical skill @dasiths Practical Takeaways
  33. ▪ Start with one context tactic and measure results ▪

    Build systematic practices around context management ▪ Recognize context engineering as core technical skill @dasiths Practical Takeaways
  34. Title Description URL [1] Pete Hodgson's Constraint-Context Matrix Primary problem

    framework https://blog.thepete.net/blog/2025/05/22/why-your-ai-coding-assistant-keeps-doing-it-wrong-and-how- to-fix-it/ [2] Drew Breunig's "How to Fix Your Context" Six practical tactics for context management https://www.dbreunig.com/2025/06/26/how-to-fix-your-context.html [3] Lance Martin's Context Engineering Four-strategy foundation https://rlancemartin.github.io/2025/06/23/context_engineering/ [4] Philipp Schmid's Context Engineering Context vs. prompt engineering distinction https://www.philschmid.de/context-engineering [5] Anthropic Research Context offloading (54% improvement) https://www.anthropic.com/engineering/claude-think-tool [6] "Less is More" Paper Smart tool selection (44% improvement) https://arxiv.org/abs/2411.15399 [7] Fred Hebert's Human-Machine Gap Operator skills importance https://ferd.ca/the-gap-through-which-we-praise-the-machine.html [8] Microsoft Azure CTO Warning Industry reality check on vibe coding hype https://www.geekwire.com/2025/reality-check-microsoft-azure-cto-pushes-back-on-ai-vibe-coding-hype- sees-upper-limit-long-term/ [9] Domain Understanding Requirement Context engineering needs domain expertise https://medium.com/@springrod/context-engineering-needs-domain-understanding-b4387e8e4bf8 [10] Breadcrumb Protocol Foundation workflow https://dasith.me/2025/04/02/vibe-coding-breadcrumbs/ [11] Manus Production Context Engineering Real-world lessons from building AI agents at scale https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus References
  35. Presentation template designed by powerpointify.com Special thanks to all people

    who made and shared these awesome resources for free: CREDITS Photographs by unsplash.com Free Fonts used: https://www.fontsquirrel.com/fonts/oswald @dasiths