What are the key takeaways from “I stopped using /grill-me for coding. Here’s what I use instead:” on Matt Pocock?
Stop Talking to AI: Why Your Codebase Needs a Shared Language
Insights from the Matt Pocock episode “I stopped using /grill-me for coding. Here’s what I use instead:”, published May 14, 2026.
Frequently asked questions about “I stopped using /grill-me for coding. Here’s what I use instead:”
What is "I stopped using /grill-me for coding. Here’s what I use instead:" about?
In "I stopped using /grill-me for coding. Here’s what I use instead:" (Matt Pocock, May 2026), generic prompting leads to hallucination and verbosity. By integrating Domain-Driven Design (DDD) principles—specifically ubiquitous language and ADRs—into an iterative AI 'grilling' session, you can achieve near-telepathic alignment between your intent and the AI's output.
What does "Ubiquitous Language" mean in "I stopped using /grill-me for coding. Here’s what I use instead:"?
In "I stopped using /grill-me for coding. Here’s what I use instead:", Derived from Domain-Driven Design, this concept suggests that the language used in code, documentation, and conversation should be identical. By forcing the AI to stick to this glossary, you eliminate the friction of translating ideas into machine-compatible concepts.
What does "Architectural Decision Record (ADR)" mean in "I stopped using /grill-me for coding. Here’s what I use instead:"?
In "I stopped using /grill-me for coding. Here’s what I use instead:", ADRs are essential for documenting decisions that are difficult to reverse. They provide the AI (and future human maintainers) with the context needed to understand why a specific approach was taken, preventing the accidental undoing of critical architectural patterns.
What does "Grill with Docs" mean in "I stopped using /grill-me for coding. Here’s what I use instead:"?
In "I stopped using /grill-me for coding. Here’s what I use instead:", This is an improvement over the original 'Grill Me' skill. It forces the AI to analyze your `context.md` file first, which ensures all recommendations are grounded in your specific project needs rather than generic programming advice.
What does "I stopped using /grill-me for coding. Here’s what I use instead:" say about move from raw prompting to 'context-aware grilling' by?
In "I stopped using /grill-me for coding. Here’s what I use instead:", Move from raw prompting to 'context-aware grilling' by forcing the AI to reference your local repository documentation. Reduces the need to re-explain domain concepts, resulting in faster and more accurate code generation.
What does "I stopped using /grill-me for coding. Here’s what I use instead:" say about use Architectural Decision Records?
In "I stopped using /grill-me for coding. Here’s what I use instead:", Use Architectural Decision Records (ADRs) for non-obvious choices that are hard to reverse. Prevents long-term technical debt by capturing the rationale behind trade-offs that simple code comments miss.
What is this episode about?
Generic prompting leads to hallucination and verbosity. By integrating Domain-Driven Design (DDD) principles—specifically ubiquitous language and ADRs—into an iterative AI 'grilling' session, you can achieve near-telepathic alignment between your intent and the AI's output.
What are the key takeaways?
Insights from the Matt Pocock episode “I stopped using /grill-me for coding. Here’s what I use instead:”, published May 14, 2026.
Move from raw prompting to 'context-aware grilling' by forcing the AI to reference your local repository documentation. — Reduces the need to re-explain domain concepts, resulting in faster and more accurate code generation.
Use Architectural Decision Records (ADRs) for non-obvious choices that are hard to reverse. — Prevents long-term technical debt by capturing the rationale behind trade-offs that simple code comments miss.
Shared language is a performance optimization for AI agents. — Aligning on terminology significantly reduces token consumption in thinking traces and generated code.
What concepts are explained?
Insights from the Matt Pocock episode “I stopped using /grill-me for coding. Here’s what I use instead:”, published May 14, 2026.
Ubiquitous Language: Derived from Domain-Driven Design, this concept suggests that the language used in code, documentation, and conversation should be identical. By forcing the AI to stick to this glossary, you eliminate the friction of translating ideas into machine-compatible concepts.
Architectural Decision Record (ADR): ADRs are essential for documenting decisions that are difficult to reverse. They provide the AI (and future human maintainers) with the context needed to understand why a specific approach was taken, preventing the accidental undoing of critical architectural patterns.
Grill with Docs: This is an improvement over the original 'Grill Me' skill. It forces the AI to analyze your `context.md` file first, which ensures all recommendations are grounded in your specific project needs rather than generic programming advice.
Notable quotes
Insights from the Matt Pocock episode “I stopped using /grill-me for coding. Here’s what I use instead:”, published May 14, 2026.
“Documenting Non-Obvious Decisions Results In Magical Alignment With AI”
— Matt Pocock, “I stopped using /grill-me for coding. Here’s what I use instead:”
Who should listen to this episode?
Software engineers and product builders using AI to architect, prototype, or refactor complex codebases.
This summary was generated by Yedapo and may contain inaccuracies. It does not represent the views of the original creators.
30-second answer
Stop Talking to AI: Why Your Codebase Needs a Shared Language
Generic prompting leads to hallucination and verbosity. By integrating Domain-Driven Design (DDD) principles—specifically ubiquitous language and ADRs—into an iterative AI 'grilling' session, you can achieve near-telepathic alignment between your intent and the AI's output.
Bottom line
Adopt 'Grill with Docs' by maintaining a context.md file and Architectural Decision Records (ADRs) to force AI agents to adhere to your project’s domain language.
Eliminating ambiguity in your LLM's understanding of your domain prevents the 'verbose loop' and ensures the generated code aligns with your specific architectural constraints.
Best moment
The host demonstrates how 'Grill with Docs' forces the AI to challenge domain-level ambiguities before writing a single line of implementation code.
Three takeaways
If you only read this, you've got it.
1
Move from raw prompting to 'context-aware grilling' by forcing the AI to reference your local repository documentation.
Reduces the need to re-explain domain concepts, resulting in faster and more accurate code generation.
2
Use Architectural Decision Records (ADRs) for non-obvious choices that are hard to reverse.
Prevents long-term technical debt by capturing the rationale behind trade-offs that simple code comments miss.
3
Shared language is a performance optimization for AI agents.
Aligning on terminology significantly reduces token consumption in thinking traces and generated code.
Get insights on every episode of Matt Pocock
Sign up free to unlock the full analysis, chapters, key concepts, and Ask AI.
Strategy Comparison: Grill Me vs. Grill with Docs
Compare when to use different AI interaction patterns based on your project maturity and document availability.
Subject
Takeaway
Why it matters
Caveat
Grill Me
Best for ideation, general problem solving, or tasks without an existing codebase.
Provides high-quality, generic probing to help you flesh out concepts without needing project-specific context.
—
Grill with Docs
Best for active codebase development and complex domain logic.
Forces the AI to adhere to predefined constraints and glossary, ensuring generated code is consistent with existing architecture.
—
Grill Me
Best for ideation, general problem solving, or tasks without an existing codebase.
Provides high-quality, generic probing to help you flesh out concepts without needing project-specific context.
Grill with Docs
Best for active codebase development and complex domain logic.
Forces the AI to adhere to predefined constraints and glossary, ensuring generated code is consistent with existing architecture.
One thing to do · 30min
Create a `context.md` file at the root of your project.
Establishes a single source of truth for domain-specific terminology that AI agents can use for immediate context.
“The same shared language that enables developers and domain experts to build software effectively also serves as the optimal mental model for Large Language Models to reduce token usage and improve code accuracy.”
Full Context
A 1-minute read.
Effective communication with LLMs is primarily an exercise in linguistic alignment. The host explains that standard prompting fails because it lacks a shared definition of the problem space, forcing the user to constantly 'patch' the AI's understanding during the conversation. The solution is to adopt the Domain-Driven Design (DDD) principle of 'ubiquitous language' by codifying terminology into a `context.md` file that the agent must reference. By doing so, the user ensures that the AI's 'thinking' is constrained by the same vocabulary used by the developers and the business domain experts.
Beyond simple terminology, the host emphasizes the importance of Architectural Decision Records (ADRs) for handling complexity. ADRs serve as a persistent memory layer, allowing the AI to understand the 'why' behind irreversible decisions, which prevents the recurrence of bad architectural patterns. During a live demonstration, the host shows how the 'Grill with Docs' skill forces the AI to surface terminology collisions early, such as clarifying the relationship between 'pitches' and 'standalone videos' before any code is written. This rigorous 'grilling' process prevents the user from drifting into implementation details too early, ensuring that the underlying data model is robust and well-defined.
The result of this ceremony is a tighter feedback loop, where the AI can provide more concise, context-aware responses because it no longer requires the user to explain non-obvious domain nuances. This isn't just about cleaner code; it is about cognitive alignment. The host suggests that this approach mimics how effective human teams function, where shared definitions allow for high-bandwidth communication and reduced ambiguity. Ultimately, the host argues that the best LLM performance is unlocked by creating a structured, documented environment that guides the model's reasoning process.
If you liked this
Save this summary
Export to Markdown, Obsidian, or Notion — a Pro feature.
