I Stopped Using /grill-me for Coding. Here's What I Use Instead — Deep Dive

Matt Pocock ⏱ ~15:16 Watch on YouTube ↗

Overview

Matt Pocock reveals the successor to his wildly popular Grill Me Claude skill. The new skill, Grill with Docs, combines the relentless interviewing of Grill Me with ubiquitous language documentation inspired by Domain-Driven Design. The result: AI that aligns with your thinking using fewer tokens, sharper language, and better code.

The Phenomenon of Grill Me ▶ 0:00

Matt wrote 4 sentences that became the most influential he's ever written — the Grill Me skill. It interviews you relentlessly until reaching shared understanding, walking down each branch of the design tree, resolving dependencies between decisions one by one. He receives ~5 messages daily from people praising it.

"Absolute game-changer" · "Goated" · "Might save time — you just one-shot everything after gathering context."

Where Grill Me Falls Short ▶ 1:03

Despite the praise, Matt found recurring problems:

Core frustration: communication was effective, but lacked persistent shared vocabulary.

Ubiquitous Language — The Missing Piece ▶ 3:30

Inspired by Domain-Driven Design (Eric Evans' "Big Blue Book"), Matt created a ubiquitous language skill. The concept: code, developers, and domain experts should all use a shared language documented in a single source of truth.

He started using it alongside Grill Me — calling the ubiquitous language skill mid-session to sharpen terms and create a context.md file. Then realized: why not combine both into one skill?

Grill with Docs — The New Skill ▶ 4:41

Grill with Docs = Grill Me + Ubiquitous Language + ADRs, combined into one skill.

Three Layers

  1. context.md — Documents all shared language (glossary of domain terms). Uses DDD's concept of bounded contexts. One context.md per bounded context; scales to monorepos with a context map.
  2. Grill session enhancements — Challenges language usage against the existing glossary, sharpens fuzzy language, discusses concrete scenarios, cross-references with code, and updates context.md as you go.
  3. Architectural Decision Records (ADRs) — Simple markdown files documenting non-obvious decisions. Created only when a decision is: hard to reverse, surprising without context, and the result of a real trade-off.

Live Demo — Grilling in Action ▶ 8:28

Matt demos adding a "pitch" entity to his app (courses/lessons/videos platform). The pitch concept: Mr. Beast-style packaging — create the video title/description/framing before the actual content.

What Happens During the Session

"This might seem like bike-shedding, but this language affects every variable name, every file name in the generated code."

The context.md Updates ▶ 11:14

After the grilling session, Claude updates context.md with new entries:

The language becomes precise and persistent across all future sessions.

Benefits of Grill with Docs ▶ 12:25

Three concrete benefits:

  1. Concise replies — AI uses fewer tokens because shared vocabulary eliminates verbose re-descriptions
  2. Aligned thinking traces — AI's internal reasoning uses the same language, so fewer tokens spent thinking = more aligned with your intention
  3. Easier to navigate code — Code mirrors the documented language, making search and navigation natural
"These are the same benefits described in Domain-Driven Design. The same techniques that work with humans also work with AI."

When to Use Which Skill ▶ 13:27

Amazing non-coding use case: someone wrote a eulogy using Grill Me to surface stories about their mom.

🔑 Key Takeaways

  1. Grill Me's weakness: no persistent shared vocabulary between sessions
  2. Ubiquitous language (from DDD) solves this — code, developers, and domain experts use the same terms
  3. context.md is the single source of truth for domain vocabulary
  4. Language decisions cascade into everything: variable names, file names, UI, deletion rules
  5. ADRs capture non-obvious decisions that context.md can't
  6. The payoff compounds: 4–5 sessions in, Claude "magically aligns with your thoughts"
  7. Concise replies + aligned thinking + navigable code = the three concrete benefits
  8. DDD techniques work just as well with AI as with human teams
  9. Grill with Docs for codebases, Grill Me for everything else
  10. "Getting the language right is absolutely crucial for feeling aligned with the AI"

⏱ Timestamp Index

0:00 The Grill Me phenomenon
1:03 Where Grill Me falls short
2:06 The re-explanation problem
3:30 Ubiquitous language from DDD
4:41 Grill with Docs: the new skill
5:03 context.md and bounded contexts
5:42 Grill session enhancements
6:02 User feedback: "magically aligned"
7:27 Architectural Decision Records
8:28 Live demo: adding a pitch entity
9:12 Terminology collision detection
10:05 Language decisions cascade into code
11:14 context.md updates after grilling
12:25 Three benefits: concise, aligned, navigable
13:27 When to use which skill
14:27 Newsletter and skill updates