This video extends Matt Pocock's Grill Me planning skill with an iterative adversarial code review from Codex, so Claude Code's plan gets challenged by a neutral second model over up to five rounds instead of Claude grading its own work. It demos the resulting Grill Me Codex skill by planning an email-capture gate for a website and shows Codex catching real security holes and false fixes before any code ships.
Chase AIWatchTranscript found
Quick learning frame
Read this before watching.
Coding-agent workflow is the loop of inspect, plan, edit, verify, summarize, and route the next task to the right tool.
New playlist item from Chase AI; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Skill you build: The ability to set up a two-model adversarial planning loop — Claude Code drafting and Codex critiquing across multiple rounds — so a non-engineer can trust a plan that both leading AI tools have signed off on.
Watch for the shift from claim to mechanism. The learning value is the point where the transcript reveals a repeatable action, tool boundary, context move, review habit, or artifact.
Concept diagram
Where this video fits.
01Inspect
02Plan
03Edit
04Verify
05Review
06Route
Deep lesson
Turn this video into working knowledge.
2,646 cleaned transcript words reviewed across 754 timed caption segments.
Thesis
I Updated /grill-me And Solved Claude Code teaches a practical codex + claude workflows move: This video extends Matt Pocock's Grill Me planning skill with an iterative adversarial code review from Codex, so Claude Code's plan gets challenged by a neutral second model over up to five rounds instead of Claude grading its own work. It demos the resulting Grill Me Codex skill by planning an email-capture gate for a website and shows Codex catching real security holes and false fixes before any code ships.
The goal is not to remember the video. The goal is to extract the operating principle, tie it to timestamped evidence, test how far the claim transfers, and make something reusable.
0:00
Models grade themselves kindly
“Plan mode is not enough. Skills like Matt Pocock's Grill Me or even larger orchestration layers like GSD or Superpowers are all trying to solve the same problem. Take that fuzzy idea in your head and turn it...”
Planning tools like Grill Me, GSD, and Superpowers close the gap between your fuzzy idea and Claude Code, but they still rely on one model to plan, build, and evaluate its own work — and Claude, as Anthropic itself has noted, reliably praises whatever it wrote. If you lack a technical background you can't judge the plan either, so the fix is a neutral third party: an iterative Codex review bolted onto Grill Me. Take a plan Claude Code recently produced for you and ask a second model (Codex, or any other LLM) to critique it; note how many concrete issues the outside reviewer finds that Claude's self-assessment missed.
6:21
Two files, five rounds
“database, which already exists. So, we're not just creating some feature from thin air. It needs to take a look at the code base that already exists to make it coherent. So, this is the prompt I'm giving...”
After the standard Grill Me question phase, the Codex portion creates two markdown files: plan.md, the source of truth the build runs from, and plan-review-log.md, where Claude Code and Codex argue through up to five iterations. The review is headless but Codex keeps the session ID, so it remembers the whole back-and-forth rather than starting each round from a blank slate. Run Grill Me Codex (or replicate the pattern) on a small feature and read the plan-review-log.md end to end to see exactly which objections Codex raised and how the plan changed each round.
8:00
Catching false fixes
“files for us. We have the plan.md and then the plan review log. So, the plan.md is the source of truth for what we're going to create. This is what our final deliverable is going to be. The...”
In the demo's email-gate plan, Codex found 11 issues in round one — including an unbounded client skill slug, a case-sensitive dedupe bypass, a raw list-bombing vector, and a table-scanning rate limit — then in round two flagged Claude's 'fixes' that weren't real, like a double opt-in that was claimed but never wired and an expression-index dedupe Supabase JS can't target, before approving in round three with only three low-level nits. For your next AI-generated plan, write down each issue the reviewer raises, then verify in the follow-up round whether every fix was actually implemented rather than just claimed.
01
Inspect
Start with this video's job: This video extends Matt Pocock's Grill Me planning skill with an iterative adversarial code review from Codex, so Claude Code's plan gets challenged by a neutral second model over up to five rounds instead of Claude grading its own work. It demos the resulting Grill Me Codex skill by planning an email-capture gate for a website and shows Codex catching real security holes and false fixes before any code ships. Treat "Inspect" as the outcome you are trying to make visible, not a topic label. Anchor it to 0:00, where the video says: “Plan mode is not enough. Skills like Matt Pocock's Grill Me or even larger orchestration layers like GSD or Superpowers are all trying to solve the same problem. Take that fuzzy idea in your head and turn it...”
02
Plan
Use "Plan" to locate the part of the codex + claude workflows workflow the video is demonstrating. Ask what changes in your real setup if this claim is true. Anchor it to 6:21, where the video says: “database, which already exists. So, we're not just creating some feature from thin air. It needs to take a look at the code base that already exists to make it coherent. So, this is the prompt I'm giving...”
03
Edit
Turn "Edit" into the reusable artifact for this lesson: A routing matrix for when to use Codex, Claude, browser checks, or manual review. This is where watching becomes something you can inspect and reuse.
04
Verify
Use "Verify" as the application surface. Decide whether the idea touches a browser flow, a local file, a model choice, a source document, a UI, or a review step.
05
Review
Use "Review" to prove the lesson. The evidence should connect back to the video title, transcript anchors, and a concrete output, not a generic best-practice claim.
06
Route
Use "Route" to carry the idea forward: save the prompt, checklist, diagram, or operating rule that would make the next agent run better.
Example
Source-backed work packet
Convert the video into a scoped task that includes the transcript claim, target workflow, acceptance criteria, and proof. The output should be a routing matrix for when to use codex, claude, browser checks, or manual review..
Example
Claim vs. demo brief
Separate what the speaker claims, what the demo actually proves, and what still needs outside verification before you adopt the workflow.
Example
Teach-back module
Transform the lesson into a definition, a mechanism diagram, one misconception, one practice exercise, and a check-for-understanding question.
Do not learn it wrong
Treating the title as the lesson without checking what the transcript actually says.
Letting the prompt drift into generic advice that could apply to any video in the playlist.
Copying the tool setup without identifying the operating principle that transfers to your own stack.
Skipping the artifact, which means the learning never becomes operational or inspectable.
Do not count this as learned until these are true.
01
State the transcript-backed claim in your own words: This video extends Matt Pocock's Grill Me planning skill with an iterative adversarial code review from Codex, so Claude Code's plan gets challenged by a neutral second model over up to five rounds instead of Claude grading its own work. It demos the resulting Grill Me Codex skill by planning an email-capture gate for a website and shows Codex catching real security holes and false fixes before any code ships.
02
Explain the practical stakes without hype: New playlist item from Chase AI; queued for transcript-backed review, topic mapping, and a practical learning artifact.
03
Map the idea onto the Inspect -> Plan -> Edit -> Verify -> Review -> Route sequence and name the weakest link.
04
Produce the artifact and include the evidence that proves it: A routing matrix for when to use Codex, Claude, browser checks, or manual review.
Put it into practice
Give this grounded prompt to Codex or Claude after watching.
You are helping me turn one specific YouTube video into real, durable learning.
Source video:
- Title: I Updated /grill-me And Solved Claude Code
- URL: https://www.youtube.com/watch?v=ENCRw5-uJBA
- Topic: Codex + Claude Workflows
- My current learning frame: Install the Grill Me Codex skill (you need Codex and a $20/month OpenAI plan, or swap in a local model), plan one small gated feature for a real site, and only start building once the plan-review-log shows Codex approving the plan.
- Why this matters: New playlist item from Chase AI; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Transcript anchors from this exact video:
- 0:00 / Evidence 1: "Plan mode is not enough. Skills like Matt Pocock's Grill Me or even larger orchestration layers like GSD or Superpowers are all trying to solve the same problem. Take that fuzzy idea in your head and turn it..."
- 1:49 / Evidence 2: "going to give you a better insight into what you're actually trying to build because whether you want to admit it or not, you probably suck at actually articulating what you want. And if you can't articulate what..."
- 4:07 / Evidence 3: "In comes Codex. And this Codex review is what I've added to Pocock skills and it's what I'm going to be giving you today. So, the first half is exactly the same as girl me. Questions back and..."
- 6:21 / Evidence 4: "database, which already exists. So, we're not just creating some feature from thin air. It needs to take a look at the code base that already exists to make it coherent. So, this is the prompt I'm giving..."
- 8:00 / Evidence 5: "files for us. We have the plan.md and then the plan review log. So, the plan.md is the source of truth for what we're going to create. This is what our final deliverable is going to be. The..."
- 9:32 / Evidence 6: "And here on round three, we see that the verdict is now approved. It's at this point that Codex and Claude Code are now on the same page. Codex has still flagged a couple things, but they're just..."
- 11:35 / Evidence 7: "downloaded, which is relatively simple to do and there's no reason you need anything beyond the $20 month open AI plan to get a lot out of this. This system we've created is also something you could easily..."
Your task:
1. Use the transcript anchors above as the primary source packet. If you add outside context, label it clearly as outside context and keep it secondary.
2. Create a source-check table with columns: timestamp, claim, what the demo proves, confidence, and what still needs verification.
3. Extract the actual teachable claims from the video. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A routing matrix for when to use Codex, Claude, browser checks, or manual review.
5. Include:
- a plain-English definition of the core idea
- a diagram or structured model using this sequence: Inspect -> Plan -> Edit -> Verify -> Review -> Route
- 3 concrete examples that apply the video idea to real agentic work
- 2 failure modes the video helps prevent
- a checklist I can use the next time I run Codex or Claude
- one practical exercise with a clear done signal
6. Add a "learning transfer" section: what changes in my workflow tomorrow if I actually learned this?
7. Add a "source check" section that cites which transcript anchor supports each major takeaway.
Quality bar:
- Make this specific to "I Updated /grill-me And Solved Claude Code", not a generic Codex + Claude Workflows essay.
- Prefer operational examples, failure modes, and reusable artifacts over broad definitions.
- Call out uncertainty instead of smoothing over weak evidence.
- If evidence is weak, say what transcript segment or timestamp needs review instead of guessing.
- Finish with a concise artifact I could paste into my learning app.
Misconceptions
What to stop believing.
One agent should do every task.
Different tools have different strengths. Routing is part of the workflow.
More context is always better.
Relevant context helps; stale context causes drift and cost.
Practice studio
Learning only counts when you make something.
01
Transcript evidence map
Separate what the video actually says from what you already believe about the topic.
3 source-backed takeaways with timestamps, confidence, and a transfer note.02
One useful artifact
Apply the video to a real workflow and produce a routing matrix for when to use codex, claude, browser checks, or manual review..
A reusable artifact with a done signal and one verification step.03
Teach-back card
Explain the lesson to someone who has not watched the video yet.
A 90-second explanation, one diagram, one example, and one misconception to avoid.
Recall check
Answer first, then reveal — without rewatching.
Why can't Claude Code be trusted to evaluate its own plan, according to the video?
What two markdown files does the Codex review phase create, and what is each for?
What happened across the three review rounds in the email-capture gate demo?
Source shelf
Use the video as a doorway, then verify with primary sources.
