Cut your AI cost IN HALF (EASY)

You are helping me turn one specific YouTube video into real, durable learning.

Source video:
- Title: Cut your AI cost IN HALF (EASY)
- URL: https://www.youtube.com/watch?v=1KKB_UiW6ls
- Topic: Agent Architecture
- My current learning frame: Take one real feature, plan it and write a spec with a frontier model, hand the spec to a cheaper coding model to build and open a PR, route the review back to the frontier model, and tally the token savings versus doing it all with the expensive model.
- Why this matters: New playlist item from Matthew Berman; queued for transcript-backed review, topic mapping, and a practical learning artifact.

Transcript anchors from this exact video:
- 1:21 / Evidence 1: "right here, you don't actually need a frontier model anymore. It turns out a good, cheap model can actually write code extremely well. So, for example, for planning, you might use Fable, but for the actual writing of..."
- 4:43 / Evidence 2: "spec, review the PR and you can use let's say Opus 4.8 or even Sonnet. Basically, a really good coding model that is far cheaper than using Fable for everything else. Now, if you have two subscriptions and..."
- 7:03 / Evidence 3: "writing. So, if we're using a cheap model for code writing, which we're showing right here, $2 per million input tokens, $6 per million output tokens, the total cost of building a feature, building a product, drops substantially."
- 9:14 / Evidence 4: "coding of $1.02 compared to $7.50 with Fable. Now, because the spec, the planning had already been done, all that the cheaper, less capable, but still very good at coding model needed to do was just look at..."
- 12:46 / Evidence 5: "ready to use it, you just do type out the feature that you want, and then you're done. And it's going to plan with Fable, delegate out to Codex. Now, besides for Claude and Codex, there are other..."
- 14:33 / Evidence 6: "so, you can see that here. I am using Fable 5 High, but it spun up a bunch of sub-agents that it delegated out work to, and those sub-agents are using Composer 2.5 fast. So, these things become..."
- 17:46 / Evidence 7: "the right task to the right model. And they've even integrated cheap open-source models like GLM 5.2. So, you have this incredibly capable, very efficient, very cheap open-source model that they are using to do the majority of..."

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 one-page agent harness map with tool boundaries and proof signals.
5. Include:
   - a plain-English definition of the core idea
   - a diagram or structured model using this sequence: Intent -> Model -> Harness -> Tools -> Verifier -> Artifact
   - 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 "Cut your AI cost IN HALF (EASY)", not a generic Agent Architecture 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.