Nate Herk stress-tests the open-source GLM 5.2 model inside Claude Code — head-to-head design, homework, and research tasks against Opus 4.8 — then shows the full setup: a z.ai plan, an API key, and rerouting Claude Code's ANTHROPIC_BASE_URL in settings.local.json so the same harness runs a model that's roughly five times cheaper.
Nate Herk | AI Automation16 minTranscript found
Quick learning frame
Read this before watching.
Creative automation uses agents to accelerate production while keeping human taste in story, pacing, selection, and critique.
New playlist item from Nate Herk | AI Automation; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Skill you build: The ability to route the right model to the right task — using a cheap open-source model like GLM 5.2 for the ~80% of knowledge work that doesn't need heavy reasoning while reserving frontier models like Opus for judgment-heavy steps — and to wire an alternative model into the Claude Code harness.
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.
01Brief
02Source
03Generation
04Selection
05Edit
06Taste Review
Deep lesson
Turn this video into working knowledge.
3,857 cleaned transcript words reviewed across 1,024 timed caption segments.
Thesis
GLM 5.2 in Claude Code is Blowing My Mind teaches a practical creative automation move: Nate Herk stress-tests the open-source GLM 5.2 model inside Claude Code — head-to-head design, homework, and research tasks against Opus 4.8 — then shows the full setup: a z.ai plan, an API key, and rerouting Claude Code's ANTHROPIC_BASE_URL in settings.local.json so the same harness runs a model that's roughly five times cheaper.
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.
0:32
Cheap can compete
“bit long, but this was the session that we did it in. It was GLM 5.2 1 million context. As you can see right here, it used about 357,000 tokens. But, as I've been playing around with it...”
GLM 5.2 (1M context) one-shot a landing page in 3:59 versus Opus 4.8's 14:59 with a comparable result at roughly one-fifth the token cost, and it even edited the video's intro from raw footage via a single /goal — though that took 1h15m and ~357k tokens, showing speed is task-dependent: more reasoning means slower GLM. Run the same one-shot design prompt through a cheap model and a frontier model, time both, and score the outputs blind before you assume you need the expensive one.
6:49
Split work by reasoning
“100% be comfortable with GLM 5.2 doing this because I felt comfortable with the way that I orchestrated that storm skill. Bunch of different agents, bunch of different verification checks. And that is way more important ultimately than...”
After a Codex-judged homework test (Opus won on a subtle duplicate-record edge case) and a multi-agent STORM research run, his rule emerges: trust GLM 5.2 for gathering data, opinions, and sources inside a well-orchestrated skill with verification checks — the orchestration matters more than the model — but lean on Opus for thinking through what the data means; only 10–20% of daily tasks need frontier power. Map one of your multi-step workflows and label each step 'gathering' or 'judgment', then assign the cheap model to the first group and the frontier model to the second.
10:52
Swap the engine
“Cloud Code or Open Code or Hermes Agent or wherever you want to plug it in, you're going to click on this button up in the top right and that's going to take you to the actual API...”
Setup is just config: get a z.ai plan ($16/$64/$144 a month, with 5-hour and weekly quotas like Claude's), create an API key, then set ANTHROPIC_BASE_URL to Z's API plus the auth token and default models in .claude/settings.local.json — and because settings are per-directory, one folder can launch Claude Code on GLM while another stays on your Opus Max plan. Create a test directory with a settings.local.json that reroutes Claude Code to an alternative model endpoint, launch it there, and confirm the model banner changed while your other projects stay untouched.
01
Brief
Start with this video's job: Nate Herk stress-tests the open-source GLM 5.2 model inside Claude Code — head-to-head design, homework, and research tasks against Opus 4.8 — then shows the full setup: a z.ai plan, an API key, and rerouting Claude Code's ANTHROPIC_BASE_URL in settings.local.json so the same harness runs a model that's roughly five times cheaper. Treat "Brief" as the outcome you are trying to make visible, not a topic label. Anchor it to 0:32, where the video says: “bit long, but this was the session that we did it in. It was GLM 5.2 1 million context. As you can see right here, it used about 357,000 tokens. But, as I've been playing around with it...”
02
Source
Use "Source" to locate the part of the creative automation workflow the video is demonstrating. Ask what changes in your real setup if this claim is true. Anchor it to 6:49, where the video says: “100% be comfortable with GLM 5.2 doing this because I felt comfortable with the way that I orchestrated that storm skill. Bunch of different agents, bunch of different verification checks. And that is way more important ultimately than...”
03
Generation
Turn "Generation" into the reusable artifact for this lesson: A creative workflow board with critique criteria and review checkpoints. This is where watching becomes something you can inspect and reuse.
04
Selection
Use "Selection" 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
Edit
Use "Edit" 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
Taste Review
Use "Taste Review" 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 creative workflow board with critique criteria and review checkpoints..
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: Nate Herk stress-tests the open-source GLM 5.2 model inside Claude Code — head-to-head design, homework, and research tasks against Opus 4.8 — then shows the full setup: a z.ai plan, an API key, and rerouting Claude Code's ANTHROPIC_BASE_URL in settings.local.json so the same harness runs a model that's roughly five times cheaper.
02
Explain the practical stakes without hype: New playlist item from Nate Herk | AI Automation; queued for transcript-backed review, topic mapping, and a practical learning artifact.
03
Map the idea onto the Brief -> Source -> Generation -> Selection -> Edit -> Taste Review sequence and name the weakest link.
04
Produce the artifact and include the evidence that proves it: A creative workflow board with critique criteria and review checkpoints.
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: GLM 5.2 in Claude Code is Blowing My Mind
- URL: https://www.youtube.com/watch?v=2OD14-0cot4
- Topic: Creative Automation
- My current learning frame: Wire GLM 5.2 into Claude Code via a per-project settings.local.json, then run one identical real task in a GLM directory and an Opus directory, and record time, cost, and quality to build your own model-routing rule of thumb.
- Why this matters: New playlist item from Nate Herk | AI Automation; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Transcript anchors from this exact video:
- 0:32 / Evidence 1: "bit long, but this was the session that we did it in. It was GLM 5.2 1 million context. As you can see right here, it used about 357,000 tokens. But, as I've been playing around with it..."
- 3:32 / Evidence 2: "did a few more/goal prompts. And I hate when the terminal does this, but this first one that I did was I did /goal and I literally said like, "Hey, get creative. Show me how good your design..."
- 5:03 / Evidence 3: "miss as far as when is GLM 5.2 actually faster. Typically, the more reasoning, the slower it's going to be. And so, just remember, cloud code is a harness. It's a harness for AI models, and typically cloud..."
- 6:49 / Evidence 4: "100% be comfortable with GLM 5.2 doing this because I felt comfortable with the way that I orchestrated that storm skill. Bunch of different agents, bunch of different verification checks. And that is way more important ultimately than..."
- 10:52 / Evidence 5: "Cloud Code or Open Code or Hermes Agent or wherever you want to plug it in, you're going to click on this button up in the top right and that's going to take you to the actual API..."
- 12:29 / Evidence 6: "that kind of stuff. And you can also set environment variables. So if you guys have activated agent teams, this might live here like mine or it might live globally. Wherever it lives, that's an environment variable. And..."
- 14:53 / Evidence 7: "bringing you guys a ton more content on local models and maybe even some other stuff like open code because you don't always want to be locked into maybe cloud code's hardest. So, please let me know in..."
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 creative workflow board with critique criteria and review checkpoints.
5. Include:
- a plain-English definition of the core idea
- a diagram or structured model using this sequence: Brief -> Source -> Generation -> Selection -> Edit -> Taste Review
- 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 "GLM 5.2 in Claude Code is Blowing My Mind", not a generic Creative Automation 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.
Creative AI removes the need for taste.
It increases the need for taste because output volume explodes.
The best prompt is enough.
References, critique, iteration, and post-production matter just as much.
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 creative workflow board with critique criteria and review checkpoints..
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.
How much cheaper is GLM 5.2 than Opus 4.8 per token, and what were the head-to-head landing page results?
What division of labor between GLM 5.2 and Opus does the creator land on after his tests?
How do you plug GLM 5.2 into Claude Code, and how can different projects use different models?
Source shelf
Use the video as a doorway, then verify with primary sources.
