Learn AI Harness Engineering in 14 Minutes | With Kimi K3 & Qwen 3.8-Max
Edward Donner explains 'harness engineering', the scaffolding of workflow, state, sub-agents, permissions/sandboxing, observability, and self-improvement built around an LLM agent, then demonstrates it live by building a sales CRM with Qwen 3.8 Max and Kimi K3 in Open Code, including a 'recursive self-improvement' step where the agent rewrites its own self-improvement instructions.
Edward Donner14 minTranscript found
Quick learning frame
Read this before watching.
A model becomes useful when it is wrapped in a harness: tools, state, permissions, memory, routing, and verification.
New playlist item from Edward Donner; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Skill you build: The ability to design an agent harness (process file, sub-agent roles, sandbox, review loop, and a self-improvement step) so a coding agent produces better results on hard tasks and gets better at improving itself over repeated runs.
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.
01User intent
02Model role
03Tool surface
04State and memory
05Verification loop
06Reusable operating rule
Deep lesson
Turn this video into working knowledge.
2,564 cleaned transcript words reviewed across 748 timed caption segments.
Thesis
Learn AI Harness Engineering in 14 Minutes | With Kimi K3 & Qwen 3.8-Max teaches a practical agent harness move: Edward Donner explains 'harness engineering', the scaffolding of workflow, state, sub-agents, permissions/sandboxing, observability, and self-improvement built around an LLM agent, then demonstrates it live by building a sales CRM with Qwen 3.8 Max and Kimi K3 in Open Code, including a 'recursive self-improvement' step where the agent rewrites its own self-improvement instructions.
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.
0:31
CRM built for $8.71
“this project, I brought back one of my favorites. It is the CRM platform asking a coding agent to build a sales CRM like a Salesforce or a Pipedrive. Build the whole thing from scratch for me to...”
Using Kimi K3 and Qwen 3.8 Max together in an agent harness, Edward built a full personal sales CRM (dashboard with expected vs. full pipeline revenue, Kanban pipeline board with drag-and-drop, contacts, organizations, deals) in about the same time as a prior Fable build but for a total spend of $8.71, calling it his best CRM build so far. Pick a familiar demo project (like a CRM) and track total token spend and build time the next time you build it with an agent, so you have a cost baseline to compare harness setups against.
7:20
Sub-agents and sandboxed workflow
“the workflow laid out. Now, I'm going to be using open code today, and for open code, there's a JSON file called open code.json that describes the different sub-agents. And this is how I've set up my sub-agent...”
The harness is defined by a process.md workflow file (build, verify against success criteria, use a sub-agent to review, incorporate feedback, self-improve) and an opencode.json that sets up a primary build agent on Qwen 3.8 Max plus a separate product-review sub-agent on Kimi K3 with read-only permissions, all running inside a VS Code dev container sandbox with a Vercel agent-browser skill installed so the agent can test its own work. Write your own process.md-style workflow file for a project: list the build steps, name which sub-agent reviews the output, and state what permissions that reviewer does and doesn't have.
10:20
Recursive self-improvement
“it is. And now I'm going to press shift tab to flip over to my build and improve primary agent, and we're ready to go. And I'm just going to say to it like, "Go ahead, build the...”
Beyond having the agent update its own process.md with learnings after each run, Edward's self-improve.md includes a step instructing the agent to rewrite self-improve.md itself so it gets better at improving itself; the run produced concrete meta-learnings like categorizing learnings, recording detection methods, writing them as executable instructions, and reproducing bugs with real user events before assuming the app is broken. Add one instruction to your own agent's self-improvement file that asks it to also revise the self-improvement instructions themselves, then compare the notes it writes before and after a run.
01
User intent
Start with this video's job: Edward Donner explains 'harness engineering', the scaffolding of workflow, state, sub-agents, permissions/sandboxing, observability, and self-improvement built around an LLM agent, then demonstrates it live by building a sales CRM with Qwen 3.8 Max and Kimi K3 in Open Code, including a 'recursive self-improvement' step where the agent rewrites its own self-improvement instructions. Treat "User intent" as the outcome you are trying to make visible, not a topic label. Anchor it to 0:31, where the video says: “this project, I brought back one of my favorites. It is the CRM platform asking a coding agent to build a sales CRM like a Salesforce or a Pipedrive. Build the whole thing from scratch for me to...”
02
Model role
Use "Model role" to locate the part of the agent harness mechanism the video is demonstrating. Ask what changes in your real setup if this claim is true. Anchor it to 7:20, where the video says: “the workflow laid out. Now, I'm going to be using open code today, and for open code, there's a JSON file called open code.json that describes the different sub-agents. And this is how I've set up my sub-agent...”
03
Tool surface
Turn "Tool surface" into the reusable artifact for this lesson: A one-page agent harness map with tool boundaries, state ownership, and proof signals. This is where watching becomes something you can inspect and reuse.
04
State and memory
Use "State and memory" 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
Verification loop
Use "Verification loop" 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
Reusable operating rule
Use "Reusable operating rule" to carry the idea forward: save the prompt, checklist, diagram, or operating rule that would make the next agent run better.
Example
Source-backed artifact packet
Convert the video into a scoped artifact request that includes the transcript claim, mechanism, acceptance criteria, and proof. The output should be a one-page agent harness map with tool boundaries, state ownership, and proof signals..
Example
Agent harness proof brief
Separate what the speaker claims, what the demo actually proves, and what still needs outside verification before you adopt the agent harness pattern.
Example
Teach-back module
Transform the lesson into a definition, a User intent -> Model role -> Tool surface -> State and memory -> Verification loop -> Reusable operating rule 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.
treating model choice as architecture
ignoring tool permissions
missing verification evidence
Letting the lesson drift into generic agent definitions.
Letting the lesson drift into model leaderboard claims.
Letting the lesson drift into tool list without operating boundaries.
Do not count this as learned until these are true.
01
State the transcript-backed claim in your own words: Edward Donner explains 'harness engineering', the scaffolding of workflow, state, sub-agents, permissions/sandboxing, observability, and self-improvement built around an LLM agent, then demonstrates it live by building a sales CRM with Qwen 3.8 Max and Kimi K3 in Open Code, including a 'recursive self-improvement' step where the agent rewrites its own self-improvement instructions.
02
Explain the practical stakes without hype: New playlist item from Edward Donner; queued for transcript-backed review, topic mapping, and a practical learning artifact.
03
Map the idea onto the User intent -> Model role -> Tool surface -> State and memory -> Verification loop -> Reusable operating rule sequence and name the weakest link.
04
Produce the artifact and include the evidence that proves it: A one-page agent harness map with tool boundaries, state ownership, and proof signals.
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: Learn AI Harness Engineering in 14 Minutes | With Kimi K3 & Qwen 3.8-Max
- URL: https://www.youtube.com/watch?v=t1PXvg2FceU
- Topic: Creative Automation
- My current learning frame: Set up a minimal harness for one of your own coding-agent projects: a process.md workflow, a separate read-only review sub-agent, and a self-improve step, then run it twice and compare what the agent's own process notes look like after each run.
- Why this matters: New playlist item from Edward Donner; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Transcript anchors from this exact video:
- 0:31 / Evidence 1: "this project, I brought back one of my favorites. It is the CRM platform asking a coding agent to build a sales CRM like a Salesforce or a Pipedrive. Build the whole thing from scratch for me to..."
- 2:11 / Evidence 2: "agent one more time. An agent is an LLM that's been equipped with tools and it's running in a loop in order to achieve a goal. That's the definition of an agent. And another way you sometimes hear..."
- 3:41 / Evidence 3: "harness engineering? So obviously the first part of harness engineering, of getting the right scaffolding, is doing everything we're already doing around context engineering, thinking about the prompts, the tools, the memory that we're equipping an LLM with..."
- 5:14 / Evidence 4: "in my last video on the Groove Box, if you haven't seen that. That is observability and evaluations. And then last but not least, self-improvement. Self-improvement is about building features into your agent harness so that it's able..."
- 7:20 / Evidence 5: "the workflow laid out. Now, I'm going to be using open code today, and for open code, there's a JSON file called open code.json that describes the different sub-agents. And this is how I've set up my sub-agent..."
- 10:20 / Evidence 6: "it is. And now I'm going to press shift tab to flip over to my build and improve primary agent, and we're ready to go. And I'm just going to say to it like, "Go ahead, build the..."
- 13:13 / Evidence 7: "the effort we put in as part of Harness engineering. And you should try doing some Harness engineering yourself. There'll be instructions in the description below to bring up this repo, play around with it, get it to..."
Video-aware target:
- Prompt lane: Agent harness
- Mechanism to extract: Identify what surrounding harness makes the model more useful than chat alone.
- Artifact to produce: A one-page agent harness map with tool boundaries, state ownership, and proof signals.
- Artifact must include: model role; tools; state/memory; permission boundary; verification proof
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, transcript support, what the demo proves, confidence, and what still needs verification.
3. Extract the actual teachable mechanism from the video: Identify what surrounding harness makes the model more useful than chat alone. 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, state ownership, and proof signals.
5. Include:
- a plain-English definition of the core idea
- a diagram or structured model using this sequence: User intent -> Model role -> Tool surface -> State and memory -> Verification loop -> Reusable operating rule
- answers to these source questions: What does the video claim the agent can do? | What surrounding system makes that claim plausible? | What proof is shown instead of merely asserted?
- 3 concrete examples that apply the video idea to real agentic work, such as a repo-editing harness; a local research assistant; a recurring refresh agent
- 2 failure modes the video helps prevent, chosen from the transcript evidence and these likely risks: treating model choice as architecture; ignoring tool permissions; missing verification evidence
- a checklist for the next real workflow, focused on: tool boundaries, state ownership, done signal, recovery path
- one practical exercise with a clear done signal: Map one current coding workflow as a harness and mark the first missing proof 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 "Learn AI Harness Engineering in 14 Minutes | With Kimi K3 & Qwen 3.8-Max", not a generic Creative Automation essay.
- Tie each harness element to a transcript anchor that names a tool, state boundary, permission, model behavior, or verification step.
- Prefer operational examples, failure modes, and reusable artifacts over broad definitions.
- Call out uncertainty instead of smoothing over weak evidence.
- Avoid these generic drifts: generic agent definitions; model leaderboard claims; tool list without operating boundaries.
- If evidence is weak or missing, stop and 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 one-page agent harness map with tool boundaries, state ownership, and proof signals..
A reusable artifact with a done signal and one verification step.03
Agent harness teach-back card
Explain the agent harness mechanism 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.
What was the total token spend for building the sales CRM demoed in the video, and which two models were used?
What role does the Kimi K3 sub-agent play in the harness, and what permission restriction does it have?
What makes the self-improve step 'recursive self-improvement' rather than ordinary self-improvement?
Source shelf
Use the video as a doorway, then verify with primary sources.
