Kimi K3, Qwen 3.8, GLM 5.2 — Is Open Source Winning in 2026?

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

Source video:
- Title: Kimi K3, Qwen 3.8, GLM 5.2 — Is Open Source Winning in 2026?
- URL: https://www.youtube.com/watch?v=ffRn2bTo04Y
- Topic: Creative Automation
- My current learning frame: Take one open-weight model release you're curious about and write a short scorecard against the three-question framework (weights available now, license permissiveness, independently verified benchmarks) before deciding whether to trust its 'open source' label.
- Why this matters: New playlist item from Panda Making Money; queued for transcript-backed review, topic mapping, and a practical learning artifact.

Transcript anchors from this exact video:
- 0:00 / Evidence 1: "In a span of about 4 months, five different labs shipped open weight AI models that are genuinely competing with the best closed systems in the world. That is not a small claim, and it is not something..."
- 4:54 / Evidence 2: "you're going to see repeated across this entire video. Open-source labs are moving so fast that even their production models often stay in a permanent state of preview. Under the hood, V4 Pro is a massive model on..."
- 10:25 / Evidence 3: "parameters using a mixture of experts design where only around 40 billion of those parameters activate per task. It supports a 1 million token context with a maximum output of 128,000 tokens. And like DeepSeek V4 Pro, it..."
- 13:42 / Evidence 4: "agentic models get trained. Normally, when a lab trains a model to use tools, break down tasks, and recover from errors, they build a fixed harness ahead of time, a predefined structure that tells the model how to..."
- 19:21 / Evidence 5: "context window. When it comes to actual performance, the results back up the scale. Independent testing places KimmyK3 fourth among every frontier model tested, open or closed, sitting just behind Claude Fable 5 and GPT 5.60 and ahead..."
- 26:23 / Evidence 6: "through all five models individually. So, let's pull the whole picture together and actually answer the question this video set out to ask. Is open source genuinely winning in 2026 or is this a story that sounds better..."
- 30:49 / Evidence 7: "agent decoding, in cost efficiency, and in the kind of independence that comes from being able to self-host a model instead of depending entirely on someone else's API. We also saw real limits from licensing that isn't quite..."

Video-aware target:
- Prompt lane: Creative automation
- Mechanism to extract: Extract the creative production loop, especially where the human keeps taste, selection, and final judgment.
- Artifact to produce: A creative production board with source inputs, prompt recipe, selection criteria, edit pass, and taste-review checkpoints.
- Artifact must include: brief; source inputs; generation recipe; selection criteria; edit/review checkpoint

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: Extract the creative production loop, especially where the human keeps taste, selection, and final judgment. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A creative production board with source inputs, prompt recipe, selection criteria, edit pass, and taste-review checkpoints.
5. Include:
   - a plain-English definition of the core idea
   - a diagram or structured model using this sequence: Brief -> Source material -> Generation -> Selection -> Edit -> Taste review -> Reusable recipe
   - answers to these source questions: What asset is being produced? | What inputs and tools drive it? | Where does human taste intervene?
   - 3 concrete examples that apply the video idea to real agentic work, such as Claude-generated video campaign; image-to-site workflow; voice or video editing loop
   - 2 failure modes the video helps prevent, chosen from the transcript evidence and these likely risks: mistaking novelty for quality; no source/brief discipline; shipping generated media without taste review
   - a checklist for the next real workflow, focused on: brief, inputs, generation, selection, critique
   - one practical exercise with a clear done signal: Build one reusable creative recipe and define what would make the result rejectable.
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 "Kimi K3, Qwen 3.8, GLM 5.2 — Is Open Source Winning in 2026?", not a generic Creative Automation essay.
- Anchor each creative step to transcript evidence about inputs, model/tool choices, iteration, editing, or critique.
- 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 content advice; tool hype; creative output without selection criteria.
- 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.