I Tested Every Major AI Model For Deep Thinking. Two Surprised Me.
Argues for using AI not to replace your thinking but to challenge it: feed years of your private data into a local 'living archive' and engineer disagreement so the model attacks your ideas for blind spots, and shares a model bake-off where GLM 5.2 (cloud) and Qwen 3.6 27B (local) beat frontier models for high-level strategic thinking.
Manolo Remiddi34 minTranscript found
Quick learning frame
Read this before watching.
A local runtime lesson is about fit: model, quantization, hardware, endpoint, latency, privacy, tool integration, and task limits.
New playlist item from Manolo Remiddi; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Skill you build: The ability to build a private, disagreement-engineered AI thinking partner from your own archived data run locally, and to judge which models actually reason well for strategy rather than just flatter you.
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.
01Task
02Hardware
03Model/quantization
04Runtime endpoint
05Agent tool loop
06Benchmark task
07Fallback
Deep lesson
Turn this video into working knowledge.
5,241 cleaned transcript words reviewed across 1,682 timed caption segments.
Thesis
I Tested Every Major AI Model For Deep Thinking. Two Surprised Me. teaches a practical local model/runtime move: Argues for using AI not to replace your thinking but to challenge it: feed years of your private data into a local 'living archive' and engineer disagreement so the model attacks your ideas for blind spots, and shares a model bake-off where GLM 5.2 (cloud) and Qwen 3.6 27B (local) beat frontier models for high-level strategic thinking.
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
Engineer disagreement
“Most people are using AI just for coding. And the idea is great, okay? Build your own business, and create your own app, develop your own games. All good, all fun, great, okay? You don't need to be...”
Most AI defaults to agreeing with you, which fortifies your bubble; the real value is a system architected to challenge you, brutally attacking every weakness of an idea so you defend and sharpen it, revealing blind spots and opportunities no yes-man advisor would surface. Take one idea you're currently excited about and prompt an AI to act as a hostile challenger whose only job is to destroy it, then write down every blind spot it exposes.
14:49
Local for private data
“mainly for coding and agentic work, but it is actually incredible. It's incredible how analyze the the problem. It's incredible how well it writes. And it's it's just And plus, it's it's a such a small model, 27...”
Because this depends on years of intimate personal data, you can't trust cloud providers that may train on it, so the speaker keeps a 'living archive' on his own computer and lets only a local model read it; token-per-second speed matters because too slow (he needs 50+/sec) breaks focus, though ~30/sec is comfortable for most. Start a 'living archive' folder on your own machine and add one source of your thinking (a journal entry, transcript, or article), noting how you'd keep it private from cloud training.
30:22
Small models surprised
“thinking and strategy and so on. Anyway, that is something that it's out there that we should check. But what is the point here? So, what I can see, this is the trend. More time pass and more...”
In a blind test feeding the same knowledge and prompt to many models and having another AI rate reasoning quality, the two open-source models won: GLM 5.2 (~750B params) ranked first, and the tiny Qwen 3.6 27B ranked third by the AI and second by the human, beating frontier models like GPT 5.5 whose no-thinking mode was unusable and whose output read too code-like. Run your own blind bake-off: give the same rich-context prompt to two or three models, strip the names, and rank the outputs for depth of reasoning before revealing which model produced each.
01
Task
Start with this video's job: Argues for using AI not to replace your thinking but to challenge it: feed years of your private data into a local 'living archive' and engineer disagreement so the model attacks your ideas for blind spots, and shares a model bake-off where GLM 5.2 (cloud) and Qwen 3.6 27B (local) beat frontier models for high-level strategic thinking. Treat "Task" as the outcome you are trying to make visible, not a topic label. Anchor it to 0:00, where the video says: “Most people are using AI just for coding. And the idea is great, okay? Build your own business, and create your own app, develop your own games. All good, all fun, great, okay? You don't need to be...”
02
Hardware
Use "Hardware" to locate the part of the local model/runtime mechanism the video is demonstrating. Ask what changes in your real setup if this claim is true. Anchor it to 14:49, where the video says: “mainly for coding and agentic work, but it is actually incredible. It's incredible how analyze the the problem. It's incredible how well it writes. And it's it's just And plus, it's it's a such a small model, 27...”
03
Model/quantization
Turn "Model/quantization" into the reusable artifact for this lesson: A local model/runtime fit sheet with hardware constraints, model choice, endpoint setup, task benchmark, and fallback rule. This is where watching becomes something you can inspect and reuse.
04
Runtime endpoint
Use "Runtime endpoint" 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
Agent tool loop
Use "Agent tool 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
Benchmark task
Use "Benchmark task" to carry the idea forward: save the prompt, checklist, diagram, or operating rule that would make the next agent run better.
07
Fallback
Connect "Fallback" to I Tested Every Major AI Model For Deep Thinking. Two Surprised Me. by naming the claim, the evidence, and the artifact it should produce.
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 local model/runtime fit sheet with hardware constraints, model choice, endpoint setup, task benchmark, and fallback rule..
Example
Local model/runtime proof brief
Separate what the speaker claims, what the demo actually proves, and what still needs outside verification before you adopt the local model/runtime pattern.
Example
Teach-back module
Transform the lesson into a definition, a Task -> Hardware -> Model/quantization -> Runtime endpoint -> Agent tool loop -> Benchmark task -> Fallback 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.
using a local model like ChatGPT
ignoring latency/context limits
no benchmark task
Letting the lesson drift into local-model ideology.
Letting the lesson drift into hardware specs without workflow fit.
Letting the lesson drift into benchmarks unrelated to the actual task.
Do not count this as learned until these are true.
01
State the transcript-backed claim in your own words: Argues for using AI not to replace your thinking but to challenge it: feed years of your private data into a local 'living archive' and engineer disagreement so the model attacks your ideas for blind spots, and shares a model bake-off where GLM 5.2 (cloud) and Qwen 3.6 27B (local) beat frontier models for high-level strategic thinking.
02
Explain the practical stakes without hype: New playlist item from Manolo Remiddi; queued for transcript-backed review, topic mapping, and a practical learning artifact.
03
Map the idea onto the Brief -> Source material -> Generation -> Selection -> Edit -> Taste review -> Reusable recipe sequence and name the weakest link.
04
Produce the artifact and include the evidence that proves it: A creative production board with source inputs, prompt recipe, selection criteria, edit pass, and taste-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: I Tested Every Major AI Model For Deep Thinking. Two Surprised Me.
- URL: https://www.youtube.com/watch?v=Zfgl5rA6ehg
- Topic: Creative Automation
- My current learning frame: Assemble a small living archive of your own writing on your machine, feed it with a challenge-me prompt to both a cloud model like GLM 5.2 and a local model like Qwen 3.6 27B, and compare which gives sharper, less sycophantic strategic pushback.
- Why this matters: New playlist item from Manolo Remiddi; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Transcript anchors from this exact video:
- 0:00 / Evidence 1: "Most people are using AI just for coding. And the idea is great, okay? Build your own business, and create your own app, develop your own games. All good, all fun, great, okay? You don't need to be..."
- 3:58 / Evidence 2: "thing, okay? How can we make it so wide that we don't feel constrained with this the limit of a narrow-minded system? That is our job. And we do this by breaking this constant bubble that we build."
- 7:33 / Evidence 3: "There are a few elements that are extremely important. One, like I explained, you need to collect your data. You need to have a place where everything goes there. I'm going to explain what kind of software you..."
- 12:24 / Evidence 4: "my computer. Now, who has access to this? My local AI, okay? My local model. That is the only safe way to do it. And at the moment, I've been using different kind of model uh because I'm..."
- 14:49 / Evidence 5: "mainly for coding and agentic work, but it is actually incredible. It's incredible how analyze the the problem. It's incredible how well it writes. And it's it's just And plus, it's it's a such a small model, 27..."
- 18:40 / Evidence 6: "when we when I explained the the living archive, okay? This long-term memory. Why is so important? Is because the model, which is extremely important, but is nothing, okay? It's nothing as unless has our knowledge. So, the..."
- 30:22 / Evidence 7: "thinking and strategy and so on. Anyway, that is something that it's out there that we should check. But what is the point here? So, what I can see, this is the trend. More time pass and more..."
Video-aware target:
- Prompt lane: Local model/runtime
- Mechanism to extract: Identify why the local setup works or fails for this specific agent task, not whether local models are generally good.
- Artifact to produce: A local model/runtime fit sheet with hardware constraints, model choice, endpoint setup, task benchmark, and fallback rule.
- Artifact must include: hardware; runtime; model/quantization; endpoint; agent integration; benchmark/fallback
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 why the local setup works or fails for this specific agent task, not whether local models are generally good. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A local model/runtime fit sheet with hardware constraints, model choice, endpoint setup, task benchmark, and fallback rule.
5. Include:
- a plain-English definition of the core idea
- a diagram or structured model using this sequence: Task -> Hardware -> Model/quantization -> Runtime endpoint -> Agent tool loop -> Benchmark task -> Fallback
- answers to these source questions: What machine/runtime is shown? | What task exposes the model limit? | What setup change improves the loop?
- 3 concrete examples that apply the video idea to real agentic work, such as Ollama or LM Studio coding endpoint; MLX Apple Silicon runner; DGX-backed Hermes session
- 2 failure modes the video helps prevent, chosen from the transcript evidence and these likely risks: using a local model like ChatGPT; ignoring latency/context limits; no benchmark task
- a checklist for the next real workflow, focused on: task fit, runtime setup, latency/context, tool loop, fallback
- one practical exercise with a clear done signal: Choose one real coding task and specify the pass/fail benchmark for a local model.
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 Tested Every Major AI Model For Deep Thinking. Two Surprised Me.", 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: local-model ideology; hardware specs without workflow fit; benchmarks unrelated to the actual task.
- 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 local model/runtime fit sheet with hardware constraints, model choice, endpoint setup, task benchmark, and fallback rule..
A reusable artifact with a done signal and one verification step.03
Local model/runtime teach-back card
Explain the local model/runtime 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.
Why does the speaker say you should engineer disagreement rather than let AI mirror you?
Why does the speaker run this on a local model, and what token-per-second speed does he personally need?
Which two models topped the blind reasoning test, and what was notable about them?
Source shelf
Use the video as a doorway, then verify with primary sources.
