Qwen3.8 27B: Same Model, Three Harnesses, One Clear Winner
Credibility score: 53/100 — Mixed Credibility. Several questionable claims detected. Watch with healthy skepticism.
BSmeter analyzed "Qwen3.8 27B: Same Model, Three Harnesses, One Clear Winner" and rated it 53/100 for credibility (a BS score of 47/100 — mixed credibility), on 2026-09-16. Its weakest claim — "Compares DeepSeek's output to a hypothetical ChatGPT 5.5, a model that doesn't exist. 💀" — scored 20/100 and was flagged as false equivalence. 64 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Of 64 claims analyzed: 1 scored under 40, 47 between 40 and 69, and 16 at 70 or above.
Claims analyzed
Setting the stage for a comparison of AI harnesses — straight talk, no tricks. 😈 — No Frame (75/100)
At 0:00
Just laying out the plan, mortal. A rare moment of clarity before the inevitable human mess. 🔥
Why this score: The speaker is simply introducing the video's objective: to test and compare three different AI harnesses with the Qwen 3.8 27B model on coding tasks. No hidden agenda, just a straightforward setup.
Original quote: “Today we're going to be testing out the top three harnesses to use with Quen 3.827B. Now I'm going to be using this on my 5090 and we're going to see which one does the very best at each coding task.”
Defining 'harness' with specific examples like Codeex and Claude Code. 😈 — No Frame (75/100)
At 0:15
Explaining the jargon for the uninitiated. A basic definition, nothing to twist here. 💀
Why this score: The speaker provides a clear and concise definition of what an AI 'harness' is, using well-known examples like Codeex and Claude Code. This is helpful context for the audience and is presented without any manipulative framing.
Original quote: “Now for those of you that don't know what a harness is, a harness is essentially a program like Codeex or Claude Code that are used to manipulate the model. For most tasks like coding, the harness will manipulate the model in a way where it'll use tool calling”
Introducing three harnesses for testing — straightforward setup. — No Frame (75/100)
At 0:35
Just laying out the tools for the experiment. No tricks here, mortal. Just the setup. 😈
Why this score: The speaker is simply listing the three different AI harnesses they plan to test in the video. This is a clear, direct statement of intent, not a claim that requires external verification or contains manipulative framing.
Original quote: “So today I'm going to be testing out Hermes agent. We're going to be testing out a very popular harness called high coating agent. And we're also going to be testing out Deep Seek Harness, which just came out a couple weeks ago.”
Explaining the first test: a custom 'massive' Minecraft prompt. — No Frame (75/100)
At 0:53
He's just setting the stage for his custom test. A 'massive prompt' is his baby, not a universal truth. 😈
Why this score: The speaker is describing the first of two tests they will conduct: a custom-designed 'massive prompt' intended to generate a playable Minecraft game. This is a clear explanation of their methodology and a personal claim about their prompt's design, not a factual assertion to be verified. It's his test, his rules.
Original quote: “Now, we're going to be doing two different tests today with these harnesses. Now, the first one is a Minecraft test. It is a massive prompt that I have come up with in order to see if it can actually produce a playable version of Minecraft with all of the different functions that you would see in a…”
Describing the second test: generating a 'complicated' color palette website. — No Frame (75/100)
At 1:13
Another test, another custom challenge. He's just telling you what's coming, mortal. 😈
Why this score: The speaker is outlining the second test, which involves generating a 'pretty complicated' color palette generator website. This is a straightforward description of their experimental design, not a claim that is subject to external verification or manipulative framing. It's part of his setup.
Original quote: “And the second test we're going to be doing is to see how well it can generate a specific pretty complicated website. In this case, this is a color pallet generator website. So, let's see what these harnesses can do with Quen 3.827B.”
Claiming only 'Fable 5 with Claude' ever 'oneshot' his extensive Minecraft prompt. — Personal Story (60/100)
At 1:30
He's flexing his 'classic' prompt and its difficulty. It's his personal benchmark, mortal, not a scientific study. 😈
Why this score: The speaker is describing their 'classic Minecraft clone prompt' and stating that only 'Fable 5 with Claude' has ever successfully completed it in one go. This is a personal anecdote about their experience with their own custom prompt and models they've tested. It's not a universally verifiable claim but a statement of their personal testing history, so it's treated as a personal story rather than a fact to be externally verified.
Original quote: “Now, for the first test, we're going to be using my classic Minecraft clone prompt. Now, this is a very extensive long prompt that I use with just about every single model to test it. And to put in perspective, the only model that was ever able to actually oneshot this prompt was Fable 5 with…”
Introducing Pi Coding Agent, citing 'a lot of recommendations' from comments and Reddit. — Anonymous Authority (45/100)
At 2:10
He says 'a lot of people' and 'Reddit' recommended it. That's not data, mortal, that's just vibes. 💀
Why this score: The speaker introduces Pi Coding Agent and supports its 'pretty cool' status by referencing 'a lot of people in the comments' and 'recommendations as well, like on places like Reddit and whatnot.' This relies on vague, unsourced community sentiment rather than specific, verifiable evidence or named authorities. It's a classic appeal to anonymous authority to bolster a claim.
Original quote: “Now, the first harness that I tried with Quen 3.827 2827B was Pi Harness or Pi Coding Agent. Now, this is a pretty cool project. I know a lot of people in the comments were recommending this. I saw a lot of recommendations as well, like on places like Reddit and whatnot.”
Speaker offers their settings and promotes Unslaw Studio as their engine. — Plain Sales Pitch (45/100)
At 2:56
He's sharing his setup, but also subtly pushing 'Unslaw Studio' as the 'engine' he's using. It's a soft sell, mortal. 😈
Why this score: While sharing personal settings is helpful, the repeated mention of 'Unslaw Studio' as 'the engine that I'm using' and 'essentially as the engine' positions it as a recommended tool, subtly promoting it within the context of a technical explanation. It's not a hard sell, but it's definitely a plug.
Original quote: “So, you can copy my settings here. Essentially, what you'll need to do is import your API key from Unsllo Studio. At least that's the engine that I'm using. I'm using Unslaw Studio essentially as the engine and then this is the coding harness that I'm using to test out some different harnesses to…”
Speaker claims Pi harness is 'very similar' to Claude code. — Confidence Mismatch (45/100)
At 3:18
He says 'very similar' but doesn't actually show the Claude code to compare. That's not a comparison, mortal, that's a declaration. 💀
Why this score: The speaker asserts that 'PI harness is very similar to something like cloud code' but doesn't provide a visual or functional comparison to Claude code. The claim of similarity is stated with confidence but lacks direct evidence within the presentation to support the degree of 'very similar' for the viewer to judge.
Original quote: “So PI harness is very similar to something like cloud code in the terminal window. So as you can see here I have a folder that I made just for this PI harness and I have some different projects in here. So what I can do here I'm just going to do like a test folder to show you guys how this works.…”
Speaker demonstrates immediate interaction with Qwen, claiming it's 'just like your actual terminal'. — No Frame (75/100)
At 3:52
He shows it working, mortal. The model responded. No tricks, just a demo. 😈
Why this score: The speaker directly demonstrates the interaction with the Qwen model within the terminal, showing it responding to a 'hi' command. This is a straightforward demonstration of functionality, and the claim that it's 'just like your actual like terminal' is visually supported by the interface shown.
Original quote: “So, as you can see here, it is just like your actual like terminal. Like I can actually interact with Quen right here. Just say hi and respond immediately. It said, "Hey, how are you doing today?" And then it also is very similar in the fact you can do a forward slash and you've got all your…”
The harness is 'manipulating' reasoning levels for efficiency. 😈 — Confidence Mismatch (45/100)
At 4:43
He says 'manipulation' like it's a dark secret — it's just how different systems interact, mortal. 💀
Why this score: The speaker uses 'manipulation' to describe the harness's interaction with the model's reasoning levels. This implies a negative, intentional deception, when it could simply be a difference in how the harness interprets or utilizes the model's capabilities. It's a loaded term for a technical interaction.
Original quote: “Whereas with Quen 3.827B,
the only actual thinking levels that are
baked into the model itself are low,
medium, and X high. So, I do think that
the harness is doing a little bit of a
manipulation here to change some of
these different reasoning levels to make
it a little more or less efficient.”
'Low' reasoning level is the 'best mode' for Qwen 3.827B. 😈 — Confidence Mismatch (45/100)
At 5:28
He declares 'low is the way to go' based on his 'testing' — but shows no data, just a feeling. That's not proof, mortal. 💀
Why this score: The speaker makes a definitive claim ('low is the way to go,' 'best mode') based on his personal experience and 'all the testing that I've done.' While personal experience is valid, without presenting the actual test results, methodology, or comparative data, it's a confident assertion lacking the evidence to back up such a strong, universal recommendation. It's an opinion presented as a definitive conclusion.
Original quote: “And
to be honest, low is the way to go. I
know I got a lot of people telling me
medium is the way to go. From everything
that I've seen and all the testing that
I've done, the low thinking level or low
reasoning level is the best mode for
Quen 3.827B.”
It took 'a lot of back and forth' and 'a couple times' to get the harness working. 😈 — No Frame (75/100)
At 5:49
He's just being honest about the struggle to get the tech working. Refreshing, for a mortal. 🔥
Why this score: The speaker openly admits to the difficulties and iterative process required to get the system working correctly. This is a straightforward account of his experience, acknowledging challenges rather than glossing over them. No manipulative framing detected.
Original quote: “Now, one
thing with this specific harness, it did
take a lot of back and forth to finally
get it into like a playable working
state. It kept essentially like not
showing certain textures like on the
sides of the different blocks. And I
finally prompted it the right way so it
knew what it needed to…”
Introducing the 'low generation' output — just setting the scene. — No Frame (75/100)
At 6:30
Just introducing the output of the 'low generation' model. Nothing tricky here, mortal. 😈
Why this score: The speaker is simply presenting the results from one of the model's outputs, setting the stage for his observations. No claims or persuasive techniques yet.
Original quote: “All right. And this is what the low”
Citing 240 FPS as a 'qualification' from the prompt — but the actual performance is wildly inconsistent. — Missing Context (45/100)
At 6:41
He mentions 240 FPS as a 'qualification' from the prompt, then immediately shows it dropping to 30. That's not a qualification, mortal — that's a wish. 💀
Why this score: The speaker highlights 240 FPS as a 'qualification' from the prompt, implying it was a target or requirement. However, he immediately follows this by stating the performance drops to 30 FPS, which contradicts the idea of it being a consistent 'qualification.' The context of the prompt's actual wording and the model's ability to meet it is missing, making the 'qualification' sound more like an aspirational goal than a met standard.
Original quote: “place. It's going from 240 frames per second, which was one of the qualifications in that full prompt, and”
Acknowledging the 'full Minecraft looking experience' despite initial rendering flaws — a balanced observation. — No Frame (75/100)
At 6:48
He's just giving a fair assessment of what he sees, flaws and all. Refreshing, for a mortal. 😈
Why this score: The speaker is providing a balanced observation, noting both the positive (a 'full Minecraft looking experience') and the negative (initial rendering issues with block sides). This is a straightforward description of the model's output without any obvious rhetorical manipulation.
Original quote: “it's also going down to like 30. So, but it does look like it produced, you know, like a full Minecraft looking experience. Now, like I said, with this one, I had to go back and forth with it pretty often where the sides of these blocks were completely clear, like it was not rendering the sides of…”
Declaring the crafting system 'does not work' after testing it — a direct, verifiable observation. — No Frame (75/100)
At 7:29
He tried it, it failed. Simple as that. No fancy words needed for a broken system. 😈
Why this score: The speaker directly tests the crafting system within the generated environment and concludes that it 'does not work.' This is a clear, observable outcome of his interaction with the model's output, not a claim based on external evidence or loaded language.
Original quote: “So, see, it doesn't look like the crafting system works. The first step where we need actual wood is not even working. And I know I think it's just actually just one. So if we just have one here. No, it doesn't look like it's working. Let's try to put it in all these slots. So yeah, the crafting…”
Highlighting positive aspects like 'infinite generation' and easy water exit, contrasting with past issues. — No Frame (75/100)
At 7:51
He's just pointing out what works well, and how it's an improvement over previous attempts. Fair enough. 😈
Why this score: The speaker is giving credit where it's due, noting the 'infinite generation' and the improved ability to exit water, contrasting it with a previous 'issue.' This is a straightforward comparison and observation of the model's features.
Original quote: “But I will say it does look pretty cool. Looks pretty good. And it does look like it's like an infinite generation here. Uh it's got nice water. Let's see if we can actually get out of the water. Oh wow, you can jump way out of the water. So it does have a way that you can actually jump out of the…”
Concluding with the lack of crafting but speculating about 'zombies and everything built in' — a mix of fact and assumption. — Confidence Mismatch (45/100)
At 8:25
He knows the crafting system is broken, but then 'thinks' it has zombies. One is a fact, the other is a guess. That's not how certainty works, mortal. 💀
Why this score: The speaker confidently states the crafting system 'doesn't have' functionality, which is a direct observation. However, he then shifts to 'I do think it has zombies and everything built in,' which is an assumption or speculation, not a confirmed fact. This creates a mismatch between the certainty of his observations and the confidence of his unverified claims.
Original quote: “Doesn't have the crafting system. And I do think it has zombies and everything built in.”
Dismisses 'medium' performance without showing it first. Classic setup. 😈 — Confidence Mismatch (45/100)
At 8:47
Declares it 'not very impressive' before showing a single frame. That's not a review, mortal, that's a spoiler. 💀
Why this score: The speaker pre-judges the 'medium' generation's performance as 'not very impressive' before presenting any visual evidence. This sets a negative expectation, influencing the viewer's perception before they even see the results, rather than letting the visuals speak for themselves.
Original quote: “it was not very impressive on the medium”
Calls the 'medium' generation 'pretty bad' and 'alien looking.' Just vibes. 👽 — Just Vibes (50/100)
At 9:26
He's just reacting to the visuals, calling them 'alien looking' and 'pretty bad.' It's a subjective take, not a claim. 🤷♂️
Why this score: The speaker is expressing a subjective opinion about the visual quality of the 'medium generation,' describing it as 'pretty bad' and 'alien looking.' This is a personal aesthetic judgment, not a factual claim that can be verified or refuted. It's just his reaction to the visuals on screen.
Original quote: “medium generation. Like I said, it's pretty bad. It has all these like alien looking things. None of the textures even generated. Like I don't even know what these things are.”
Suggests 'X high' 'might have been worse' than 'medium.' More subjective takes. 📉 — Just Vibes (50/100)
At 10:04
Another subjective comparison, saying 'might have been worse.' It's just his feeling about the visuals. 🙄
Why this score: The speaker is offering another subjective assessment, stating that the 'X high' generation 'might have been worse' than the 'medium.' This is an opinion based on his personal perception of the visual output, not a verifiable claim about objective performance metrics. It's a casual observation, not a definitive statement.
Original quote: “And X high actually might have been worse than the medium.”
Declares X high 'might even be worse' based on one specific control issue. Confidence Mismatch. — Confidence Mismatch (45/100)
At 10:30
One specific control issue and suddenly it's 'might even be worse.' That's not a full review, mortal, that's a pet peeve with a megaphone. 💀
Why this score: The speaker extrapolates a general negative assessment ('might even be worse') from a single, specific issue with WSAD controls. This overstates the impact of one problem on overall performance without broader evidence.
Original quote: “S goes forward, W goes backwards, and A goes the wrong way, and D also goes the wrong way. So, WSAD is completely backwards with every single key. So, not very impressed with X high either. In fact, X high might even be worse if we're counting the WSAD controls for PI harness.”
Declares 'low generation' the winner after limited testing. Confidence Mismatch. — Confidence Mismatch (45/100)
At 10:46
One quick test and we've got a 'winner.' That's not a conclusion, mortal, that's a premature coronation. 😈
Why this score: The speaker declares a 'winner' based on the limited testing shown so far. This is a premature conclusion, as more comprehensive evaluation across various tasks would be needed to definitively name a winner.
Original quote: “So overall, it looks like the low generation is the winner. All right.”
Describes extensive back-and-forth with Hermes to achieve basic functionality. No Frame. — No Frame (75/100)
At 11:54
He's just laying out the struggle, mortal. Sometimes, even for me, things take a few tries. 🔥
Why this score: The speaker is simply describing the iterative process required to get the Hermes agent to perform as desired. This is a straightforward account of his experience, not a manipulative framing.
Original quote: “So, I went back and forth with Hermes quite a bit to try to get it to like an actual playable state of Minecraft.”
Concludes Hermes is 'absolute worst' for efficiency and places it third. Confidence Mismatch. — Confidence Mismatch (45/100)
At 12:00
From 'quite a bit' to 'absolute worst' and 'third place' in a blink. That's not analysis, mortal, that's a quick judgment with a definitive label. 💀
Why this score: The speaker moves from describing a 'back and forth' process to a definitive statement that Hermes was the 'absolute worst' in terms of efficiency and places it 'in third place.' While his experience suggests inefficiency, the 'absolute worst' label is a strong, subjective judgment based on limited comparative data.
Original quote: “So, in terms of efficiency, I think Hermes was the absolute worst. And in this case, I probably am going to put Hermes in third place.”
Hermes was 'pretty bad' in early generations — a subjective assessment presented as fact. 💀 — Loaded Language (45/100)
At 12:47
Calling something 'pretty bad' without specific metrics is just vibes, mortal. It's an opinion dressed as a verdict. 😈
Why this score: The speaker uses subjective language ('pretty bad') to describe performance without providing concrete examples or quantifiable data from those 'first few generations.' It's an emotional assessment, not a technical one.
Original quote: “But Hermes overall was honestly pretty bad when it came to like the first few generations.”
Speculating the model 'searched the web' for humans — a guess presented as a possibility. 🧐 — Confidence Mismatch (45/100)
At 13:06
He's guessing the model 'searched the web' for humans. That's not how these things work, mortal. It's a wild leap. 💀
Why this score: The speaker speculates about the model's internal workings ('searched the web and like found like humans') without any evidence or understanding of how the model might have generated humans. It's a confident guess about a technical process he doesn't seem to grasp.
Original quote: “So I don't know if it like searched the web and like found like humans.”
Claiming 'no block breaking animation at all' — a definitive statement on a visual feature. 🚩 — No Frame (75/100)
At 13:12
He's just stating what's clearly not there on screen. Hard to argue with a missing animation, mortal. 😈
Why this score: The speaker is directly observing and stating a visible absence in the game's graphics. This is a straightforward observation of a missing feature, not a loaded claim or misdirection.
Original quote: “Then another thing I noticed is the breaking animation. You can't even see it. Doesn't have any kind of block breaking animation at all as you can see here.”
Declaring the crafting system 'a failure' because it's 'partially working' — a subjective judgment. 💀 — Loaded Language (45/100)
At 14:19
Calling 'partially working' a 'failure' is a bit dramatic, isn't it? It's not a total loss, just not perfect. 😈
Why this score: The speaker uses the strong term 'failure' to describe a system that is 'partially working.' While it might not meet expectations, labeling it a 'failure' is an emotional and subjective judgment rather than a neutral assessment of its functionality.
Original quote: “It looks like it does, but it actually is only partially working. So, this one also is a failure when it came to the full actual like crafting system.”
Declares Hermes third place based on missing zombies. That's a leap, mortal. 💀 — Confidence Mismatch (45/100)
At 14:43
No zombies, therefore third place? That's not a conclusion, that's a tantrum with a thesis. 😈
Why this score: The speaker declares the Hermes harness will come in third place based solely on the absence of zombies during the night cycle. This is a massive leap in logic, as a single missing feature doesn't automatically determine overall ranking in a 'coding test' without a clear rubric or other comparative data. It's an emotional reaction, not an objective assessment.
Original quote: “All right. All right, so I waited for it to turn into night and now it is officially nighttime. So, let's go ahead and look around here. And you know what? I'm not even sure I'm seeing any zombies or anything. Like, I don't even think it got that there should be zombies at night. So, I think Hermes…”
Complains about skeletons when he asked for zombies and creepers. Picky, picky. 🙄 — Just Vibes (50/100)
At 15:16
He asked for specific monsters, got different ones, and now he's complaining. The audacity. 😈
Why this score: The speaker expresses mild annoyance that skeletons appeared instead of the requested zombies and creepers. This isn't a factual claim or a manipulative frame; it's just a personal observation and a bit of a comedic moment about the AI's interpretation. It's a 'just vibes' moment where the humor comes from the specific, slightly off-kilter outcome.
Original quote: “I don't even think I asked for skeletons. I actually just asked for the zombies and the creepers, but okay, they do hurt you.”
Declares the generation 'pretty bad' based on biome variety and tree density. A bit dramatic, mortal. 💀 — Loaded Language (45/100)
At 15:37
From 'not that cool' to 'pretty bad' in one breath. That's not analysis, that's a mood swing. 😈
Why this score: The speaker uses subjective, emotionally charged language ('not really that cool,' 'pretty bad') to describe the game's generation based on limited biome variety and dense trees. While these are observations, the strong negative conclusion is driven by personal preference and loaded phrasing rather than objective metrics or a balanced comparison.
Original quote: “Another thing that I noticed with this generation is it's not really that cool in terms of like different biomes. Like if we go through here, it looks like it's just grass and trees and there's really no like mountains and the trees get like almost too thick. If we look at this here, like look at…”
Declares a 'favorite harness' and 'best one-shot' based on personal experience. — Confidence Mismatch (45/100)
At 16:48
Calling it the 'best one-shot I've ever seen' is a bold claim, mortal, for something so subjective. That's not data, that's just a feeling. 😈
Why this score: The speaker presents a subjective preference ('my favorite harness') and a strong, unqualified superlative ('the best one-shot I've ever seen') as if it's an objective truth. While personal experience is valid, declaring something 'the best' without any comparative metrics or broader testing is a confidence mismatch. It's an opinion stated as fact.
Original quote: “Now, so far, this is still my favorite harness to use with this model. It is the best oneshot from a local model that I've ever seen from this Minecraft prompt.”
Dismisses 'medium' setting as a 'complete failure' without showing the results. — Missing Context (45/100)
At 17:08
He says 'completely failed' on medium, but doesn't show the failure. You're just supposed to take his word for it, mortal. 💀
Why this score: The speaker makes a strong claim that the 'medium' setting 'completely failed' but doesn't provide any visual evidence or specific examples of this failure. He mentions 'a lot of you guys are saying to use medium' but then dismisses it without showing why, leaving the audience to trust his assertion without proof. It's missing the context that would justify such a definitive dismissal.
Original quote: “this is all on low again. By the way, I did do a test on medium and it completely failed. Again, I know a lot of you guys are saying to use medium. I'm just not seeing medium as the best thinking model to use with this model.”
Reiterates 'best one-shot' claim, adding a single exception. — Confidence Mismatch (45/100)
At 18:01
Again with the 'best one-shot I have seen,' mortal. One exception doesn't make it a scientific finding. That's just a slightly more specific opinion. 😈
Why this score: The speaker repeats the claim of this being 'the best one-shot I have seen from any model with any harness,' only adding a single exception ('other than Fable 5'). This still relies heavily on personal observation and subjective judgment rather than objective, comparative data. It's a strong, sweeping statement based on limited, unquantified experience, which is a confidence mismatch.
Original quote: “And this is the original oneshot generation, which this is, like I said, the best oneshot I have seen from any model with any harness other than Fable 5.”
Declares Deep Seek the 'best harness' for any model, based on one example. 💀 — Confidence Mismatch (45/100)
At 18:54
One example, one model, and suddenly it's the 'best' for *any* model. That's not a conclusion, mortal — that's a leap of faith. 😈
Why this score: The speaker presents a single, minor example of a bug fix and then immediately extrapolates this to declare Deep Seek the 'best harness' for 'any of these models.' This is a massive overgeneralization from limited evidence, showing a confidence that far outstrips the data presented.
Original quote: “And like I said, this is the best one I've ever seen from a local model with any harness. And like I said, this is with Deep Seek harness. I think so far, Deep Seek is the best harness that you can use with any of these models.”
Claims 'even Opus won't do that' regarding floating blocks, without showing Opus failing. 🚩 — Confidence Mismatch (45/100)
At 19:37
He says 'Opus won't do that' like it's a known fact, but shows zero evidence. That's not a comparison, mortal — that's just talk. 💀
Why this score: The speaker makes a strong claim that 'even models like Opus won't do that' (referring to floating blocks) to highlight the difficulty and superiority of the current model. However, no demonstration or evidence is provided to show Opus failing to perform this task, making the claim an assertion of superiority without proof.
Original quote: “Like that's pretty tough to do. And even like models like Opus won't do that, which is pretty crazy.”
The game takes a long time to break wood blocks — a direct observation. — No Frame (75/100)
At 20:30
He's just stating what he's seeing in the game. No trickery here, mortal. Just a slow axe. 😈
Why this score: The speaker is describing a direct observation of the game's mechanics, specifically the time it takes to break wood blocks. This is a straightforward statement of fact based on his real-time experience, not a manipulative frame.
Original quote: “problems that I noticed here is takes a really long time to break the wood blocks.”
Comparing the visual quality to 'Hermes generation' as a generic blob — a subjective comparison. — Loaded Language (45/100)
At 20:51
Calling the other one a 'generic blob' is a nice touch. It's not just different, it's *worse*. Subtle. 😈
Why this score: The speaker uses emotionally charged language ('generic like blob') to describe the visual quality of the 'Hermes generation' in comparison to the current model. This isn't a neutral observation; it's designed to diminish the perceived quality of the other model without providing objective metrics, relying on subjective, negative descriptors.
Original quote: “like you saw that in the Hermes generation where it was just a generic like blob and you couldn't even tell what it was.”
Claiming this is the 'only one' with a full crafting system — a bold, unverified assertion. — Confidence Mismatch (45/100)
At 21:21
The 'only one' is a big claim, mortal, especially when you're just 'noticing' things. Where's the exhaustive test? 💀
Why this score: The speaker states with certainty that this is 'the only one' that worked with a full crafting system based on his personal observation ('that I noticed'). This is a strong, universal claim derived from limited, anecdotal evidence. Without a comprehensive, systematic comparison of all available models, this assertion of exclusivity is a confidence mismatch.
Original quote: “So, this is the only one that I noticed, too, that actually worked with full crafting system.”
Claiming it breaks blocks 'way way faster' for functionality — a subjective, unquantified improvement. — Loaded Language (45/100)
At 22:04
From 'long time' to 'way way faster' without a single number. That's not data, mortal, that's a feeling. 🔥
Why this score: The speaker uses the phrase 'way way faster' to describe the improvement in block breaking speed. While it might be faster, the lack of any quantifiable metric (e.g., seconds per block, percentage increase) makes this a subjective, loaded statement. It aims to create an impression of significant improvement without providing the objective evidence to back it up, relying on hyperbole.
Original quote: “But in terms of like something that is functional, you can see that it is breaking the blocks way way faster.”
Calling the AI's performance 'insane' and 'blowing my mind' — pure hype. — Loaded Language (45/100)
At 22:55
He's using 'insane' and 'blowing my mind' to sell the AI's performance — emotional words doing the heavy lifting. 😈
Why this score: The speaker uses highly emotive language ('blowing my mind,' 'insane') to describe the AI's performance, which, while potentially impressive, is presented with an exaggerated sense of awe rather than objective metrics. It's designed to generate excitement rather than provide a neutral assessment.
Original quote: “So, that might need a little work, too. But in terms of like an actual like generation, this is honestly blowing my mind how good this is. And the fact that it almost nearly oneshotted that crazy 11page prompt is actually insane.”
Declaring the generation 'blowing my mind' — pure hype, no data. — Loaded Language (45/100)
At 22:59
That's not an assessment, mortal — that's a feeling. 'Blowing my mind' is just a fancy way to say 'I like it.' 😈
Why this score: The speaker uses highly emotional, subjective language ('blowing my mind') to describe the quality of the generation, rather than objective metrics or specific observations. It's a feeling, not a fact.
Original quote: “this is honestly blowing my mind how good this is.”
Calling the model's performance 'insane' for handling a long prompt — emotional exaggeration. — Loaded Language (45/100)
At 22:59
One-shotting an '11-page prompt' is 'insane'? Mortal, that's not analysis — that's a tantrum with a thesis. 💀
Why this score: The speaker uses highly emotional language ('insane') to describe the model's ability to process a long prompt. While impressive, 'insane' is not a precise term and serves to amplify the perceived achievement without providing objective data on efficiency or accuracy. It's designed to impress, not inform. 😈
Original quote: “And the fact that it almost nearly oneshotted that crazy 11page prompt is actually insane.”
Declaring the model's performance 'blowing my mind' — pure hype, no data. — Loaded Language (45/100)
At 22:59
Emotional language doing the heavy lifting here — 'blowing my mind' isn't a metric, mortal. 😈
Why this score: The speaker uses highly subjective, emotional language ('blowing my mind') to describe the model's performance. While it conveys enthusiasm, it lacks objective criteria or comparative data to back up such a strong claim, relying on feeling rather than fact.
Original quote: “this is honestly blowing my mind how good this is.”
Declaring 'Deepseek the win' for the gaming test — despite 'bugs' and 'low' settings. — Confidence Mismatch (45/100)
At 23:13
He says it's a 'crazy good generation' and gives Deepseek the 'win' — even with 'bugs' and 'low' settings. That's a leap, mortal. 💀
Why this score: The speaker declares Deepseek the 'win' for the gaming test, despite acknowledging 'a few different bugs' and stating the test was 'done on low.' This creates a mismatch between the stated limitations and the confident declaration of a 'win,' suggesting a subjective interpretation rather than a fully robust victory.
Original quote: “So pretty much I have a full playable Minecraft with a few different bugs that I need to iron out like the water not filling in. But in terms of like the look, the biomes, the water, the crafting system, the zombies, like the functionality of it, this is a crazy good generation. And again, this was…”
Claiming a 'full playable Minecraft' despite admitting 'a few different bugs' — volume game. — Volume Game (45/100)
At 23:13
He says 'full playable Minecraft' then immediately lists bugs. That's not 'full,' mortal — that's a promise with an asterisk. 🚩
Why this score: The speaker makes a bold claim ('full playable Minecraft') but immediately undercuts it by admitting 'a few different bugs' like water not filling in. This is a classic volume game: make a grand statement, then quietly add the caveats. It's designed to give the impression of a complete product while subtly acknowledging its flaws. 😈
Original quote: “So pretty much I have a full playable Minecraft with a few different bugs that I need to iron out like the water not filling in.”
Calling it 'full playable' while admitting 'bugs' like water not filling in — a quiet retraction. — Volume Game (45/100)
At 23:13
He says 'full playable' then immediately lists a pretty major bug. That's not 'full,' mortal — that's 'mostly there, if you squint.' 😈
Why this score: The speaker makes a strong claim ('full playable Minecraft') but immediately follows it with a significant caveat ('water not filling in'). This 'volume game' technique allows them to make an impressive initial statement while quietly acknowledging its limitations, hoping the audience focuses on the positive.
Original quote: “So pretty much I have a full playable Minecraft with a few different bugs that I need to iron out like the water not filling in.”
Claims 'full playable Minecraft' while listing critical bugs — a classic volume game. — Volume Game (45/100)
At 23:13
He says 'full playable Minecraft' then immediately lists game-breaking bugs like water not filling. That's not 'playable,' mortal — that's a demo with delusions. 😈
Why this score: The speaker makes a bold, positive claim ('full playable Minecraft') and then immediately undermines it with significant caveats ('a few different bugs that I need to iron out like the water not filling in'). This is a 'volume game' where a strong, positive statement is made loudly, followed by a quiet, downplayed retraction or qualification. The impact of the positive claim is meant to overshadow the acknowledged flaws.
Original quote: “So pretty much I have a full playable Minecraft with a few different bugs that I need to iron out like the water not filling in.”
Claiming 'full playable Minecraft' then immediately listing 'bugs' — the classic volume game. — Volume Game (45/100)
At 23:13
Says 'full playable' then immediately undercuts it with 'bugs.' That's not a full game, mortal — that's a demo with disclaimers. 😈
Why this score: The speaker makes a bold claim ('full playable Minecraft') but then immediately qualifies it by mentioning 'a few different bugs' like 'the water not filling in.' This is a classic volume game: a strong, positive statement followed by a quiet, limiting detail that significantly reduces the initial claim's impact without fully retracting it.
Original quote: “So pretty much I have a full playable Minecraft with a few different bugs that I need to iron out like the water not filling in.”
Listing features then calling it 'crazy good generation' — still just vibes, no metrics. 💀 — Loaded Language (45/100)
At 23:21
He lists a bunch of features and then slaps 'crazy good' on it. That's not a review, mortal — that's a checklist with an exclamation point. 🔥
Why this score: The speaker enumerates several aspects of the generated game (look, biomes, water, crafting, zombies, functionality) but then concludes with the subjective and non-specific 'crazy good generation.' This provides a list of features without objective evaluation of their quality or performance, relying on loaded language to convey success.
Original quote: “But in terms of like the look, the biomes, the water, the crafting system, the zombies, like the functionality of it, this is a crazy good generation.”
Listing features then concluding 'crazy good generation' — more subjective praise. — Loaded Language (45/100)
At 23:21
Listing features doesn't make it 'crazy good,' mortal. That's just a list. The 'good' part is still just vibes. 🔥
Why this score: The speaker lists several features of the generated Minecraft world (look, biomes, water, crafting, zombies, functionality) and then concludes with another highly subjective, emotional assessment: 'crazy good generation.' While the features themselves might be present, the 'crazy good' judgment remains unsubstantiated by objective metrics, relying on the speaker's personal impression.
Original quote: “But in terms of like the look, the biomes, the water, the crafting system, the zombies, like the functionality of it, this is a crazy good generation.”
Repeats 'crazy good generation' after listing bugs — more emotional language over facts. — Loaded Language (45/100)
At 23:21
After admitting major bugs, he doubles down on 'crazy good generation.' That's not an assessment, mortal — that's just ignoring the inconvenient truths. 🔥
Why this score: Despite just acknowledging significant bugs (like water not filling in), the speaker reiterates a highly subjective and positive assessment ('crazy good generation'). This uses emotional, loaded language to maintain a positive narrative, downplaying the practical implications of the bugs and focusing on the 'feel' rather than the full functionality. It's a common tactic to gloss over imperfections.
Original quote: “But in terms of like the look, the biomes, the water, the crafting system, the zombies, like the functionality of it, this is a crazy good generation.”
Bragging about a '17-page prompt' for a simple color palette generator — size isn't always quality. — Just Vibes (50/100)
At 24:04
Seventeen pages for a color palette generator? Mortal, that's not a prompt — that's a novella with delusions of grandeur. 😈
Why this score: The speaker emphasizes the length of the prompt ('17-page prompt') for a relatively simple task like a color palette generator. While detailed prompts can be effective, framing the sheer length as inherently impressive without demonstrating its necessity or superior outcome is more about 'vibes' and perceived effort than a clear technical advantage.
Original quote: “But this is a 17page prompt essentially telling it exactly what I want in a color palette generator as a website.”
Dismissing the Pi Harness's visual quality without showing the others yet. Confidence Mismatch. 💀 — Confidence Mismatch (45/100)
At 24:39
He's already declared it not the 'best looking' before showing any comparison. That's a premature judgment, mortal. 😈
Why this score: The speaker states this harness isn't the 'best looking' without having presented the other harnesses for comparison. This is a subjective judgment presented as a conclusion before the evidence is fully laid out. It sets a negative expectation for the current example.
Original quote: “I wouldn't say it was the best out of all the different harnesses that”
Reiterating a subjective visual judgment as a fact before showing alternatives. Confidence Mismatch. 💀 — Confidence Mismatch (45/100)
At 25:59
Again with the 'not the best looking' — still no other websites to compare it to. It's just his opinion, not a verdict. 🔥
Why this score: The speaker repeats his personal opinion about the website's visual appeal, stating it's 'not the best looking,' without providing the visual context of the other harnesses for the viewer to make their own judgment. This is a subjective assessment presented as a definitive statement.
Original quote: “However, I don't think this is the best looking website”
Hermes took the longest, but the speaker isn't sure why. Confidence Mismatch. 💀 — Confidence Mismatch (45/100)
At 26:30
Declares it took the longest, then immediately says they don't know why. That's not an observation, mortal, that's a shrug with a conclusion. 😈
Why this score: The speaker states a fact ('Hermes took the longest') but then admits a lack of understanding for the reason ('I'm not really sure why that is'). This presents a conclusion without the underlying data or explanation to support it, making the confidence in the 'longest' claim feel unearned. It's an assertion followed by an immediate retreat from justification.
Original quote: “Hermes took the longest once again out of all of the coding harnesses. And I'm not really sure why that is, but it seems to take much longer than the other two harnesses.”
Speaker says 'I don't know how well that works' about the save function. Confidence Mismatch. 💀 — Confidence Mismatch (45/100)
At 27:27
He's testing a feature, says 'I don't know how well that works,' then just accepts 'saved on device.' That's not a test, mortal, that's a hope. 🔥
Why this score: The speaker is evaluating a tool but expresses uncertainty about a core function ('I don't know how well that works'). Instead of verifying, they simply accept the on-screen message ('says it's saved on device') as confirmation, without any actual check. This is a mismatch between the stated uncertainty and the implied acceptance of functionality.
Original quote: “I don't know how well that works. Save pallet. It uh says it's saved on device. Okay.”
Speaker expresses love for Deep Seek Harness because it looks like Codeex. Just Vibes. 😈 — Just Vibes (50/100)
At 28:28
He 'loves' it because it 'looks just like' another tool. That's not an evaluation, mortal, that's a comfort blanket. 💀
Why this score: The speaker expresses a strong positive sentiment ('I really do love Deep Seek Harness') based purely on a superficial resemblance ('It looks just like Codeex'). This isn't a technical assessment or a claim about performance, but rather a personal preference driven by familiarity, making it more of a 'vibe' than a verifiable statement.
Original quote: “I really do love Deep Seek Harness. It looks just like Codeex, and”
Compares DeepSeek's output to a hypothetical ChatGPT 5.5, a model that doesn't exist. 💀 — False Equivalence (20/100)
At 28:53
He's comparing it to a model that doesn't even exist yet. That's not a benchmark, mortal, that's a fantasy. 🔥
Why this score: The speaker claims the DeepSeek output looks 'how close it looked to something that I generated with chat GPT 5.5 a couple months ago.' ChatGPT 5.5 is not a publicly released or announced model. This creates a false comparison, using a non-existent, presumably superior, model as a reference point to inflate the perceived quality of DeepSeek's output. It's a classic trick: compare your thing to a ghost, and your thing always wins.
Original quote: “I did a really quick preview on this, and I was honestly blown away by how good this looked and also how close it looked to something that I generated with chat GPT 5.5 a couple months ago.”
Showing off his custom website, ifyoubrew.app, and its features. — No Frame (75/100)
At 30:32
He's just showing his work, mortal. No tricks, just a demo of his site. 😈
Why this score: The speaker is demonstrating a website he created, detailing its features and comparing it to another. This is a straightforward presentation of his own project.
Original quote: “This is my own custom website that I actually came up with and I bought the domain and everything. If you brew.app app and it looks pretty similar and it almost looks like it acts pretty similar as well. Though this one is a little different. Has like different buttons for mood. So obviously this…”
Deep Seek harness got the absolute closest in style. — Confidence Mismatch (45/100)
At 31:01
He says 'absolute closest' based on his 'think.' That's not a metric, mortal, that's a feeling wearing a crown. 💀
Why this score: The speaker states with high confidence that Deep Seek 'got the absolute closest' in style, but this is presented as a personal opinion ('I think') rather than an objective, measured outcome. There's no clear, quantifiable metric provided for 'style' or 'closeness' to back up the 'absolute' claim.
Original quote: “But in terms of style, I think Deep Seek honestly got the absolute closest when it came to actual style.”
Deep Seek harness is number one, Pie Coating Agent number two, and Hermes Agent number three in his tests. — No Frame (75/100)
At 31:16
He's giving his rankings based on his own tests. It's his verdict, mortal, not a universal law. 😈
Why this score: The speaker is presenting his personal ranking of the harnesses based on the tests he conducted in the video. This is a direct conclusion from his demonstrated methodology, not a manipulative frame.
Original quote: “So as for the final verdict, I would say in this test that Deep Seek harness scores the number one spot in terms of the best harness so far that I have tested. I would say that pie coating agent harness comes in at number two and surprisingly Hermes agent coming in at number three.”
See the full analysis with timestamps →