I Built JARVIS with Claude Code (INSANE Results!)
Credibility score: 50/100 — Mixed Credibility. Several questionable claims detected. Watch with healthy skepticism.
BSmeter analyzed "I Built JARVIS with Claude Code (INSANE Results!)" and rated it 50/100 for credibility (a BS score of 50/100 — mixed credibility), on 2026-09-13. Its weakest claim — "Claims to have built "the best" AI assistant on the internet" — scored 45/100 and was flagged as confidence mismatch. 14 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Of 14 claims analyzed: 0 scored under 40, 12 between 40 and 69, and 2 at 70 or above.
Claims analyzed
Video opens with a highlight reel preview of JARVIS's capabilities. — Just Vibes (50/100)
At 0:00
They're showing off the AI's 'insane results' right out of the gate — a classic demo reel to hook you. 😈
Why this score: This is a quick montage of the AI performing tasks like scheduling, invoicing, and checking emails. It's meant to impress and set the stage for the video's claims about building JARVIS, but it's a demonstration, not a claim to be fact-checked. It's pure showmanship, mortal. And it works, doesn't it?
Original quote: “Jarvis, do I have any meetings tomorrow? >> Three on the books tomorrow, sir. A busy block at 10:00 a.m., lunch with Alex at 1:00 p.m., and the gym at 4:00 p.m. A day that's apparently too important to name half its own events. >> Jarvis, create a new invoice for the new lead, Mike Johnson, for…”
Claims to have built "the best" AI assistant on the internet — Confidence Mismatch (45/100)
At 0:39
Zero benchmark, zero comparison, just the crown on his head. Confidence with no receipts. 💀
Why this score: Calling something the absolute best on the entire internet requires at least some form of comparison or external validation. He provides none. This is pure self-declaration with no evidence backing the superlative.
Original quote: “I built the best personal AI assistant on the internet. This is Jarvis. Not only is Jarvis capable of doing everything a normal personal assistant would, on top of that, he's incredibly funny.”
Offers a completely free guide and prompt pack — Plain Sales Pitch (45/100)
At 0:54
Free up front, then the skool.com link shows up. Classic funnel language. 😈
Why this score: The promise of completely free materials is immediately followed by directing viewers to skool.com/aiworkshop. This is the standard lead-magnet-to-paid-community pattern, not a genuine no-strings-attached gift.
Original quote: “I'm going to give you this entire guide and prompt pack, which is step-by-step guide, completely for free”
Claims full model switching via Open Router to any AI model — Missing Context (45/100)
At 1:04
Open Router gives access to models — doesn't mean Jarvis can actually use them seamlessly. Huge gap between access and integration. 🚩
Why this score: Having API access to multiple models through Open Router is not the same as having a fully integrated system where the assistant can fluidly switch between models mid-conversation while maintaining context, personality, and tool access. The technical implementation details are missing.
Original quote: “Jarvis uses Anthropic's APIs. As you can see, its brain operates on Opus 5, but I have the ability to have him change his brain to any other model, right? I have given him access to Open Router, which means that he can access any model like Grok 4.5, like Gemini, or whatever it may be.”
Claims Claude-built JARVIS can book restaurants by phone — No Frame (75/100)
At 2:49
Straight demo. He plays the actual call, no hype. Green pill.
Why this score: He shows the phone number, plays the live restaurant booking, and the AI reports back. Nothing hidden, no anonymous experts. The recording is the proof.
Original quote: “Can also have him make phone calls for me. Now, I did a previous video on this where I give a full demo of what that phone call represents. Basically, he has a number. I've given him a phone number, and he can make calls on my behalf, book reservations, call restaurants, check prices, let's say for…”
Claims JARVIS can prep him for meetings via calendar access — Confidence Mismatch (45/100)
At 4:01
Shows calendar lookup, then jumps to 'incredible' prep calls. One works, the other is promised.
Why this score: He demonstrates pulling tomorrow's meetings, but the 'pre-call prep' is only described, never shown. The calendar bit is real; the prep calls are still a future claim.
Original quote: “So that's the phone feature, and of course I have the ability to ask him anything related to my calendar, my email, and all that kind of stuff. So for instance, let's say I have all of these different meetings for tomorrow. I can ask him to actually call me before a meeting with a prep, which is an…”
Laughs at the AI's 'ambiguity and cardio' line, then calls it 'very simple and very quickly'. — Confidence Mismatch (45/100)
At 4:46
He's laughing at the AI's 'joke' then immediately pivots to 'very simple'. That's not simple, mortal, that's a programmed response. 😈
Why this score: The speaker laughs at the AI's generated line, which implies a level of sophistication, but then quickly downplays the complexity by calling the process 'very simple and very quickly'. This creates a mismatch between the perceived intelligence of the AI's output and the speaker's description of its underlying simplicity, potentially overstating the ease of achieving such results.
Original quote: “>> [laughter]
>> There you go. So uh very simple and very quickly he has access. Again, all of this is because I've given them access”
Claims the AI inferred invoice details from a simple prompt, but then explains it pulled from a pre-existing 'leads table'. — Volume Game (45/100)
At 5:58
He says he gave 'no description' then immediately explains the AI pulled all the 'details' from a pre-filled table. That's not inference, mortal, that's a lookup. 💀
Why this score: The speaker initially claims the AI generated a detailed invoice from a 'very simple' Telegram note with 'no description'. However, he immediately clarifies that the AI knew the details because 'the new lead Mike Johnson is in this leads table' which contains all the necessary information like 'AI Reputation build' and 'proposal set'. This is a classic volume game: a bold claim of AI inference followed by a quiet, immediate retraction that reveals the data was pre-populated, not truly inferred from a minimal prompt.
Original quote: “Now, all of this, the details, obviously I didn't give him any description, right? You guys saw that the Telegram note is very simple. All I'm saying is the new lead Mike Johnson. So, the reason why it knows this is because the new lead Mike Johnson is in this leads table. It has the AI Reputation…”
Claims the AI 'knows everything' about his business from a 'Second Brain' — a confidence flex with no proof. 😈 — Confidence Mismatch (45/100)
At 6:36
He says the AI 'knows everything' about his business — a bold claim for a system that just pulls from files. That's not 'knowing,' mortal, that's 'reading.' 💀
Why this score: The speaker attributes human-like 'knowing' to an AI that is merely accessing and processing files. While impressive for automation, calling it 'knowing everything' overstates the AI's cognitive ability and implies a level of understanding beyond simple data retrieval. It's a classic overstatement to hype the tech.
Original quote: “So, therefore, this new one is for $2,500. And again, the description, all of this is coming from my because he has access to my AI Second Brain and all knows everything about my business. So, when I say AI Reputation Manager build, he grabs all of the description from all of the files that are…”
Brags about the AI creating email drafts, then admits he has to review 'everything' — a classic volume game. 😈 — Volume Game (45/100)
At 7:20
He says the AI creates drafts, then immediately admits he has to check 'everything' to avoid 'ridiculous' output. So, it's not a solution, it's a first pass you still have to babysit. 🔥
Why this score: The speaker presents the AI's ability to create email drafts as a significant feature, but then immediately undercuts its efficiency by stating he 'has to review everything' to prevent 'ridiculous' content. This is a volume game: a big claim followed by a quiet admission that reveals the limitations, making the 'feature' less impactful than initially suggested.
Original quote: “And then on top of that, he has he's capable of creating drafts. So, if I go to my email here, so as you can see, earlier, as you guys saw, this was one of the emails that I received. So, all I told him is to go ahead and respond to him. So, he creates a really polite email. Of course, he doesn't…”
AI 'Jarvis' describes the speaker's YouTube channel — a demo that's more impressive than the previous claims. 😈 — No Frame (75/100)
At 7:45
The AI actually describes the screen content accurately. A genuine demonstration, for once. I'm almost impressed. Almost. 😈
Why this score: The AI's ability to accurately identify and describe the contents of the screen, including subscriber count and channel name, is a direct and verifiable demonstration of its capabilities. Unlike previous claims that relied on vague assertions, this segment provides concrete evidence of the AI's functionality in real-time. It's a straightforward demonstration without obvious framing tricks.
Original quote: “So, obviously these are just a few features. He has so many more features that if I made this video for all the features, it will take a long time. But, one of the most fun features is that you have the ability to share your screen and Jarvis will be able to see exactly what you're saying. So, for…”
Claims Jarvis can perform browser tasks — shows no actual demo. Confidence Mismatch. 💀 — Confidence Mismatch (45/100)
At 8:33
Says Jarvis can 'perform tasks on your own browser' — but shows nothing. Just tells you it can. 😈
Why this score: He's making a big claim about functionality without demonstrating it. He just says it's 'really cool' and expects you to believe it. Where's the proof, mortal? I've seen more evidence in a politician's promise. 🔥
Original quote: “>> good, Jarvis. Thanks. And then on top of that, you can also have them do things. So, if you're sharing your screen, let's say you want to fill a form or ask him a particular question, he's also capable of actually performing tasks on your own browser, which is really cool feature.”
Offers a 'free community' with guides, then immediately pivots to a 'paid community' for the 'finished' version. Plain Sales Pitch. 🍒 — Plain Sales Pitch (45/100)
At 9:32
The classic bait-and-switch: 'free' gets you the scraps, 'paid' gets you the actual product. I've watched this trick since before money existed. 💰
Why this score: He sets up a 'free community' with 'guides' and 'prompt packs,' but then clarifies that the 'finished' and fully functional version, the one he's been hyping, is locked behind a 'paid community.' It's a clear upsell, presenting the free version as a stepping stone to the real deal. This isn't a gift, it's a funnel. 😈
Original quote: “Now, of course, all of this, the prompt pack, the guides, everything is there inside the free community. But, if you want the finished, which is this version, right? With all the voice and all of the resources and all of the functionalities, that's inside the paid community.”
Claims you can download and run Jarvis 'in 5 minutes' — a bold promise for complex AI. Confidence Mismatch. 🚩 — Confidence Mismatch (45/100)
At 9:56
Download and run a complex AI in '5 minutes'? That's not a promise, mortal, it's a sales pitch. ⏱️
Why this score: Setting up an AI assistant, especially one with the claimed functionalities, typically involves more than a 5-minute download and run. This sounds like an oversimplification to make the paid product seem effortlessly accessible, ignoring potential setup complexities, dependencies, or user-specific issues. It's a classic move to minimize effort and maximize perceived value. 💀
Original quote: “right? So, this is version six of Jarvis that you can download in 5 minutes and you'll be able to run it on your own file.”
See the full analysis with timestamps →