Scott Hanselman
Region: US
Monday 02 February 2026 00:05:44 GMT
76709
3874
157
629
Music
Download
Comments
Savage Meerkat :
I’ve been saying this from the beginning, the idea that AGI will come from large language models is the equivalent of training horses to run faster and faster with the exception that one day, magically, one of them will give birth to a locomotive.
2026-02-02 03:17:25
215
Barry Carey :
The security risk is crazy. Lots of less technical people jumping on the hype and giving it the keys to the kingdom. I'm a very technical person and the first time I saw Claude using Dev tools MCP I was in awe. I get it feels like magic. But I understand the risks of it's access. The average person doesn't stand a chance. It won't be long before the first story of someone getting accounts drained from a prompt injection
2026-02-02 01:22:57
59
Men4Matriarchy :
Great description of LLMs. The history of Markov chains provides an enlightening look into the start of LLMs. It's also fascinating to dive deep into what exactly is the difference between an LLM and its component parts (memory via DB, input organs, output organs). Is the manner in which humans process information any different than LLMs and their component organs as a whole? Humans without our component organs (ears, speech, eyes, fingers to type) would not be recognized as intelligent. I'm not claiming AI is sentient, but diving deep into a comparison raises unsettling questions about the nature of human intelligence. Are humans not also context completion engines? Our neurology, with a lifetime of input and training, seems similar to pre processing layers and weights.
2026-02-05 15:21:57
3
Max Steel :
I love how everyone is singing.Its praises, but no one is actually giving examples of real world use. having It run a YouTube video on my computer While i'm away isn't helpful
2026-02-02 02:07:57
34
Eternity :
I worry about token usage from an enthusiastic bot 😅
2026-02-02 07:35:48
2
azsportsfan9 :
As someone who has spent probably 300+ hours and $300+ toying with his Clawd bot in concept, it is an amazing idea, but I can’t seem to get my bot to be stable with any architecture structure. Agent routing, orchestrator. It is a great tool to be able to switch LLMs while building. Maybe in the future, with further versions, it’ll be what is promised, but it isn’t currently, or maybe it’s just a skill issue on my end. My 2 cents.
2026-02-12 03:08:57
4
cliffrosen185 :
I wish people would stop saying ‘statistically most likely next word’ because it’s downright misleading - it makes it sound like shallow word frequencies. Predicting language well requires understanding it. At scale, statistics capture meaning, not just which words tend to follow which.
2026-02-02 04:30:15
9
1izbon :
You’re an incredibly talented communicator!
2026-02-11 04:38:12
5
craftygurl :
That was a yap. Fascinating yap
2026-02-10 03:16:31
8
vbprgmr :
did you see that movie?
2026-02-04 02:04:20
0
Spartan B-312 :
Forget AGIs, Chipotle taking reservations is the true dystopian future
2026-02-05 02:56:50
7
Niki & Ritika from Sani :
This was so helpful and easy to understand, thank you
2026-02-14 23:04:53
4
Kathryn :
I see even small amounts of hallucination in routine requests. Right now I need to specify NOT using information that has not been explicitly provided, but that should be the default for those requests
2026-02-04 20:31:14
6
Piratesmama :
The problem with all of this is that the worst of humanity will use this to hurt people. Despite all the good, this will be have devastating consequences.
2026-02-04 04:15:37
11
Lorentz Factr :
Wu Tang Financial mentioned 😂
2026-02-02 03:17:01
30
Eric Williams :
how many keys have been leaked?
2026-02-02 06:48:37
7
Hobbes :
You need to diversify your bonds...
2026-02-02 03:26:23
25
gutterwalker :
The Meseeks episode of Rick and Morty is essentially an allegory for AI with agents personafied. The gap between what we ask for and what we actually want and the ambiguity that clutters even the smallest asks.
2026-02-03 18:00:06
8
Bryan Valencia :
This was actually so incredibly useful/insightful for understanding how these things work a bit more. I’ve been wanting to pick up a mini to dedicate to playing around with this stuff more.
2026-03-05 20:11:01
1
FiddyMint :
Hands-down, the most valuable piece of AI coaching and description I’ve heard in many months thank you so much
2026-02-03 15:30:26
4
interdurozw :
Then think about the upcoming legal implications. Who’s liable?
2026-02-04 15:21:47
2
Caféconcuriosity :
Great explanation
2026-02-03 21:09:26
2
Engelsb809 :
I wish things life were explained this way. Good video
2026-02-11 18:06:01
1
Mcrnlip :
This came at the exact right time when I was about to ask ChatGPT about all this, but I’d rather get it from you. Lol.
2026-02-04 03:04:50
2
GasCanCan :
Awesome explanation thank you
2026-02-23 21:53:22
1
To see more videos from user @shanselman, please go to the Tikwm
homepage.
