GigaQian
Region: US
Wednesday 22 April 2026 00:26:18 GMT
77878
6921
193
771
Music
Download
Comments
g_lifts :
Kimi is so BAD.
2026-04-22 11:51:00
11
Wahala :
The benchmarks on Kimi 2.6 are crazy so far but mannnn this glm 5.1 is liquid gold. I’m not even exaggerating, I’m looking at it unit test each module by itself as we speak. While decomposing a previous code. I’m honestly speechless
2026-04-22 00:56:19
97
creativesecurity :
As soon it released I gave to my Hermes and it went super power
2026-04-22 01:17:04
8
ً :
This is an ad
2026-04-22 06:29:17
41
Jeremy S :
I set up kronos
2026-04-22 01:20:29
16
Never Settle :
I like you man, cutting through the BS with just raw uncut unhype content w facts
2026-04-23 00:46:44
7
AugieC :
Yeah, I have officially been left behind
2026-04-22 02:02:11
5
veryveryodd :
What's the dif between swarm from kimi and multiple sub agent from the harness itself
2026-04-22 01:47:22
7
JustinLeon97 :
Bro I love how excitement just pours from you when you’re teaching us
2026-04-22 04:02:44
6
Eman-dingo :
It's the end of the world as we know it. But I feel fine.
2026-04-22 01:28:04
5
Genie :
100 agents running on what, carl? compute isn't free
2026-04-22 04:11:38
6
b0gie :
Kimi 3 is gunna be WILD.
2026-04-22 07:16:14
6
Diana Teresh :
How does it compare to Claude?
2026-04-22 09:47:05
8
Just_Me :
Commenting for the algo.
2026-04-22 05:55:06
5
Chief :
Is Kimi 2.6 with Qwen 3.6 a super power?
2026-04-22 11:28:50
6
ForwardCompatible :
kimi k2. 5 rocks. 2.6 is simply amazing
2026-04-22 01:00:57
7
THE ONE AND ONLY :
Great content
2026-04-22 00:46:13
5
⛪️Nestor Diaz🖼🙋♂️ :
kimi is pretty strong in reasoning
2026-04-22 21:22:33
7
UU :
alot of these models are just benchmarkmaxxing, they train the models to specifically do well in select benchmark, tested kimi yesterday by having it create a AI detector tool and used gemini 3.1 flash lite as the model for the detector, it one shot the entire task perfectly although I spent a good 30 minutes refining my prompt with chatgpt
2026-04-22 10:52:16
7
Jonn Nova :
Everyone seem kimi 2.6
2026-04-22 10:01:13
7
Zach :
Great content! I’m definitely gonna check it out, I’ve been using Claude a lot more lately, but the usage has been screwing me
2026-04-22 01:05:15
10
❂TacticalTac𖤓🪽 :
I think they removed usage limits too. I used to have only so many agent credits but now it doesn't show agent credits diminishing anymore. I'm having kimi research llm integrated TDD right now because this is an area where it has been very difficult to remove the human element but could massively improve agent performance if it is solved. I'm sure there has got to be research out there solving this soon.
2026-04-22 01:49:41
5
cheec :
Comprehension is downstream of intelligence; and intelligence that leaves no upstream is intelligence you can never comprehend, or trust. The blackbox is a product feature; it’s a token burning machine.
2026-06-18 22:09:23
0
Johnohhh1 :
yoooo Goose with K2.6. puts out on the first date
2026-04-22 11:52:43
4
See_no_hear_no_speak_no :
I like the closing question and have been thinking about it too. We can do cool stuff now for sure. Make money. Maybe. But how can we use tech to make the world a better place? How can we help the people - who have no idea about swarm AI’s - lives better?
2026-04-22 07:17:03
4
To see more videos from user @gigaqian, please go to the Tikwm
homepage.
