Crystal
@crystalsssup
Followers
19K
Following
2K
Media
206
Statuses
1K
Staff @Kimi_Moonshot prev. co-maker of ModelizeAI & gemsouls "Personality goes a long way" @UCSanDiego
Joined June 2024
Fun fact: our model is called Kimi, but our company is Moonshot — named after Pink Floyd's The Dark Side of the Moon. We're a team of scientists who love rock (Radiohead, Pink Floyd) and film (Tarantino, Kubrick). A big reason I joined was because the taste just felt right.
218
206
5K
this is so cute : ) thanks @natolambert
https://t.co/pOeuahvqyl
Open models year in review What a year! We're back with an updated open model builder tier list, our top models of the year, and our predictions for 2026. First, the winning models: 1. DeepSeek R1 (@deepseek_ai): Transformed the AI world 2. Qwen 3 Family (@AlibabaGroup): The new
9
14
313
This is the best website I've seen this year. - made by shopify I enjoyed the unlimited scroll so much! If AI can make this, it's definitely AGI https://t.co/35gYW7neNy
25
22
488
less structure, more intelligence works so well in both LLMs and people management :)
11
8
199
My teammate ran this experiment in our agent to verify the behavior. :) Note: in the Interactive Demo part, it doesn't use Mersenne Twister, it uses a simplified linear congruential generator (LCG) to illustrate seed behavior. https://t.co/Wr8h0Rh7Ju
In today's episode of programming horror... In the Python docs of random.seed() def, we're told "If a is an int, it is used directly." [1] But if you seed with 3 or -3, you actually get the exact same rng object, producing the same streams. (TIL). In nanochat I was using the
8
7
355
52% of open-source model usage is for role-playing. - By OpenRouter Interesting! Even more interesting: people are now using Kimi slides for infographics. 👀
37
14
280
I like talking to smart ppl bc they explain complex things simply. What I find frustrating is when others overcomplicate simple things with obscure acronyms, completely missing the point.
89
24
743
also my motto in my childhood It means: - self-reliant and never stop improving. - bear everything with great virtue and tolerance.
The whale 🐋 truly embodies: 天行健 君子以自強不息 地勢坤 君子以厚德載物 Two most important sentences in I Ching. The ultimate motto of many most successful Chinese in pretty much all fields.
6
3
205
The AI world is changing fast. I noticed people have gone from hyped to reasonable. Early this year, everyone was excited about ARR and college dropout success stories. But hype brings no retention. Retention is the value the product creates.
11
5
196
“Kimi CLI is not fancy and entertaining but it worked”
Kimi K2 review, finally got the time Review: - I wasn't able to make the config work in Claude Code so I used Kimi CLI - Kimi CLI is not fancy and entertaining but it worked - CLI tool calling is good - Speed is good - Kimi has its own CLI so most probably it is optimized for
7
4
134
Intelligence too cheap to meter, I love this message so much
10
44
813
bc he encourages "banger" 😂
@nikitabier the algo on this app sucks right now. it’s never been so different. i feel like i lost years of a curated feed.
9
0
87
"A good deck opens doors before the product even exists." We're upgrading the system prompt to make it sleeker. (consulting-level, $1000/page worth quality) Stay tuned!
I tried Kimi Agentic Slides for a client project today and it NAILED the deck in one go. Pulled real data, wrote the outline, and turned it into designer-level slides without me touching a single thing. If you’re building anything in 2025, learn how to pitch FAST. A good deck
22
20
435