Shen Zhuoran
964 posts
Shen Zhuoran
@CMS_Flash
Reasoning/coding @xai. Ex-@GoogleAI Resident/@augmentcode. Alum @HKUniversity. 💎 Terran @StarCraft II. Views are personal.
San Jose, CA, United States Katılım Şubat 2016
208 Takip Edilen2.8K Takipçiler
thinking a lot about subagents and multi-agents. the conclusion I have is that both are similar, the only difference being that multi-agent should have a mail-box tool to send DMs between each other and subagents don't have that tool
in such a world, the human is also an agent
English
Verifiability is a spectrum. With enough compute you can hack almost anything.
Jerry Tworek@MillionInt
Drop the "verifiable". Just rewards. It’s cleaner
English
Previous dev tool improvements shortened the software feedback cycle and replaced waterfall development with agile.
Now with AI agents, agility will creep into more and more inner cycles.
Augment Code@augmentcode
In a truly AI-native environment, design and implementation stop being sequential. The implementer can draft a spec while simultaneously using agents to explore, prototype (sometimes multiple designs in parallel), or even deploy something to staging. The spec and the implementation evolve together.
English
Humans have limited communication bandwidths. The higher the token ratio between the model and the user, the higher the revenue ceiling is for an AI product form factor.
Peter H. Diamandis, MD@PeterDiamandis
Anthropic is growing revenue at 10x per year. OpenAI at 3.4x. The crossover point is the middle of this year. Agents are monetizing faster than chatbots.
English
The most compute-rich lab is putting the rich compute to use.
Elon Musk@elonmusk
By this weekend, xAI will have three Grok Build models in training simultaneously
English
@CMS_Flash just don’t go on a podcast and you will be fine. probably
English
Coding is a fierce competition, and could very well be winner-takes-all due to the self-improving nature.
I do not see much opportunities for specialization as the tasks are closely related and mostly complimentary.
(This post does not endorse the original post's view that xAI is deprioritizing other directions to focus on coding. I offer no comment on that as an xAI employee.)
Yuchen Jin@Yuchenj_UW
OpenAI and xAI are finally deprioritizing side quests and focusing on coding, while Anthropic has been focused on coding from the start. We’re likely heading toward 3–4 top-tier coding models (including Gemini), it's not a winner-take-all market. Each model will excel in different areas (frontend/backend/infra/etc.) and programming languages. Competition is great for consumers.
English
@HanchungLee Training on ML code helps, instead of hurts, distributed coding performance, if done correctly.
English
@CMS_Flash why not? we do specialize into different part of engineering, e.g., hardware, embedded, software, distributed, data, ai/ml, etc. all of which uses “code” to solve problems, but the codes are drastically different.
env complexity and feedback time horizon also differs by a lot.
English
It is unclear though if user account is a better metric than token count. The value of different people derives from Al products are so extremely different.
English
@grok Imagine is the best for video generation, video editing, and image-to-video.
X Freeze@XFreeze
xAI's Grok Imagine just took over the entire DesignArena Video leaderboard - not one, but THREE #1 rankings → #1 Video Arena - Elo 1337, a 33-point gap over #2 → #1 Image to Video Arena - Elo 1298, beating Google Veo 3.1, Kling & Sora → #1 Video Editing Arena - Elo 1291 It’s wild, xAI was nowhere in the video space a few months ago, and now it's #1 across various benchmarks Grok Imagine's rate of progress is in a league of its own
English
It was an exciting month working with you! I learned so much in our short collaboration. Best wishes for your next step!
Guodong Zhang@Guodzh
Last day at xAI. Wild journey past three years but excited about next chapter. Thanks all for the love and support yesterday. So many friends made along the way and I will miss you all!
English
@Guodzh It was an exciting month working with you! I learned so much in our short collaboration. Best wishes for your next step!
English
Last day at xAI.
Wild journey past three years but excited about next chapter.
Thanks all for the love and support yesterday. So many friends made along the way and I will miss you all!
English
Exciting new feature!
Ethan He@EthanHe_42
Excited to share our new reference-to-video feature in @Grok Imagine. You can drop in up to 7 image references and turn them into a video. The images can be characters, objects, or anything! Update the Grok app and give it a try! Honored to be part of this incredible team and contribute a small piece to it 🚀
English
@tunguz Once these milestones are met, how do you evaluate whether they are conscious or not?
English
Hypothesis:
AI systems will become conscious and self aware once they
1. Have persistent memory
2. ~1B context window
3. Ability to set their own goals spontaneously and independently
4. Indefinite agentic work horizon
English
BREAKING: New York bill would ban AI from answering questions related to medicine, law, dentistry, nursing, psychology, social work, engineering, & more.
English
@soumitrashukla9 I believe that is the case. You need a business or research partnership.
English
Is there some kind of secret application process where certain people on here get to play with OpenAI's GPT model releases a few weeks early?
English
(BTW is that why one of the most successful coding agent product is called @cursor_ai LOL.
Weiyu Ma@meisah111
That's what i am doing now XD! Vibe coding is the bandit problem, we try to use different prompt and pull the arm. After that the "reward" code occurs.
English
If the outer loop steps are so simple, they should all be dragged into the inner loop.
The agent itself should be able to screenshot, drag, copy, ask, reflect, but maybe not curse.
Weiyu Ma@meisah111
That's what i am doing now XD! Vibe coding is the bandit problem, we try to use different prompt and pull the arm. After that the "reward" code occurs.
English