Shen Zhuoran
@CMS_Flash
Reasoning/coding @xai. Ex-@GoogleAI Resident/@augmentcode. Alum @HKUniversity. 💎 Terran @StarCraft II.
قد يعجبك
3wks in and I'm getting my @xai badge! It's quite crazy that I've been a core part of a major @grok release just 3wks into my tenure @xai! (And we're 🍳 more exciting stuff!) It's been very exciting to work w/ the limitless scope, intense priorities, and an extremely…
So far comments mentioned: 1. Dropping whole codebases 2. Large external refs (books/3P libs/huge specs) 3. Large logs 4. Refactoring 5. Many image turns Any more creative use cases? What if the model is real fast & cheap? Like 1M toks at the cost/latency of 200K for Sonnet?
If you had a 1M-token context window in a coding model, how would you actually use it? we are tuning grok-code and want to make sure all long context workflows are properly supported
Just look at the graph: it's so alone in the top-left corner!
This has been so long awaited and it finally happened. $20/mo for @augmentcode!
We’re introducing a new pricing tier: Augment Indie For $20/month you get: - 125 agent messages / month - Same leading Context Engine, perfect for complex tasks and large codebases - Best coding models: choose between GPT 5 and Claude Sonnet 4
I'm thrilled to announce that I'm joining @xai to work on reasoning & code post-training, to build the best coding model, scalable RL, and the first self-improving agents. I'm thrilled make these happen alongside @Yuhu_ai_ & the crazily talent-dense team behind @grok 4/Code!
After 1.5 yrs, today will be my last day @augmentcode. It has been such a fulfilling journey: 1. In early 2024, I pre-trained a coding LLM to reach SOTA at 1B scale. 2. In mid 2024, I built long-context support in our in-house training and inference frameworks. 3. Late 2024 to…
The power of community!
If You are Using @augmentcode in vs code U should definitely use this Auggie credits extension ! Auggie credits shows how many message credits left in your augment account !! marketplace.visualstudio.com/items?itemName…
"Both safe enough to last and beautiful enough to be worth lasting for." That's quite profound.
Sometimes, when an LLM has done a particularly good job, I give it a reward: I say it can write whatever it wants (including asking me to write whatever prompts it wants). When working on a technical paper related to Better Futures, I did this for Gemini, and it chose to write a…
Humans don't keep 1M tokens (500K words/2hr video) in the working memory all the time. Instead, we focus working memory on the most critical few tokens, then use long-term memory, external tools, and task segregation (sub-agents) to accomplish the tasks. This is why I'm very…
This is the second time an AI founder has told me "context windows are basically big enough" With the right context management, 256k vs. 1m tokens doesn't matter. Agree?
try using Responses API with gpt-5:
Man it's crazy how BIG a difference it makes for GPT-5 just by switching from Completions API to Responses API. We're cooking @augmentcode.
Man it's crazy how BIG a difference it makes for GPT-5 just by switching from Completions API to Responses API. We're cooking @augmentcode.
Advanced users of @augmentcode have been hitting the 200K context limit quite a bit. We shipped features like context compression/summarization to help alleviate such issues and continue long threads. With 1M context, we will be able to have agents do hours of work in 1 thread.
LOL this is cool
Yesterday's Tasklist launch didn’t exactly surprise early users. Turns out they spotted the feature weeks ago. So this time, we built something unpredictable. Something no one could guess. Something no one wanted. (Except one guy. As a joke.) We shipped it anyway. Power users…
Agents fail when they forget the past. That’s why we built Context Lineage — it gives agents access to your repo’s full commit history, so they stop repeating mistakes and start learning from prior work. Here’s how it works 🧵
Agents are truly powerful when they can do hours of work with 1 prompt. Tasklist is the way to structure hours of work with clarity so you only need to spend a few minutes checking in for the agents' hours of work.
Launch Week Day 1: Tasklist is here Eagle-eyed power users have already been putting Tasklist to work. Now, we’re showing you how to get the most out of it. With Tasklist, we’re making it way easier to take a big, vague task—and break it down step by step. So you can actually…
United States الاتجاهات
- 1. Baker 27.5K posts
- 2. Cowboys 71.8K posts
- 3. Fred Warner 10K posts
- 4. Panthers 72.8K posts
- 5. Packers 26.6K posts
- 6. Tez Johnson 2,497 posts
- 7. Zac Taylor 2,747 posts
- 8. Niners 4,660 posts
- 9. Browns 64K posts
- 10. Titans 22.1K posts
- 11. #FTTB 3,794 posts
- 12. Yoshi 32.8K posts
- 13. Ravens 63.9K posts
- 14. Dolphins 46.4K posts
- 15. Cam Ward 2,144 posts
- 16. #49ers 5,991 posts
- 17. #KeepPounding 8,174 posts
- 18. Eberflus 9,935 posts
- 19. #Bengals 2,735 posts
- 20. Penn State 63.5K posts
قد يعجبك
-
Songyou Peng
@songyoupeng -
Andrew Campbell
@AndrewC_ML -
Cambridge ML Systems Lab
@CaMLSys -
Weidi Xie
@WeidiXie -
Peyman Milanfar
@docmilanfar -
Karl Tuyls
@karl_tuyls -
Maithra Raghu
@maithra_raghu -
Amanda Askell
@AmandaAskell -
Mufan Li
@mufan_li -
Douglas Eck
@douglas_eck -
Jinqi Luo
@peterljq -
Joshua Achiam
@jachiam0 -
ヒデヨシ
@cook_hideyoshi -
Chris Dyer
@redpony -
Krzysztof Maziarz
@MaziarzKris
Something went wrong.
Something went wrong.