Xiao Ma
@infoxiao
Followers
12K
Following
15K
Media
340
Statuses
4K
gemini post-training @googledeepmind. views are mine.
New York
Joined August 2014
🎨 Gemini 2.5 tech report just dropped! So proud to have led the development of RL*F (Reinforcement Learning from Human and Critic Feedback) - our breakthrough in AI training inspired by... art school crits? Here's the thing: How do you teach taste? Style? Things without clear
13
32
356
We're happy to support the Human Centered LLMs course, on topics close to our hearts. We'd like to support more classes with free credits for students to use on assignments and projects. If you're an instructor interested in using Tinker in your course, please reach out to
11
45
519
Concerned about biases and politics in your children's education? Be an advocate with these 6 simple steps. Follow THINC to learn more about what you can do as a parent.
0
1
5
Fine-tuning APIs are becoming more powerful and widespread, but they're harder to safeguard against misuse than fixed-weight sampling APIs. Excited to share a new paper: Detecting Adversarial Fine-tuning with Auditing Agents ( https://t.co/NqMeGSCQIF). Auditing agents search
arxiv.org
Large Language Model (LLM) providers expose fine-tuning APIs that let end users fine-tune their frontier LLMs. Unfortunately, it has been shown that an adversary with fine-tuning access to an LLM...
10
45
449
it is underrated how funny @karpathy is
13
21
936
We're honored to be the Exclusive Prediction Market Partner for the fastest growing sports league in the world. Polymarket is proud to be partnering with the PPA & MLP to bring pickleball polymarkets to the masses.
55
20
196
that was way too fast
1
0
4
the way i thought this is the new @AnthropicAI ad
0
0
11
so hm how many aura points do i lose if I implement a gmail 'unsubscribe' mcp by just marking it as spam
3
0
6
Even one of the top imaging companies in the US admits how much better ultrasound is than mammography for breast imaging. What is not said is the radiation risk of mammograms is worse than the benign sound waves of U.S. I never Rx'd mammograms, choosing ultrasounds instead.
3
14
73
The ultimate test of code maintainability? Let an agent run free and see if it descends into unmanageable chaos.
0
1
4
galaxy brain is realizing you can use cli to access reminders https://t.co/WXFmtEnyJe and give all the coding clis access to it 🤯
github.com
A simple CLI for interacting with macOS reminders. Contribute to keith/reminders-cli development by creating an account on GitHub.
0
0
9
no one: absolutely no one: literally no one on the planet earth: claude:
171
44
1K
Mark your calendars. After 4 years of building. $PRDT launches November 1st, 2025 - 12PM CET. Let’s make history together. 💚
266
327
1K
very funny that when we watched matrix before: fiction and now: product roadmap
0
0
9
out: context switching in: agent switching more in: skill maxing
The number of coding agent's you can manage at one time, productively, is your Ma number. It is decreed. I'm at 2 - maybe 3 :(
1
0
3
Dunbar's number: but for the number of agents you can manage at the same time. can we name it the Ma's number
1
0
9
That strange noise under the hood? You don’t have to ignore it anymore.
0
1
18
Let me pitch you Skills: - They allow anyone to customize agents with a simple primitive — files - They give us continuous learning until continuous learning arrives - They are powerful, composable, and sufficiently AGI-pilled - you can now say “probably a skill issue” at work
Claude can now use Skills. Skills are packaged instructions that teach Claude your way of working.
49
70
1K
😂 not me using the exact clip at work
At a high level, the best analogy I've heard for Skills is something like Neo learning Kung Fu in seconds in the Matrix. We're "loading in" specialized knowledge to our general agents at runtime.
0
0
9
Introducing NotebookLM for arXiv papers 🚀 Transform dense AI research into an engaging conversation With context across thousands of related papers, it captures motivations, draws connections to SOTA, and explains key insights like a professor who's read the entire field
57
529
3K