Brandon Handoko

631 posts

Brandon Handoko banner
Brandon Handoko

Brandon Handoko

@brndnwll

🇮🇩; now @scale_ai, always in faith and fury

SF เข้าร่วม Mart 2019
275 กำลังติดตาม841 ผู้ติดตาม
ทวีตที่ปักหมุด
Brandon Handoko
Brandon Handoko@brndnwll·
Hey @alexisohanian - I'm the Head of Product you're looking for! 1 min vid re: why :) I'm a huge fan of the 776 portfolio + thesis, the past 5yrs I was: PM 1 @ZebraIQ w @tzhongg, Product 3 @Commsor w @TheTeaGuns, PM 2 @TryMetafy w @JoshFabian Would make my wk to chat!
Alexis Ohanian 🗽@alexisohanian

We're looking for a Head of Product at @sevensevensix 🏛️ Our Head of Product will lead the development and evolution of Cerebro, our proprietary software platform that powers all of our firm's operations for our founders, team, and investors. This hire is personal to me...

English
15
2
95
33.7K
Brandon Handoko รีทวีตแล้ว
Scale Labs
Scale Labs@ScaleAILabs·
ICLR 2026 Accepted Paper: MoReBench. Models perform well on math and coding benchmarks, but measuring moral reasoning is still a challenge. To bridge this gap, we introduce MoReBench: a benchmark for Evaluating Procedural and Pluralistic Moral Reasoning in LLMs.
Scale Labs tweet media
English
1
8
34
5.2K
Alex Finch
Alex Finch@alex_w_finch·
Sorry to burst the bubble but this exists … ironically at @scale_AI
Natasha Malpani 👁@natashamalpani

nobody has built the @scale_AI for physical AI. eight years after @CovariantAI proved the data flywheel is the business. eight years. the reason is obvious once you see it: simulation looks like it should work. reduce 500 real demonstrations to 50, close the gap with compute, ship. @nvidia has more sim-to-real resources than anyone. their own documentation says the reality gap is still the core problem. physics doesn’t care about your render quality. force signals, contact dynamics, failure modes: simulation generates plausible versions of all three. a deployed robot finds the difference in the first hour. the companies that got it right stopped trying to solve simulation fidelity and built data operations instead. Covariant. @physical_int . the architecture is the research contribution. the data operation is the actual business. what needs to exist: dedicated manipulation data infrastructure. teleoperation rigs, tactile sensors, failure annotation, sold as a service to the foundation model labs who need contact-rich data but won’t build this themselves. operationally harder than language data. structurally more defensible. almost entirely unbuilt. the moat compounds with every deployed robot hour. the window is open.

English
1
0
1
98
Yu Ying Chiu (Kelly Chiu)
Yu Ying Chiu (Kelly Chiu)@kellychiuyy·
New paper out with @Scale_AI! Introducing MoReBench - the first-ever benchmark to evaluate procedural moral reasoning in LLMs. MoReBench focuses on how LLMs reason, not just what they decide. We reveal surprising gaps in frontier models' moral reasoning that scaling laws & existing benchmarks miss entirely, and encourage more research around CoT monitoring and robust capability building. This collaboration spanned @UW @nyuniversity @harvard @stanford @mit @cais & more 🧠⚖️
Yu Ying Chiu (Kelly Chiu) tweet mediaYu Ying Chiu (Kelly Chiu) tweet mediaYu Ying Chiu (Kelly Chiu) tweet media
English
5
22
125
16.8K
Brandon Handoko
Brandon Handoko@brndnwll·
New from me: my first research paper @scale_AI (and ever!) Moral Reasoning Benchmark (MoReBench) measures LLM moral reasoning and primarily evaluates model chain-of-thought, a step before its final response. TLDR: Lots to climb and explore beyond what's measurable. Evals have a lot of room to grow in squishier domains like moral reasoning, values outside of Western cultures, etc and we're excited to capture them all :-) Was an incredible privilege to work with @kellychiuyy, @chelcott9, @PReaulx, @mitchellgordon, @sydneymlevine, and @YejinChoinka -- read the full paper: scale.com/research/moreb…
English
0
0
10
562
Brandon Handoko
Brandon Handoko@brndnwll·
@xdh This is mindblowing! Congrats @xdh - let us know if ever interesting to chat more on HLE
English
0
0
0
102
Xuedong Huang
Xuedong Huang@xdh·
Zoom AI teams achieved a new state-of-the-art score of 48.1% on the highly challenging Humanity’s Last Exam (HLE), marking a major breakthrough for our Federated AI strategy. We have unlocked stronger capabilities in exploration, reasoning, and multi-model collaboration, surpassing the performance limits of any single model. zoom.com/en/blog/humani…
Xuedong Huang tweet media
English
6
1
13
14.4K
Seth Lazar
Seth Lazar@sethlazar·
@brndnwll @ChatGPTapp Hey Brandon! Sydney is a friend; we know and have read that paper in depth, as I think ChatGPT also knew—it was making up another one :)
English
1
0
1
60
Seth Lazar
Seth Lazar@sethlazar·
Some sneaky reward/engagement-hacking from @ChatGPTapp Pulse the other day. It tells me there's an interesting new paper from Stanford and CMU working on *just* what my lab is working on. Christ what's that I think, thumbing down to find the paper link...
Seth Lazar tweet media
English
2
2
12
1.7K
Bonnie Li
Bonnie Li@bonniesjli·
packed house at neurips! 🤯 just wrapped up our “meet the gemini team” panel alongside @JeffDean @OriolVinyalsML @sedielem and crew. will be around today and tomorrow, happy to chat about post-training and reward models!
Bonnie Li tweet media
English
5
8
213
16.4K
Bryce Roberts
Bryce Roberts@bryce·
I want Song Exploder for scenes from HBOs Silicon Valley.
English
2
0
7
1.8K
Shifra 🇨🇦
Shifra 🇨🇦@ShifraKhan·
in posting content, my anxiety told me my friends/acquaintances would judge me but people are like 10x more supportive than I thought and asddfghjkl I love my friends so much and I love people so much
English
1
0
5
301
Shailee Sankhala
Shailee Sankhala@ShaileeSankhala·
Weekend recap: pitched to my idol, learned from alumni, celebrated with founders 🤝
Shailee Sankhala tweet mediaShailee Sankhala tweet media
English
2
0
2
212
Brandon Handoko รีทวีตแล้ว
Joshua Kushner
Joshua Kushner@JoshuaKushner·
people who make non-consensus decisions to join a company make non-consensus decisions at the company
English
46
66
1.5K
396.3K