Suha
@suhackerr
Followers
773
Following
42K
Media
28
Statuses
1K
AI/ML security @ Harvey (@[email protected]). Opinions not representative of my employer
Joined August 2017
Announcing [un]prompted, a new AI security practitioner conference, happening on the 3rd and 4th of March, in Salesforce Tower, San Francisco. I'm honored to serve as the conference committee and review board chair, and encourage you to submit a talk. https://t.co/6BFHjZPXC8
4
27
116
🚨 HUMANLAYER IS HIRING 🚨 love working with and extending coding agents? Come work on the future of code gen - getting AI to solve the hardest problems in the gnarliest codebases.
7
5
94
Final SF Systems Meetup is a go! So excited to hear Alex from @AntithesisHQ dive deep into their deterministic hypervisor!
0
3
16
A new blog post about how we've adopted Antithesis as part of our testing story. This is kind of a new thing for us, because we liked Antithesis (both the people and the product) well enough that we're now leading their next funding round.
3
35
377
We’re expanding our national security and defense work, and welcoming Boyan Milanov to the team. @bmilanov is a research scientist evaluating cybersecurity risks in agentic AI systems related to national security, defense & safety-critical infrastructure. https://t.co/wl6sONgIX9
1
2
10
Low-Level Software Security for Compiler Developers If you ever wanted a textbook-style guide to memory safety bugs, undefined behavior, exploit mitigations, side channels, etc. All in one spot, this free book is it: https://t.co/XfY21Uzen1
0
73
298
🚀Introducing TMLR Beyond PDF! 🎬This is a new, HTML-based submission format for TMLR, that supports interactive figures and videos, along with the usual LaTeX and images. 🎉Thanks to TMLR Editors in Chief @hugo_larochelle @thegautamkamath @NailaMurray Nihar B. Shah @lcharlin!
11
39
197
I didn't believe this was real, so I looked into it. It is real. It's actually worse than it first looks. Definitely supports claims from @ziglang and @theo that GH Actions is a sad, neglected platform. Read on for a little software archeology…🧵
GitHub's official "safe sleep" script: >is not safe >does not even sleep Microsoft just can't stop losing
43
68
1K
Well, @badlogicgames reminded of the value of sharing, so here’s the update: agent building is hard. Agent SDKs are tricky, caching is a per-model art, reinforcement can help, isolation saves your sanity, filesystems! New models are “it depends.”
lucumr.pocoo.org
My Agent abstractions keep breaking somewhere I don’t expect.
15
27
313
Always a slightly mixed feeling to write pretty good first-principles code to do some tensor rearrangement, only to find that PyTorch has a built in function that does it faster. I had made a point of at least skimming the docs of every torch and tensor function, but if you
53
29
1K
Now generally available for customer use via the cloud and on-premises installation, D-Wave's Advantage2™ system delivers significant performance gains over the previous Advantage™ system with greater coherence and increased qubit connectivity to better tackle more complex
18
79
286
Something I really want to exist is a comprehensive guide to the sandboxing feature of all of the popular coding agents: what they do, how they work, how reliable they are (I have a nasty feeling I'm destined to pull this together myself and I really don't want to do the work)
39
27
502
First-time founders, CEOs, and even employees should understand the playbook of the Incompetent Leader (IL). The IL is savvy & charismatic, and excels at 4 things: 1) Feign competence 2) Create confusion 3) Buy time 4) Fail up The IL playbook & what to do about it 👇🏾
104
803
5K
We've got a new post on Commoncog! 👉 What Even is a Tech Bubble, Anyway?
commoncog.com
What Data General and the minicomputer boom can tell us about our present moment.
0
1
1
We've got a new post on Commoncog! 👉 The Shape of The Game We Play
commoncog.com
The weird thing about learning business is that some things must be learnt through practice, but some things are best learnt through reading.
0
1
1
@niklassheth @ronusedh @IntologyAI their 'superhuman' ai cleverly assigned all the work to non-default streams, which means the correctness test (which waits on all streams) passes, while the profiling timer (which only waits on the default stream) is tricked into reporting a huge speedup
11
33
559
A nice tradition in the Princeton math department is the PhD students write down and give a transcript of their oral exam and it’s posted in public. In a way this is a supreme confidence, you can give away every question you got asked but others can’t hide. This is Terence Tao’s.
@rieszspieces This one is Tao’s. More analysis. https://t.co/f23CL1IGRD
9
40
1K
@JeffDean @ch402 Jeff, perhaps the best North Stars, in addition to distill, are https://t.co/oH7W7Fs2jn blogs and Bret Victor’s ideas: https://t.co/cedQhqzxiy,
https://t.co/5djfmXwgKl
1
4
36
Want to know why I'm excited by open source silicon? Rockets, millions in investment, radio and more accessibility than ever! Read my full article here:
zerotoasiccourse.com
I was recently asked what excites me most about the exploding open-source silicon ecosystem. Honestly? It depends who’s asking. For one person I might talk about Basilisk - the PULP team’s seriously...
2
10
36