A father and hexdump tattoo owner. Opinions are my own, except when they are my wife's. he/him.
Israel
Joined November 2008
- Tweets14.9K
- Following1.3K
- Followers6.3K
- Likes7.5K
Pinned Tweet
fuzzing UTF-8 strings pro-tip: Ⱥ (U+023A) and Ⱦ (U+023E) are the *only* code points to increase in length (2 to 3 bytes) when lowercased.
It’s not scope creep, it’s featuremaxxing ✨
More to come from the magnificent Check Point AI Research team 🧙♂️✨ blog.checkpoint.com/ai-secur…
Shahar Tal retweeted
"Revisiting SecureROM Adventures".
Talk is about the SecureROM and an usbliter8 variant by iH8sn0w
hexacon.fr/conference/speake…
Nice project, but even more so a screenplay-worthy emotional rollercoaster tweet with shock, reveals and plot twists. I’d binge it on Netflix.
I wrote a new Task Manager that runs natively on Windows, macOS, and Linux. You can get it now at tmog.org ... but why?
Well, it started as an argument over whether you could vibe code Microsoft Word today. I didn't think so. But I figured... maybe something simpler like Task Manager. But you can't just tell Claude, "Hey, write a task manager!" Well, you can, but it'll be hot garbage.
Then my kid blew out his appendix and needed surgery, so I had a couple of days in waiting rooms to tinker with the idea. I broke out my new M5 MacBook Air, and I put together a 107-page spec, like I was some caffeine-addled PM from the 90s. Everything you could ever want to know about a task manager, how it works, what it should do, and what it should look like.
By the time he was out of the hospital, I had a rough app working. But then he started bleeding internally, needed emergency surgery and a blood transfusion. Happy to say he's all better now - but suffice to say I had the rest of the week to tinker as he healed.
But why write a Task Manager? Because when I wrote the original in '95, it was on a P90 or something equally weak. Updating, scrolling, and painting without using more than a percent or two of the CPU was the modest goal.
Now that we have 5 GHz CPUs and GPUs to help, why not push the limits? Animate everything at 60Hz, instead of once a second. History mouse tracking, a summary page, and all the features I would have added then if the hardware had allowed it.
My autism leads me to embrace neon and high contrast in a way that's not to everyone's taste, so there's a saturation slider to control the color. But this one goes to 11.
You can also select from a bunch of DEC VT220-inspired themes like Green Phosphor, Amber, or White. And I added Blue to match my VSCode theme!
The code is fully native: Direct2D/DirectComposition on Windows, UIKit on Mac, and Qt on Linux. The UI communicates with a shared CPP core, which branches out to platform-specific helpers. There's NO electron, HTML, JSON, etc.
Thanks to David and Rutger for their help along the way, I couldn't have done it this quickly on three platforms without them!
If you've got some cycles to spare, give it a try and let me know if you run into any issues!
tmog.org
Shahar Tal retweeted
Once again, I had to explain to Codex why the code it wrote was a C++ slopfest, then force it to optimize it for 10 hours straight until it achieved a 200× speedup.
I asked it to self-reflect on the whole thing and write down. TBH, this is a gold mine:🧵
Shahar Tal retweeted
New exploit: “xor dword [0xf80c2094], 1<<22”
Unlocks CPU microcode, the platform security processor, system management mode, and every internal processor register, all at once, on 100 million AMD CPUs. As far as I can tell can’t be fixed.
github.com/xoreaxeaxeax/skit…
Shahar Tal retweeted
Opus 5 is a great model for coding, data analysis, design, biology, knowledge work.
More than any of these eval scores, what is most exciting to me is something else: Opus 5 is our least prompt injectable model yet. It is a bit buried in the system card, but across PI evals and red teaming, Opus 5 is very hard to prompt inject successfully.
And when layering defenses -- strong model alignment, combined with prompt injection probes, combined with Auto Mode in Claude Code -- the success rate for prompt injection attacks drops to ~0. This is new and exciting! More about this soon.
www-cdn.anthropic.com/c5fbac…
Shahar Tal retweeted
These are the real numbers of Israeli @BlackHatEvents #BHUSA talks and speakers, thanks to @StAJect0r
Also a lot of interesting other trends in the Israeli security community
stavc.github.io/blackhat-il-…
@doritsos @meirorbach @amitaiz
Went through the @BlackHatEvents #BHUSA schedule - now I've got serious FOMO.
Tons of great talks, nice balance of veterans (e.g. @albinowax @natashenka @hdmoore @hasherezade @vanhoefm) & new faces.
Israeli 🇮🇱 speakers roughly 30% overall, 50% in AI 👀
blackhat.com/us-26/briefings…
Shahar Tal retweeted
Gotta love the LLMs' standard response - because it's exactly what any mathematician would say if you showed up with this out of the blue.
hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final
((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)
Shahar Tal retweeted
hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final
((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)
Shahar Tal retweeted
Thank you, @tomocchino, for the shout-out at @vercel Ship New York!
Two months ago, we announced a partnership with Vercel to solve the trust problem that was holding AI agents back.
Together with Vercel’s Chat SDK, we integrated the first approval system to give humans the final call on every agent action, delivered natively in Slack, WhatsApp, Teams, and a dozen other messaging channels.
Appreciate the love from a great partner 🩵🖤
Shahar Tal retweeted
If you thought AI progress was slowing down, well here's the immediate answer to that. Huge jump in capability across the board. This is going to deliver major improvement in agents across almost all knowledge work categories.
Shahar Tal retweeted
"You cannot govern a technology you have only been briefed on."
Dr. @VivianBala's challenge to his fellow legislators has become a global rallying cry.
Last week, @takahiroanno, leader of Japan's @team_mirai_jp, cited Minister Balakrishnan's hands-on use of @NanoClaw_ai in a parliamentary debate and offered to personally tutor the Prime Minister on setting it up. The PM said yes.
Officially volunteering to fly out and set NanoClaw up for Japan's PM myself.
Offer extends to any head of state whose country has good food.
Shahar Tal retweeted
Introducing NanoCo, our enterprise offering from the team behind @NanoClaw_AI, backed by a $12M seed round led by Valley Capital Partners, with participation from Docker, Vercel, Monday.com, Hugging Face's Clem Delangue, and many more incredible investors.
Shahar Tal retweeted
Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946.
For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids.
An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better.
This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.
Shahar Tal retweeted
We are investigating unauthorized access to GitHub’s internal repositories. While we currently have no evidence of impact to customer information stored outside of GitHub’s internal repositories (such as our customers’ enterprises, organizations, and repositories), we are closely monitoring our infrastructure for follow-on activity.
