word
I was quite emotional this morning. I davened (prayed) at 6am. Then, I voted. In democratic elections. I then read about the $2+ Billion acquisition of ๐ฎ๐ฑ quantum computing company Qlassic by intel. I was thrilled for our friends and colleagues who backed and built Qlassic. And, on a small personal level, I am excited by the new book I was privileged to write that is arriving in stores today.
On the voting line at 7 a.m. I met a female soldier with an M16 slung on her shoulder. A yeshiva student. Elderly people and professors. Some with yarmulkes and others without. We live in neighborhood in this city where you see everything and everyone.
I also thought about my children who are voting at special ballot boxes because theyโre in the army or on the front lines now.
I want you to stop and think for a minute. Almost any other country that would have gone through what we in Israel went through would not have made it to this moment. Massive protests, covid, a 7 front war, fractious politics and cacophonous social media. Indeed, arguments and disagreements are very present โ but even this morning, people who will vote differently from one another at the ballot box will then get in the same tank and share a shabbat dinner. They are willing to give up their body and soul, one for the other, for something greater than themselves.
Before I cast my vote this morning, I paused and thought about those who unfortunately and tragically didnโt return from the front lines and the hostages that did not return from captivity. Our biological family and our Israeli family. It is our first election since October 7th. My eyes teared up.
I then thought about the heroism and self-sacrifice of young Israelis: the hundreds of days of service by our soldiersโour childrenโand their families stationed on the front lines and holding down the home front so that we may merit to live as free people in our land after two thousand years. This is no trifling matter. Our ancestors would have given arms and legs to have this gift. Many people around the world yearn for our freedom.
We are privileged to be part of this incredible people and this amazing country at this extraordinary time. We are fortunate and blessed. I feel personally fortunate and blessed.
Am Yisrael Chai. ๐ฎ๐ฑ
1
0
9
522
ื ืืื ืืืื. ืงืจืืคืื ืืื ืื ืืื ืืืกืชืจืช ืืกืคืื.
ืืืื ืงืฉืจ - ืืืื ืืืื ืื ืฉื ืขืืื ืงืฉื ืืื ืฉืื ืืืืื ืขืืื ื ืกื ืงืฆืืืช ืืืืืืืช.
ืืื ืืกื ืืืืขืืื ืืืื ืขื ืืืื ืืช ืืื ืจืืกืื ืื ืฆืคืื ืงืืจืืื. ืืื ืืกืื ืืืืื, ืกืื ื ืืืืื ืืช ืืืืขืื ืืจืืฉืื ื. ืื ื ืื ืจืืื ืืื ืืคืฉืจ ืืืื ืืงืืื ืืช ืืืืื ื.
ืฆืื ืืืฆืืืข ืืฉืืืฉื ืืงืจืื. ืืฉืื!
ืืืื ืงืฉืจ - ืืืื ืืืื ืื ืฉื ืขืืื ืงืฉื ืืื ืฉืื ืืืืื ืขืืื ื ืกื ืงืฆืืืช ืืืืืืืช.
ืืื ืืกื ืืืืขืืื ืืืื ืขื ืืืื ืืช ืืื ืจืืกืื ืื ืฆืคืื ืงืืจืืื. ืืื ืืกืื ืืืืื, ืกืื ื ืืืืื ืืช ืืืืขืื ืืจืืฉืื ื. ืื ื ืื ืจืืื ืืื ืืคืฉืจ ืืืื ืืงืืื ืืช ืืืืื ื.
ืฆืื ืืืฆืืืข ืืฉืืืฉื ืืงืจืื. ืืฉืื!
ืชืขืฉืืืช ืืงืจืืคืื ืืืฉืจืืืืช ืืื ืืืช ืืืืืืืืช ืืขืืื ืืืืจืืช ืงืฉืืื ืจืืืืืืจืืื ืขืฆืืืื, ืืฆืืืื ืืืขืืื ืกืืจื ืืจืืื ืฉื ืืืจืืช ืืืื ืืงืืจื ืื, ืฉืืคืื ืืืืืื ืืขืืื.
ืืงืจืืคืื ืืฉ ืืช ืืคืืื ืฆืืื ืืืืจ ืืช ืืฉืจืื ืืฉืืืงืื ืืคืื ื ืกืืื ืืขืืืืืื, ืืื ืืื ืื ืืงืื ืื ืืกืชืืจ ืืืขืืจืืช ืืกืคืื. ืื ืืืจ, ืืืืืจ ืืืขืจืืช ืืฉืงืืคื, ืืืืืื ืืืืืืจื ืืืืชืจ ืืขืืื ืืืขืืจืืช ืืกืคืื.
ืืฉืจืื ืืื ืืงืืงื ืืจืคืืจืืช ืงืจืืคืื ืืืืคื ืืืืื. ืฆืจืื ืืชืงื ืืช ืื ืื ืื ื ืืืก, ืืืฆืืจ ืจืืจืืช ืืืืงืืื ืืืฉืจืื, ืืจืืื ืฉืื ืืงืจื ืืืืฉืื ืืืื ืืื ืื ื ืืคืืจื ืืืจืืช ืืงืจืืคืื, ื ืืื ืื ืืขืฉืืช ืืืช, ืืฉืืชืืฃ ืืจืืืืืืจืื ืืืืจืื ืืืงืฆืืข, ืขื ืื ืช ืืืืง ืืช ืืืืืช ืืฉืจืื ืืืช ืงืฉืจืื ืขื ืืขืืื
3
3
36
1,787
Adam Benayoun ๐ retweeted
after months of building, excited to finally share antseed's funding round, led by spark capital
our similarity to bittorrent isn't a coincidence, it's a design choice. we want a network that's as open as possible, where anyone can take part in its economic activity
looking forward to keep building this with @amosmeiri and the whole antseed community
lfg ๐
our similarity to bittorrent isn't a coincidence, it's a design choice. we want a network that's as open as possible, where anyone can take part in its economic activity
looking forward to keep building this with @amosmeiri and the whole antseed community
lfg ๐
What BitTorrent did for files, we want to do for AI inference.
Antseed lets anyone buy and sell access to AI models, on their own terms.
The Antseed Foundation has raised $2.4M in a token round led by Spark Capital to help make that happen.
15
9
50
1,784
1/ Proud to back @AntSeed, the open peer-to-peer market for AI inference ๐
The $2.4M token round was led by @sparkcapital, with participation from @ColliderVC, @DCGco, @NorthIslandVC, @ReciprocalVC, @RelayerCapital and @AskVenice.
Here's why we did it, and why it's mostly about the people ๐งต
The $2.4M token round was led by @sparkcapital, with participation from @ColliderVC, @DCGco, @NorthIslandVC, @ReciprocalVC, @RelayerCapital and @AskVenice.
Here's why we did it, and why it's mostly about the people ๐งต
What BitTorrent did for files, we want to do for AI inference.
Antseed lets anyone buy and sell access to AI models, on their own terms.
The Antseed Foundation has raised $2.4M in a token round led by Spark Capital to help make that happen.
9
3
54
8,170
5/ We also backed @kotevcode in his previous company at Collider. When we heard he was teaming up with Amos, the round became very easy to underwrite.
Two founders we already trust, choosing to build together.
Two founders we already trust, choosing to build together.
2
0
4
146
Adam Benayoun ๐ retweeted
What BitTorrent did for files, we want to do for AI inference.
Antseed lets anyone buy and sell access to AI models, on their own terms.
The Antseed Foundation has raised $2.4M in a token round led by Spark Capital to help make that happen.
Antseed lets anyone buy and sell access to AI models, on their own terms.
The Antseed Foundation has raised $2.4M in a token round led by Spark Capital to help make that happen.
136
117
777
720,940
Adam Benayoun ๐ retweeted
We have been building CoFHE for a while.
What started as an idea to bring FHE-based privacy to blockchains is now a reality, and weโre ready for everyone to use it.
Our strategy is simple:
1. Build CoFHE and drive adoption. That includes the infrastructure and SDK, together with standards like Dual Mode Token.
2. Lead the charge by building solutions on top of our infrastructure where we see the most urgent needs. More on that in a few weeks.
3. Scale across payments, stablecoins, and tokenized assets (and more) while making privacy integration seamless for users.
Chapter one starts in October.
What started as an idea to bring FHE-based privacy to blockchains is now a reality, and weโre ready for everyone to use it.
Our strategy is simple:
1. Build CoFHE and drive adoption. That includes the infrastructure and SDK, together with standards like Dual Mode Token.
2. Lead the charge by building solutions on top of our infrastructure where we see the most urgent needs. More on that in a few weeks.
3. Scale across payments, stablecoins, and tokenized assets (and more) while making privacy integration seamless for users.
Chapter one starts in October.
Fhenix goes mainnet this October.
For years, transacting onchain came with a tradeoff: participate in open, composable markets and, with it, surrender the privacy of your financial activity.
Dual mode turns what was once an inevitable tradeoff into a property of your asset time, the same asset can be public when you want and confidential when you need- without sacrificing composibility or fragmenting liquidity.
The era of dual mode money is here, more soon.
17
11
70
2,487
Super exciting!
Fhenix goes mainnet this October.
For years, transacting onchain came with a tradeoff: participate in open, composable markets and, with it, surrender the privacy of your financial activity.
Dual mode turns what was once an inevitable tradeoff into a property of your asset time, the same asset can be public when you want and confidential when you need- without sacrificing composibility or fragmenting liquidity.
The era of dual mode money is here, more soon.
0
1
4
375
Adam Benayoun ๐ retweeted
THE CITADEL. HIGH STAKES GUNSHIP BROWSER MMO
Closed beta starts this weekend. Claim your spot now thecitadel.com
Closed beta starts this weekend. Claim your spot now thecitadel.com
2,766
2,267
3,113
260,882
Adam Benayoun ๐ retweeted
The Information is hiring our first Israel Bureau Chief as we ramp up our international expansion!
Ten years ago, we opened an Asia bureau and cemented a leading position in covering China tech.
Last month, we welcomed our first reporter in Singapore.
And as I searched for our next hub, I was looking for a place whose tech industry is woefully under covered relative to its significance, and the answer was clear.
Israel is at the center of three of the most important tech trends of the day: AI, cyber and defense.
I can't wait to build this team together with an awesome bureau chief.
Check out the role here.
ats.rippling.com/theinformationโฆ
Ten years ago, we opened an Asia bureau and cemented a leading position in covering China tech.
Last month, we welcomed our first reporter in Singapore.
And as I searched for our next hub, I was looking for a place whose tech industry is woefully under covered relative to its significance, and the answer was clear.
Israel is at the center of three of the most important tech trends of the day: AI, cyber and defense.
I can't wait to build this team together with an awesome bureau chief.
Check out the role here.
ats.rippling.com/theinformationโฆ
41
39
496
123,654
Adam Benayoun ๐ retweeted
We just ran Jev on our WebMCP benchmark.
The result: basically broke the benchmark.
Jev + Mercury 2.5 (a fast, low-cost LLM) using WebMCP solved 100% of the tasks at roughly 112ร lower model cost than GPT-6 Astra using computer use with code execution. Compared to Astra using er use, the model cost was 245ร lower (!).
We also compared Jev operating the browser with and without WebMCP.
We used Browser Useโs open-source Ultrafast, with some improvements to the harness to make it more reliable across the benchmark.
Jevโs browser-control accuracy on its own was not amazing - adding WebMCP nearly doubled the number of solved tasks, from 25/49 to 49/49, while reducing model cost by 18% (more on why below).
The benchmark and methodology are fully open and reproducible.
Full results: webmcp.com/benchmark
A few words on how the Jev + WebMCP harness works and why this is exciting:
Jev receives text as input and a set of discrete options it can choose from. With WebMCP, those options are the tools exposed by the website. At each step, Jev sees the task, the available tools and previous results, then picks what to do next.
The limitation is that Jev canโt generate arbitrary text, which you need for tool arguments. For example, it can choose the search_products tool, but it canโt generate the search query itself.
So we split the work: Jev picks the tool and Mercury 2.5 generates the arguments if needed.
This works well because turns out most of the cognitive load in these tasks is around choosing the right action. The argument generation itself is relatively simple, so we can delegate to a small and very fast model. We used Mercury, which outputs 1,000+ tokens/sec and is very cheap.
The result is a pretty simple combination: Jev for tool selection + Mercury for arguments + WebMCP for the interface. It ends up being very reliable, very fast, and very cheap.
A few words about Ultrafast and why do we think it underperforms:
Without WebMCP, Jev chooses from the pageโs controls: which button to click, which field to fill, or which option to select.
But choosing a valid button is different from choosing the right next step. The agent still has to navigate menus, understand forms, recover from errors and recognize when the task is actually complete.
Our hypothesis is that WebMCP makes the decision space much simpler. Instead of figuring out a sequence of clicks through a website, Jev chooses explicit actions that directly advance the task.
@typesafeai itself documents weaker accuracy on questions requiring multiple reasoning steps. WebMCP moves much of that complexity into the websiteโs tools, leaving Jev with clearer decisions and fewer opportunities to go wrong (in a sense WebMCP "compresses" a sequence of clicks into one tool call).
Our modified Ultrafast setup solved 25/49 tasks - that is a result for our particular implementation and benchmark, not a universal limit on Jev or Browser Use. We are open to more harness optimization to get this result to perform better, feel free to directly contribute to the benchmark here: github.com/nekuda-ai/Windโฆ
Browser-use ultrafast: github.com/browser-use/jeโฆ
The result: basically broke the benchmark.
Jev + Mercury 2.5 (a fast, low-cost LLM) using WebMCP solved 100% of the tasks at roughly 112ร lower model cost than GPT-6 Astra using computer use with code execution. Compared to Astra using er use, the model cost was 245ร lower (!).
We also compared Jev operating the browser with and without WebMCP.
We used Browser Useโs open-source Ultrafast, with some improvements to the harness to make it more reliable across the benchmark.
Jevโs browser-control accuracy on its own was not amazing - adding WebMCP nearly doubled the number of solved tasks, from 25/49 to 49/49, while reducing model cost by 18% (more on why below).
The benchmark and methodology are fully open and reproducible.
Full results: webmcp.com/benchmark
A few words on how the Jev + WebMCP harness works and why this is exciting:
Jev receives text as input and a set of discrete options it can choose from. With WebMCP, those options are the tools exposed by the website. At each step, Jev sees the task, the available tools and previous results, then picks what to do next.
The limitation is that Jev canโt generate arbitrary text, which you need for tool arguments. For example, it can choose the search_products tool, but it canโt generate the search query itself.
So we split the work: Jev picks the tool and Mercury 2.5 generates the arguments if needed.
This works well because turns out most of the cognitive load in these tasks is around choosing the right action. The argument generation itself is relatively simple, so we can delegate to a small and very fast model. We used Mercury, which outputs 1,000+ tokens/sec and is very cheap.
The result is a pretty simple combination: Jev for tool selection + Mercury for arguments + WebMCP for the interface. It ends up being very reliable, very fast, and very cheap.
A few words about Ultrafast and why do we think it underperforms:
Without WebMCP, Jev chooses from the pageโs controls: which button to click, which field to fill, or which option to select.
But choosing a valid button is different from choosing the right next step. The agent still has to navigate menus, understand forms, recover from errors and recognize when the task is actually complete.
Our hypothesis is that WebMCP makes the decision space much simpler. Instead of figuring out a sequence of clicks through a website, Jev chooses explicit actions that directly advance the task.
@typesafeai itself documents weaker accuracy on questions requiring multiple reasoning steps. WebMCP moves much of that complexity into the websiteโs tools, leaving Jev with clearer decisions and fewer opportunities to go wrong (in a sense WebMCP "compresses" a sequence of clicks into one tool call).
Our modified Ultrafast setup solved 25/49 tasks - that is a result for our particular implementation and benchmark, not a universal limit on Jev or Browser Use. We are open to more harness optimization to get this result to perform better, feel free to directly contribute to the benchmark here: github.com/nekuda-ai/Windโฆ
Browser-use ultrafast: github.com/browser-use/jeโฆ
94
169
2,052
241,973