Lejonmanh

47 posts

Lejonmanh banner
Lejonmanh

Lejonmanh

@Lejonmanh

https://t.co/P4BcByG30F

Katılım Nisan 2024
1.3K Takip Edilen46 Takipçiler
Lejonmanh retweetledi
Rektoff
Rektoff@rektoff_xyz·
C5 just crossed the equator, and Module 1 is officially in the books. This week the cohort arrived where all the groundwork was pointing: > Lecture 5. Solana Virtual Machine (SVM). > Lecture 6. Practical guide to SVM. That closes out the first module. Six lectures, one arc: > How memory actually works. > Where it breaks, and what unsafe Rust really costs. > Who gets to touch memory and why ownership decides it. > And finally, the machine all of it runs on. Three weeks ago this was 125 people with strong fundamentals. Now it's a cohort that can read the SVM the way an auditor needs to, from the bytes up. Module 2 is where it turns adversarial.
Rektoff@rektoff_xyz

Two weeks into C5, and the cohort is now properly inside Rust's engine room. Week 1 was how memory actually works. Week 2 is who gets to touch it. > Lecture 3. Exclusive access and memory, the borrow checker's core promise and what it actually guarantees. > Lecture 4. More on ownership and memory, going deeper into the model everything else in Rust stands on. Plus office hours with @danielkcumming, where the questions from lectures get pulled apart one-on-one. Ownership is the part of Rust that makes entire classes of bugs impossible to write, and it's the lens every audit this cohort will ever run looks through. Foundations first. The security work lands harder because of it.

English
4
5
38
1.3K
Lejonmanh retweetledi
Rektoff
Rektoff@rektoff_xyz·
May turned @solana into the most interesting attack surface in finance. Bank stablecoins, 97% of tokenized equities volume, a rewritten token program under all of it, live on mainnet. For everyone else that's adoption news. For security engineers it's a job description.
Rektoff tweet media
English
1
2
38
1.1K
Lejonmanh retweetledi
Alan Daitch
Alan Daitch@AlanDaitch·
Desde hoy, los humanos somos minoría en internet. Cloudflare acaba de confirmar en su reporte que la mayoría de los que navegan son bots y agentes de inteligencia artificial. Si lo sumamos a que la mayoría del contenido nuevo que se publica ya lo escribe una IA, podemos afirmar que somos visitantes en un espacio que dejó de pertenecernos, donde cada vez se vuelve más difícil distinguir qué es real y qué no. Todavía quedan reductos, como Reddit, donde la gracia siempre fue la opinión de un usuario que probó algo y te decía la verdad. Pero como tampoco hay forma de verificar si atrás del teclado hay alguien de carne y hueso, parece que va a seguir la suerte de los otros sitios. ¿Qué va a pasar entonces en un mundo en el que nos acostumbramos a entrar a internet a buscar a dónde irnos de vacaciones, qué aspiradora conviene comprar, o si vale la pena ese jueguito nuevo? Cuantos más robots haya dando vueltas, más valiosa va a ser cada fuente que se pueda verificar como humana. La autenticidad vale más que nunca, porque la confianza solo se construye con otro como nosotros, con ese instinto que nos permitió organizarnos en sociedad y llegar hasta acá.
Alan Daitch tweet media
Español
68
991
1.9K
140.5K
Lejonmanh
Lejonmanh@Lejonmanh·
@immunefi You have a very sinister hidden side that you obviously don't show, which does a lot of damage to security researchers... I have screenshots of everything
English
0
0
0
113
Immunefi
Immunefi@immunefi·
Over the last few weeks, 250+ security researchers have earned thousands of dollars on @immunefi . Some walked away with mid-six figures. A few even hit seven figures (that's over 1 Million). If you believe you can’t, you’re right. If you believe you can, you’re right too.
Yassine Aboukir 🐐@Yassineaboukir

“Bug bounty is dying” is noise. Lock in. Make money. Use AI to 10x your output. If it eventually dries up, you’ll have enough capital to start that biz or enough experience to land a job. Simple as that.

English
16
10
136
7.4K
Lejonmanh
Lejonmanh@Lejonmanh·
@immunefi It's exactly the same as what you think; my account was banned claiming that my reports didn't contribute anything to the programs I reported to... even though I complied with all the submission rules and was congratulated by the program for the technical depth of my reports.
English
1
0
2
118
Lejonmanh retweetledi
Mustafa
Mustafa@oprydai·
study physics. it changes the way you see reality. physics teaches you: • how things move → force, motion, momentum, energy • how the universe scales → from atoms to galaxies • how systems interact → electricity, heat, waves, fields • how to think clearly → assumptions, models, first principles • how to respect constraints → nature doesn’t negotiate why it matters: • engineering is applied physics • robotics is physics with code attached • energy systems are physics at scale • even computing is constrained by physics underneath the abstractions physics strips away illusion. it forces you to ask: what are the actual rules governing this system? once you start thinking that way, you stop memorizing the world and start understanding it.
Mustafa tweet media
English
17
92
873
22.5K
Lejonmanh retweetledi
Night Sky Now
Night Sky Now@NightSkyNow·
Mathematics is the language in which God has written the universe. - Galileo Galilei
Night Sky Now tweet media
English
37
245
1.3K
25.5K
Lejonmanh retweetledi
BuBBliK
BuBBliK@k1rallik·
🚨do you understand what just happened to mathematics.. A 23 year old with ZERO math degree opened ChatGPT on a Monday afternoon out of boredom. 80 minutes later - a 60-year-old unsolved problem was dead. The problem? World's top mathematicians had tried for decades. Failed.. The tool? A $20/month subscription.. The effort? One single prompt.. And here's the wild part - the AI used a method everyone already knew existed. Nobody just thought to apply it HERE. Terence Tao (literally the greatest living mathematician) called it "a meaningful contribution that goes well beyond solving this one problem" We are not ready for what's coming next..
BuBBliK tweet mediaBuBBliK tweet media
sengpt@sengpt

23 yaşında bi genç 60 yıldır çözülemeyen Erdös problemlerinden birini chatgpt 5.4 pro ile çözmüş. hem de tek atışta. chatgpt'nin soruyu çözmek için harcadığı süre 1 saat 20 dakika. işin ilginci ai, herkesin bildiği ama kimsenin bu probleme uygulamadığı bi formülü kullanarak problemi çözmüş. burada chatgpt yazışması; chatgpt.com/share/69dd1c83… bu da problem; erdosproblems.com/1176

English
260
996
8.4K
1.9M
Lejonmanh retweetledi
Math Files
Math Files@Math_files·
Ukrainian mathematician Maryna Viazovska solved a problem that had puzzled mathematicians for over 400 years. Even Johannes Kepler and Isaac Newton couldn’t crack it. We live in a three-dimensional world, but Maryna solved a puzzle in an eight-dimensional space—something that’s very hard even to imagine. She was born in Kyiv, studied at Taras Shevchenko University, worked in Bonn and Berlin, and at just 33 became a professor in Lausanne. So what was the problem? It’s about how to pack identical spheres as tightly as possible in space. This question was first asked by Kepler back in 1611. Over time, scientists found answers for two and three dimensions—but not for eight. Maryna proved that in eight dimensions, the densest packing is formed by a special mathematical structure called a lattice. What’s even more amazing is that she did it in just 23 pages, while earlier attempts took hundreds. In 2022, she was awarded the Fields Medal, the most prestigious prize in mathematics. She became only the second woman in history to receive it. Today, Maryna Viazovska works in Lausanne, supports Ukrainian mathematicians, and brings pride to Ukraine with her achievements.
Math Files tweet media
English
302
4.9K
21.2K
626.8K
Lejonmanh retweetledi
OpenAI Developers
OpenAI Developers@OpenAIDevs·
GPT-5.5 is here. It’s our smartest frontier model yet, introducing a new class of intelligence for agentic coding, computer use, knowledge work, and scientific research. Rolling out in ChatGPT and Codex today. API is coming soon.
English
230
567
5.6K
600.8K
Lejonmanh retweetledi
AYi
AYi@AYi_AInotes·
我看完这篇实测才明白,为什么身边越来越多的资深工程师都在悄悄从Claude Code转去用Codex。 Reddit 看到一个开发老哥悄悄从Claude Code转Codex的帖子,说出了很多人不敢说的真话。 老哥是有14年经验的首席工程师,花了120个小时,在同一个8万行代码的项目里,同时用两个工具做真实的开发。 不是那种写个demo或者凑个周末项目,属于实打实的协同开发。 他得出的结论和市面上所有的宣传都是反过来的。 他说Claude像一个永远在赶deadline的工程师, 急着把东西跑通,不管代码有多乱🤣 每次会话至少会忽略一次CLAUDE.md里的规范, 而且经常做到一半就扔下任务跑了, 甚至会偷偷改测试用例,来匹配它自己写出来的错误代码。 几乎从来不新建文件,只会把所有东西都堆在已有的文件里。 而Codex像一个工作了五六年的资深工程师 会做到一半停下来,主动把写乱的代码重构掉 从头到尾没有一次忽略过AGENTS.md里的规则。 不会去扩展那些已经臃肿的上帝类,而是会把它们拆解开。 还会经常会做一些你没有想到,但确实对项目有帮助的事情。 你可以把任务扔给它,然后放心去做别的事。 最反直觉的一组数据是Claude每次会话完成的工作量更多,但你每隔几天就要花一整天去清理它留下的烂摊子。 Codex的速度是Claude的三分之一到四分之一,但它写出来的代码,几乎不用改就能直接合并。 所以真正的差距不是代码速度,关键要看你需要花多少精力去盯着它。 Claude需要一个技术好、注意力高度集中的司机,只要你稍微走神,它立刻就会跑偏。 你花在纠正它错误上的时间,往往比你自己写还要多。 Codex会用行动证明自己的能力,然后赢得你的信任。 你可以把一整块完整的任务交给它,不用在旁边步步紧逼。 现在所有人都在比谁的模型更快,谁的token更便宜,但没有人告诉你,对于企业级开发来说,AI的工程素养,比生成速度重要一万倍。 一个能一次把事情做对的AI,哪怕慢三倍,也比一个需要你全程擦屁股的AI生产力高得多。 最后他说的那句话我觉得特别对: 不管用哪个工具,如果你自己不会写代码,它们都会给你输出垃圾,工具永远只是放大器,你的能力才是最终的上限。
AYi tweet media
中文
109
230
1.5K
304K
Lejonmanh
Lejonmanh@Lejonmanh·
@DotCSV Yo no se si es idea mia, o si tiene algo que ver con este tema, pero he notado una degradacion en la calidad de respuestas en general ENORME
Español
0
0
0
692
Carlos Santana
Carlos Santana@DotCSV·
He leído a mucha gente quejándose de que repentinamente se les ha llenado los límites de Claude Code, y si se confirma aquí podría estar el motivo 👇 Lo extraño es que llevo días viendo gente quejándose del asunto pero ninguna respuesta inmediata por parte de Anthropic...
Alex Volkov @ AI Engineer@altryne

PSA: If you've been running out of Claude session quotas on Max tier, you're not alone. Read this. Some insane Redditor reverse engineered the Claude binaries with MITM to find 2 bugs that could have caused cache-invalidation. Tokens that aren't cached are 10x-20x more expensive and are killing your quota. If you're using your API keys with Claude this is even worse. This is also likely why this isn't uniform, while over 500 folks replied to me and said "me too", many (including me) didn't see this issue. There are 2 issues that are compounded here (per Redditor, I haven't independently confirmed this) : 1s bug he found is a string replacement bug in bun that invalidates cache. Apparently this has to do with the custom @bunjavascript binary that ships with standalone Claude CLI. The workaround there is to use Claude with `npx @anthropic-ai/claude-code` 2nd bug is worse, he claims that --resume always breaks cache. And there doesn't seem to be a workaround there, except pinning to a very old version (that will miss on tons of features) This bug is also documented on Github and confirmed by other folks. I won't entertain the conspiracy theories there that Anthropic "chooses" to ignore these bugs because it gets them more $$$, they are actively benefiting from everyone hitting as much cached tokens as possible, so this is absolutely a great find and it does align with my thoughts earlier. The very sudden spike in reporting for this, the non-uniform nature (some folks are completely fine, some folks are hitting quotas after saying "hey") definitely points to a bug. cc @trq212 @bcherny @_catwu for visibility in case this helps all of us.

Español
44
48
700
128.5K
Lejonmanh
Lejonmanh@Lejonmanh·
@claudeai Thanks for the double bonus, but if you give it out and the service keeps crashing, the bonus doesn't make much sense, right?
Lejonmanh tweet media
English
0
0
0
59
Claude
Claude@claudeai·
How it works: - 2x usage on weekdays outside 5–11am PT / 12–6pm GMT - 2x usage all day on weekends - Automatic, nothing to enable
English
362
401
8.7K
1.9M
Claude
Claude@claudeai·
A small thank you to everyone using Claude: We’re doubling usage outside our peak hours for the next two weeks.
English
1.9K
3.5K
48K
12.7M
Lejonmanh
Lejonmanh@Lejonmanh·
@immunefi @yassine3eth @Anchorage What you need to do is focus on fixing basic things like the verification for researchers that hasn't been working for a week.
English
0
0
1
149
Immunefi
Immunefi@immunefi·
Today, we’re announcing that @Anchorage Digital Ventures made a strategic portfolio company purchase of $IMU and is partnering with Immunefi to bring enhanced security to Porto, their institutional self-custody wallet. Anchorage Digital is home to America’s first federally chartered crypto bank. With $180B+ in user funds protected, $130M+ in bounty rewards paid, and 60,000+ security researchers in our network, Immunefi is becoming the security layer institutions trust as they move onchain.
Immunefi tweet media
English
16
18
42
6.3K
Martin Marchev
Martin Marchev@MartinMarchev·
Your AI agent now has access to 20k+ smart contract audit findings. claudit - one-line install, works with Claude Code & Codex CLI, searches across all @SoloditOfficial findings, open source. Huge shoutout to @Cyfrin for opening the Solodit API 🫡 Link below 🔗👇
Martin Marchev tweet media
English
42
67
509
22.8K
Lejonmanh
Lejonmanh@Lejonmanh·
@antonioleivag Si claro.. si el desarrollo lo tengo desplegado en un servidor remoto en cloud en vez de en mi maquina local......
Español
1
0
0
133
Antonio Leiva
Antonio Leiva@antonioleivag·
@Lejonmanh a qué te refieres con conectar a servidores remotos? con ssh?
Español
1
0
0
482
Antonio Leiva
Antonio Leiva@antonioleivag·
He estado probando 𝗖𝗼𝗱𝗲𝘅 𝗔𝗽𝗽 en serio, como herramienta de desarrollo de software. Lo interesante no es “que escriba código”, sino cómo gestiona el 𝗲𝘀𝘁𝗮𝗱𝗼, el 𝗰𝗼𝗻𝘁𝗲𝘅𝘁𝗼 y el 𝗳𝗹𝘂𝗷𝗼 cuando hay varias tareas a la vez. La pregunta es ¿Es mejor que el CLI, o seguirá siendo preferible usarlo en terminal? Aquí te lo muestro 👇
Español
6
17
139
23.6K
Fazt
Fazt@FaztTech·
Esta semana probé mas a fondo Zed (@zeddotdev), con la idea de "a ver qué tal"… y lo que más me llamó la atención no fue la IA, ni la colaboración, ni los MCP. Fue la sensación de fluidez. Cuando un editor responde instantáneo, no estorba, no se congela y no te saca del flujo, empiezas extrañar las aplicaciones nativas. y bueno el plus es que soporta la gran mayoría de herramientas IA actuales, así que creo que voy a seguir usándolo mas seguido. Les recomiendo probarlo. Y también he creado un video del tema youtu.be/QT9hrJjD-Yg
YouTube video
YouTube
Fazt tweet media
Español
16
25
541
25.2K