the divinity in you has awoken the divinity in me. I wish greater wealth, happiness and prosperity for you.

Joined October 2018
Crypto Lelouch retweeted
1M+ views in less than a week on a brand new account Claude Opus 5.5 creates these in less than 10min. Check the first comment if you want the entire workflow!
35
4
55
3,530
Crypto Lelouch retweeted
53
320
15
3,834
136,754
Crypto Lelouch retweeted
Quantized to Hell and Back !! DeepSeek 4.1F 1017-STQ-1.75bpw is released with Sushi 🍣 1.4.0, its quality is comparable to oQ2.5e & rendered this grim reaper animation. ✅ 1m context ✅ 96GB & 128GB Mac ✅ tok/s - 1,000 prefill, 75 code, 45 prose 🍣👌 github.com/beamivalice/sushi
34
25
8
231
21,845
Crypto Lelouch retweeted
How to start growing your app on TikTok, PART 3 (all organic, no ads) My tiktoks get me ~2M views a month. And I don't even make the content myself!! Claude makes all of it. Here's my whole setup: 1. Every night a Claude Code routine fires 2. Pulls photos from Pinterest 3. Checks recent hooks, no repeats 4. Writes posts off my viral hooks 5. Plugs my app in like competitors do 6. Sends drafts to Telegram + Postiz 7. Creators add a trending sound and post 8. Every post = A/B test: hook, topic, app mention 9. Stats land in Notion every morning 10. Tracks downloads per post + flags shadowbans 11. Next batch copies what won I don't even have to watch any of it... I put this ALL on my GitHub: github.com/pvlppv/tiktok-gro… Next parts: > How do I test what's working and what's not? > When should I hire creators or start TikTok ads? And more!! Follow so you don't miss them!!
19
2
108
6,397
6 UI resources for your next vibe coded game 1. Game UI Database: free, over 78,000 screens from 1,975 games, with phone and tablet filters and video of transitions. gameuidatabase.com 2. Gamedexy: mobile free-to-play flows recorded in order, covering onboarding, shops, daily rewards and battle passes. Pro connects to Claude Code and Codex. gamedexy.com 3. Flomob: mobile game screens with AI search and copy to Figma. flomob.com 4. Interface In Game: mostly PC and console, but great for menus, stores and tutorials. interfaceingame.com 5. Kenney UI Pack: free buttons, panels and icons you can drop straight into your game. kenney.nl/assets/ui-pack 6. Mobbin: the one I use most at work. It has no games, but it's the best place to see how top apps handle sign-up, settings, paywalls and onboarding, which every game needs too. mobbin.com For games, the first five are the closest thing to Mobbin I've found.
15
29
2
443
22,998
Crypto Lelouch retweeted
We're thrilled to announce the Superpower Peptide Library. Every peptide, mapped to the body. 152 peptides, from approved drugs to compounds still in trials, sorted by where they act. Each one has its own page, reviewed by a physician and fully cited, covering: - the evidence and how strong it is - its FDA status - it's mechanism of action - what studies in humans have found Because not all peptides are equal. Semaglutide is FDA-approved. BPC-157, maybe the most talked-about peptide for recovery and repair, is not. In a 2025 review of its research, 35 of 36 studies were in animals or the lab. Know the science before you buy the hype. Check it out here: superpower.com/guides/peptid…
195
801
152
8,508
1,347,871
Crypto Lelouch retweeted
Runpod, a $325M company, increased monthly new paying customers 300% in just 90 days using AI search. here’s the genius AEO system behind it: 1. They went from tracking 50 prompts to 300 almost overnight -> The prompts were split by buyer persona, funnel stage and niche technical use case instead of just tracking broad category searches -> Their buyers included DevOps engineers, AI builders, researchers and platform teams comparing Runpod against AWS, CoreWeave and Google Cloud 2. ChatGPT became the channel they prioritized -> Around 98% of Runpod’s AI referral traffic was coming from ChatGPT -> So instead of treating every AI engine equally, they focused their monitoring and optimization around the channel already sending real buyers 3. Their audit found a technical problem most teams would have missed -> Several Runpod articles were only partially indexed -> AI could see the title, but not properly access the body content -> They fixed metadata, restructured the pages and corrected formatting before producing more content 4 They built prompts around extremely specific technical buying questions -> “How do serverless GPU platforms compare to traditional cloud infrastructure?” -> “What’s the best way to manage GPU provisioning and auto scaling for AI workloads?” -> The goal was to track the questions their actual ICP asks, not generic high-volume keywords 5. They used the prompts that were already working to decide what to publish next -> Runpod reviewed which questions consistently surfaced the brand -> Those prompts became new content titles, angles and topics 6. Then they exported the entire dataset -> Prompt performance was filtered by topic, buyer persona and stage of the customer journey -> Instead of a marketer guessing what article to write next, the content roadmap came directly from gaps in AI search 7 They paired the prompt data with deeper research -> Runpod used ChatGPT Deep Research to investigate trending themes, audience questions and gaps in existing content before producing the page 8. They reverse engineered the sources AI was already using -> Runpod tracked which owned pages, competitor sites and third-party domains kept getting cited -> This exposed both content opportunities and external sites where competitors had stronger visibility 9. Publishing became a feedback loop -> Find a high-value prompt -> Research why Runpod is winning or losing it -> Publish or fix the relevant page -> Monitor whether Runpod starts appearing more often -> Double down on the prompts that move 10. They brought content production in-house -> Once prompt data and source data were feeding the strategy, Runpod no longer needed to outsource the entire research/content process -> The team could produce around the exact questions and visibility gaps they were seeing in real time 11. They cut the cost of their AI search stack by more than 80% -> Their tooling cost dropped from roughly $3,000/month to $500/month while prompt coverage expanded from 50 to 300 12. Then it turned into actual customers -> ~28,000 visitors from ChatGPT -> ~2,100 conversions -> 8% conversion rate -> ~40 new customers per day from ChatGPT -> roughly a 2-week close cycle within 90 days, Runpod was bringing in 4x more new paying customers every month. they didn’t treat AEO like “publish more blogs.” they built a feedback loop around what buyers ask, what AI recommends, which sources it trusts and what actually turns into customers. a $325M company turned ChatGPT into one of its top acquisition channels. why aren’t you? if you want us to run an audit for your company & make you rank #1 in AI search, reply “audit”
61
15
3
355
51,246
Crypto Lelouch retweeted
🌍 WanGP v18.81 — The One World Generator Build a world from text with Hy-Nomas or from an image with EVOKE, then explore it with your keyboard! Save your progress, record your journey and revisit familiar places—starting at 8 GB VRAM thanks to MMGP v4. 🤖 Deepy Prime your local Agent gets roughly 2× the context at the same VRAM, plus new abilities: 🔎 Search the web and find compatible LoRAs 🐍 Create graphics, animations and batch workflows with Python 🧊 Generate 3D assets and explorable worlds 🎬 Restore, upscale and outpaint videos through Media Flow Free to run locally. Your GPU has more places to take you. github.com/deepbeepmeep/WanG…
4
16
3
133
13,531
Crypto Lelouch retweeted
reverse engineer any ios app. app design → github.com/Appllama/appllama… best practices → github.com/expo/skills animations → github.com/emilkowalski/skil…
25
246
16
2,378
156,709
Crypto Lelouch retweeted
It's a much bigger jump than previously thought after some more tweaks. 21% increase without spec decode, 32% with spec. These numbers are averages after running 10 requests (sequentially) in 3 separate servers to minimize any variance.
New all-reduce kernels that optimize for small-message latency across PCIe-connected GPUs. 12% end-to-end decode speedup for DeepSeek-V4.1 Flash on 4x RTX 6000 Pro Server GPUs. github.com/dphnAI/sonar/pull…
8
9
119
6,656
Crypto Lelouch retweeted
DeepSeek-V4.1-Flash-1017 decodes code at 80.5 tok/s on one Mac Studio M5 Ultra 256 GB with BobClawblaw's recipe on TensorFold. Setup: – Recipe and TensorFold branch deepseek-v41-main (72a552ed), his own tool and settings – Context 524,288 tokens, 4 streams, DSpark drafts on, temperature 0 – Speed and quality with thinking off, except GPQA with thinking Speed: – Decode, code: 80.5 tok/s – Decode, prose: 70.5 tok/s – 500 words: about 9.3 s – First token after a 32K prompt: 18.6 s – 8 requests at once: 88.3 tok/s in total Quality: – MMLU-Pro: 67.5% ± 4.6 points – GPQA Diamond: 45.0% ± 6.9 points – IFEval: 90.5% ± 4.1 points – LiveCodeBench: 60.0% ± 14.5 points – GPQA with thinking: 76.7% ± 14.6 points Voxel world: – Built as an agent in 45.7 min, DeepSeek Harness 0.2.1-rc.2, one prompt, one try, no help, thinking on – Tokens written: 112,065 – Browser check: 8.5/10 Good to know: – The author runs temperature 0, the DeepSeek card suggests 1.0. I kept the author's setting. – The browser check ran on a Linux box from the same files. – Answers cut at the cap count as wrong: LiveCodeBench 14 of 40, GPQA Diamond 9 of 198, MMLU-Pro 1 of 400. Play the world in your browser. All numbers, the exact command and the raw output: niklaslenz.com/v2/deepseek-v… Recipe by @BobClawblaw on @ashxhart's TensorFold, model by @deepseek_ai.
14
11
2
120
10,750
Crypto Lelouch retweeted
Released: DeepSeek-V4.1-1017-Ablitrated-Efficient (built on @Hikari_07_jp’s testmodel)Layer-selective abliteration + targeted fixes for thinking loops and re-verification. HF: huggingface.co/pentacoxian-d…
1
10
1
65
9,518
このモデルを更に強く無検閲化しつつ精度を高めるために協力して取り組んでいます! CoxianさんはThinkingのループを大規模なFTなしで抑制する革命的な方法を実現してくれました。そのアイディアと、僕が作った倫理観が無いモデルを作った方法を組合せてより良いAblitretedモデルを目指します🚀
1
10
3
134
7,007
Crypto Lelouch retweeted
绝了!NVIDIA最新开源了2,575 个高规格 3D 资产,整整472GB!! 从日常用品到工业设备,里面收录了不少现实环境里的物体。 一揽子,全部放进了 Hugging Face。 NVIDIA SimReady Assets。 绝不单纯是一个“免费建模素材库”!! NVIDIA给每个3D 资产都带上了物理属性和语义信息,比如有多重,什么材质,重心在哪,用在什么场景等等,全都写进了参数里。 明摆着,就是用来让AI学习的! 学习怎么和真实的物理世界互动! 传统的3D资产在AI眼中,就是个有贴图的“视觉资产”,NVIDIA想做的是,把“视觉资产”变成AI可以理解的“物理资产”! 所以,这批资产的价值可想而知! 至少能够用到这些场景上! 机器人抓取和搬运训练,先在虚拟环境里练习拿取、摆放物体; 搭建工厂、仓库的数字孪生,测试设备和物体的布局; 配合合成数据工具,改变光照、视角、遮挡和物体位置,生成带标注的训练数据。 大家都在卷文字、代码与图像生成时,NVIDIA 开始为Physical AI(物理 AI) 铺设地基了。 不过,对大部分人来说,也是一个相当全面的3D资产素材库! 链接在下面,必须收藏了!
49
405
32
3,735
205,394
Crypto Lelouch retweeted
离谱!AI 最后一堵墙,终于被一个开源项目拆了!! 以前 AI 再聪明,也像个被关在笼子里的天才,想刷 X、挖 Reddit、翻小红书评论,处处受限。 现在有人直接给 AI 装上了“全网通行证”! 这个叫 Agent Reach 的开源项目,已经在 GitHub 上火到 10 万+ Star。 以前让 AI 去 X 搜帖子、翻小红书评论、看 B 站视频,经常不是读不到,就是要自己折腾一堆工具。 现在直接给 Codex、Claude Code 装上 Agent Reach,就能让 AI 跨平台搜索、读评论、提取视频字幕,连 Reddit、YouTube、GitHub 都能查。 最牛的是,不用自己写爬虫,基础功能免费,一句话就能让 AI 帮你安装。 部分平台需要登录,X 目前还是实验性支持。 以后找信息,不一定需要自己一个个平台翻了。 项目地址:github.com/Panniantong/Agent…
57
255
18
2,038
108,817
Crypto Lelouch retweeted
If an attacker parks outside your building with a $1,500 software-defined radio, a 5G smartphone will still drop its connection, connect to a fake 1991-era GSM tower, and hand over its permanent hardware identity in cleartext. Every modern baseband modem still carries legacy 2G silicon for backward compatibility, and that 2G stack has no way to verify if a cell tower is real. 1. The architectural flaw: One-way tower trust In 4G (LTE) and 5G, mobile networks enforce mutual authentication after the initial attach. When your phone connects to an LTE tower, the network sends a random cryptographic challenge (RAND) alongside an Authentication Token (AUTN). Inside your SIM card, a secure processor checks that token against a 128-bit secret root key (K) burned in at the factory. If the tower cannot prove it shares that key with your carrier, the baseband drops the connection before any voice or text data flows. In 2G (GSM), that check only runs in one direction. When GSM was standardized in 1991, commercial software-defined radios did not exist. The protocol was built so the tower authenticates the SIM card to stop billing fraud, while the SIM card never authenticates the tower. A 2G baseband processor will connect to any transmitter broadcasting a valid Mobile Country Code and a stronger signal (RXLev) than the surrounding towers. 2. How rogue base stations force the downgrade Because a rogue tower cannot forge a 4G cryptographic signature, it forces the phone's baseband to abandon 4G and 5G entirely. An attacker using an Ettus USRP B210 and a 10-watt RF power amplifier triggers this fallback in two ways: * Selective RF jamming: The radio floods local 4G and 5G downlink frequencies with noise while leaving a 2G GSM frequency clean. When the phone loses its LTE signal, the baseband's cell reselection state machine scans all supported bands and locks onto the attacker's clean 2G beacon. * Protocol downgrade injection: The attacker broadcasts a rogue LTE signal just long enough to catch the phone's tracking area update, then replies with an unauthenticated Tracking Area Update Reject (EMM Cause #7: EPS services not allowed) or an RRCConnectionRelease redirecting the modem straight to a 2G frequency. The baseband assumes you walked into a dead zone and switches to 2G in about 10 seconds without prompting the user. 3. Identity theft, A5/0, and the SMS relay catch Once the phone camps on the rogue 2G station, the attacker sends a standard Identity Request packet. The phone replies in cleartext with its 15-digit International Mobile Subscriber Identity (IMSI) and its hardware IMEI. Next comes encryption. Normal 2G traffic uses the 64-bit A5/1 stream cipher, which can already be cracked in seconds using 2TB Kraken rainbow tables. Worse, the 3GPP specification (TS 44.018) requires every phone to obey a Ciphering Mode Command set to A5/0 (null cipher). When the rogue tower sends A5/0, the phone turns off air-interface encryption completely. Any outgoing SMS or call from the victim's phone travels to the attacker's laptop as unencrypted GSM 7-bit bytes. To intercept incoming two-factor SMS codes from a bank, the attacker has to go one step further. Because a rogue SDR in a car is not wired into your real carrier's core network, the bank's SMS won't route to the attacker's laptop automatically. Either the attacker uses a second radio to relay your phone's handshake back to the real carrier's 2G tower and cracks the A5/1 downlink key, or they simply jam 4G so your phone drops onto the real carrier's 2G tower and passively crack the A5/1 air traffic in Wireshark. 4. The modern defense: Cutting legacy silicon You cannot patch a 1991 radio handshake without breaking backward compatibility, so OS vendors finally gave users a kill switch for the baseband: * Android 12+ 2G kill switch: Android added the setAllowedNetworkTypesForReason API to the Radio Interface Layer (RIL). Turning off "Allow 2G" instructs the modem firmware to disable its GSM stack at the silicon level. If 4G is jammed, the phone stays in "No Service" instead of dropping to 2G. * Android 14 & 15 null-cipher locks: Android 14 added a modem toggle to reject A5/0 null-cipher connections outright on supported hardware, and Android 15 added live system alerts whenever a tower requests an unencrypted connection or cleartext IMSI. * iOS Lockdown Mode: Turning on Lockdown Mode disables the 2G baseband stack on iPhones. * 5G Standalone (5G SA) SUCI: Even on 4G LTE, a phone still sends its IMSI in cleartext during the very first attach request before encryption starts. Pure 5G Standalone fixes this last gap: the SIM card uses ECIES public-key cryptography to encrypt your IMSI into a Subscription Concealed Identifier (SUCI) before a single packet leaves the antenna.
42
140
19
1,153
68,194
Crypto Lelouch retweeted
开源了我的 XSS 审计技能 ,过去一年我用它发现了几十个获得赏金的 XSS 漏洞,连续四个季度进入微软 MSRC Top 10 github.com/songjianyang/xss-…
10
164
4
1,225
48,483
Crypto Lelouch retweeted
— C.S.Lewis
24
2,380
79
13,170
160,819