Mentions

우분투 보안업데이트 늘어지는거나 snap처럼 마음에 안드는게 한두개가 아닌데 업데이트할때마다 프로 광고하는것도 그렇고.. 데비안으로 어떻게 옮기지 하다가 이런걸 찾았다 github.com/bin456789/reinstall/blob/main/README.e…

つい最近自分でマインスイーパ作ったことあったからベネ駿共テの情報第3問いつにもまして簡単だった janimaru0704/bevy_minesweeper

Online .DV to .mp4 convert tool Free DV to MP4 converter. Runs in your browser. Your files never get uploaded. Try it: dv-to-mp4-converter.pages.dev Code: swooshboots/lovable-dv-to-mp4-converter I wanted a way to convert footage during a session to be able to clip out different people’s tricks and send them the .mp4 all from mobile during the session. I couldn’t find another way to do this on mobile so I vibe coded this little app. Works great, please share any feedback any anything else you’d like to see added. Thanks for looking!

Aerolux v3: FSR + Frame Gen and Native AA become one X-Plane 12 mod Hi all! Here's an early look at where my two free X-Plane 12 mods are headed. In v3, "FSR 3.1 / 4.1 + Frame Gen" becomes **Aerolux Boost** and "Make XP Anti-Aliasing Great Again" becomes **Aerolux Pure**. They'll install together, update together, and share one new in-sim panel. The video shows the full panel concept with every feature in it, including what works now, what v3 adds, and what we're planning after that. **Working in v2 today** - AMD FSR 4.1 (FSR 3.1 on other GPUs), as native AA or upscaling - Pure: TAA + SMAA, SMAA T2x or CMAA 2 on any GPU, with no AMD files - 2× frame generation. A new "early release" change gave +18.8% real FPS in testing (31.7 → 37.6). - X-Plane's own motion data, moving-object detection, texture LOD bias, cockpit screen fix - Frame pacing profiles, auto frame cap, Anti-Lag 2, and a CPU helper with options for chiplet CPUs - Driver and GPU scheduling (HAGS) check, VR detection with safe pass-through, 7 bindable commands, Vulkan panel drawing **Coming in v3** - AI lighting (neural rendering, using your own model file). It has four strengths from Subtle to Max with local tone, Fast or Detailed models that switch without a restart, three styles, and a dark cockpit lift. - A cockpit-display mask, so glass screens are never changed - AI updates every 2 or 4 frames with a motion-carried fade, plus a change limit and edge guard - Cockpit text sharpening and a low-FPS guard for frame generation - New health checks: overlays, extra monitor windows, and auto safe-start after repeated failures - A desktop manager for every X-Plane copy. Your v2 settings carry over. **Planned after v3** - Intel XeSS upscaling and XeSS frame generation - Two-GPU frame generation (a second graphics card making the extra frames) - 3×/4× frame generation when AMD ships it for Vulkan - VR per-eye upscaling, and AI lighting on older graphics cards **Try v2 now (free, open source; Betas are marked Pre-release)** - FSR + Frame Gen: github.com/cjboy31/xplane12-fsr-framegen/releases - Native AA: github.com/cjboy31/make-xp-anti-aliasing-great-ag… - Discord (support, bug reports, news): discord.gg This is a design preview. Nothing in v3 is released yet. Feedback is welcome!

@0xtigs do you mean it's hard to differentiate the final response in claude code desktop app? i made a claude mod for that and it helps sooo much. it adds this border to its response. keshavatearth/for-your-eyes t.co

github.com/HikaruEgashira

New version of Eglantine (Nintendo Switch 2 emulator) fork by @Zortosdev A lot of improvements to gain more performance during the gameplay! Check it out! zortos293/Eglantine

💥 8GB 显卡，终于能把 Qwen3.8 开到 128K 了。 Strata 0.2.10：模拟 8GB 跑 DFlash2，解码约 63 到 73 tok/s；真机 Arc A750 大约 12.7 到 14.8。默认改走更小的 DFlash2，不再硬塞 MTP。 温馨提示：系统内存还是要够，官方测的是大内存机器。 来源：GitHub Strata👇 github.com/Niko1221/Strata/releases/tag/v0.2.10 t.co

Registration for the VOID Autumn Cup 2026 is now open⬇️ github.com/Mace-Media/VOID-Masters-2026/blob/main…

Show HN: Provenance – A Spec That Evolves with Your Code Hi all, I've been building Provenance, a tool for growing your specification of your code as your understanding of it becomes clearer. Currently I feel that the lifecycle around software improving feels too detached. We talk to our teams, spec things out with our agents, track work in tickets, but there's nothing really to connect it all together and there's no neat way to keep it all updated. Provenance is my attempt to solve that problem by building direct links between planning, requirements, implementation, and verification. I've purposely tried to keep the concepts as agnostic to everyone's own workflows as possible. I'll go over some of them here: Sources: a source is anything you can link to that might be of value to understanding your software. I.e. a diagram, meeting notes, legislation, a video even. Requirement: a requirement is a high level description of some need of your system. Resolution: A resolution is a decison made about some property of your system you believe needs to be documented. Generally resolutions are the child of one or more requirements and the parent many requirements. Rule: A rule is the child of a requirement used to describe some single property of a requirement. Rules are linkable directly to functions that implement them and tests that verify them. We're still in the early stages, but I'd be interested to see what you all think.

spektrafilm을 LUT 출력시켜서 적용시켜 본 사진. 기술적으로는 아직 갈 길이 멀지만, 오픈 라이센스 필름 룩 중에서는 잠재력이 커 보인다. LUT는 결국 색 포화시키기이고 색공간 처리가 중요한데 그 부분을 정확하게 지정할 수 있고 매끄럽게 동작한다. 필름의 광화학적 특성을 수학적으로 모사한다는 발상이 좋다. 개발자 분이 대단한 일을 해냈다. andreavolpato/spektrafilm

Failed AIGP once. Built myself a Duolingo style app, passed with 5 days of prep some context. I'm not a lawyer, not a full time privacy person either. I run product across a couple of startups and most of my time these days goes into agent identity.. how AI agents get identities, permissions, audit trails. so AIGP adds a lot of credibility to what say at work... and yeah, I failed it the first time... 😅 the material isn't even the problem. it's how IAPP writes the questions. every option sounds right, two are nearly identical. DPIA or FRIA. provider or deployer. that one law that creates no duties at all (NAIIA, I see you). for the retake I had 5 days. that's it. no way I was grinding 400 page PDFs again between work, so I built myself a duolingo style app instead. streaks, points, a confetti cannon, mindmaps you tap to lock facts in... whatever gets my brain to open it at 11pm tbh. made it public in case it helps someone: repo: mahfzal87/aigp-coach what's in it: - mock exams written to confuse you the way the real one does, with reasoning for every wrong answer - a "Law Lab".. mindmaps of every law on the BoK (EU AI Act, GDPR, Colorado, Texas, NYC LL144, Korea, NIST, ISO 42001), the trigger keywords for each, and side by side tables of the pairs everyone mixes up - streaks, XP, a timed drill. duolingo mechanics but the vocabulary is FRIA vs DPIA instead of French - progress tracking that shows your weakest competencies so you stop revising stuff you already know - updated for late 2026.. digital omnibus dates, the colorado repeal, SB 53. a lot of courses still teach the old timeline everything saves on your device. no account, no ads, nothing for sale. disclaimers: I built it so yes, this is self promotion of a free thing. all questions are original, nothing from the actual exam (NDA). and honestly 5 days only worked because I live in this field all day.. if you're starting from zero, give it longer. passed this week 🎉 happy to answer anything about the exam or the agent identity side

Marquee: a self-hosted, open-source Plex alternative with native apps, Live TV/DVR, and a music player built around how your music sounds I've been building **Marquee**, a self-hosted media server for movies, TV, anime, music and live TV. It runs in Docker on your own hardware, needs no cloud account, and is open source (AGPL-3.0). **GitHub, with screenshots:** pingpongshow/marquee **Apps:** web, Mac, iPhone, iPad, Apple TV, Android and Android TV. **Video** - Direct play when your device can, otherwise remux or hardware transcoding (NVIDIA, Intel Quick Sync or CPU) - Skip Intro and Skip Credits, chapters, seek-bar thumbnails, and watch together - Subtitles including styled ASS, OpenSubtitles search and Bazarr - Downloads for offline use on iPhone, iPad, Android and Mac, with an offline library when you can't reach your server - Library Health for duplicates, unmatched items and unplayable files, plus an optimizer for oversized files **Live TV and DVR** - Any M3U/XMLTV source or Dispatcharr - Series recordings, with commercials cut out by Comskip - On-demand films and shows **Music (the part I'm most excited about)** - **Soundprint** analyses your library on your own server, so radios and "sounds like this" follow how songs actually sound - **Muse:** describe a playlist in words - Endless Artist Radio and Track Radio, daily mixes, a DJ, and a thumbs down that keeps a song out of radios and generated playlists - Gapless playback, crossfade, loudness levelling, an equaliser and synced lyrics - Play to Sonos from the Mac app, or send audio to any Mac output on its own - **In Town:** optional concert listings for artists you like, near you, using your own free Ticketmaster key **For the household** - Profiles with optional PINs, and kids' profiles with rating and time limits - Per-person progress, ratings and recommendations - Shared playlists, and optional requests through Seerr - Web themes chosen by an admin, including holiday themes, or upload your own background **Coming from Plex?** You can import watch history, ratings, playlists, markers and accounts, and keep watch state in sync both ways. **Good to know before you try it** - It's a young project and I'm one person, so expect rough edges. Bug reports are very welcome. - The iPhone, iPad and Apple TV apps aren't in the App Store. They're unsigned, so you sideload them or build them in Xcode. The Android APK and the notarised Mac app install directly. - Away from home, Tailscale is the only supported way to connect. - You'll need a free TMDB key for movie and TV details. Music analysis is optional. - Unraid templates are available, and any NAS or server that runs Docker should work. I'd love feedback, especially on what would stop you switching from what you use now.

模块先查表。 @0xGky：Awesome Android Root——650+ 应用/模块合集，覆盖 Magisk/KernelSU/APatch/LSPosed，含机型教程与救砖、免 Root 替代专题。GitHub：awesome-android-root/awesome-android-root 判断：已 Root 或明确要折腾模块的人先当索引用；只想日常稳定用机的可跳过。动手前完整备份，银行/DRM App 兼容风险自负。

Pokémon Pokopia recompiled to run natively on your PC! No emulator needed, just the game files! Amazing! drakolordx7/pokopia-recomp

你的密钥、证书和敏感数据，值得一个真正开放的安全保险箱 OpenBao 是一款采用 MPL 2.0 许可的开源项目，支持安全管理、存储与分发密钥、证书等敏感信息 开源、透明、可控，为云原生时代的秘密管理提供新选择 openbao/openbao

Marquee: a self-hosted, open-source Plex alternative with native apps, Live TV/DVR, and a music player built around how your music sounds I've been building **Marquee**, a self-hosted media server for movies, TV, anime, music and live TV. It runs in Docker on your own hardware, needs no cloud account, and is open source (AGPL-3.0). **GitHub, with screenshots:** pingpongshow/marquee **Apps:** web, Mac, iPhone, iPad, Apple TV, Android and Android TV. **Video** - Direct play when your device can, otherwise remux or hardware transcoding (NVIDIA, Intel Quick Sync or CPU) - Skip Intro and Skip Credits, chapters, seek-bar thumbnails, and watch together - Subtitles including styled ASS, OpenSubtitles search and Bazarr - Downloads for offline use on iPhone, iPad, Android and Mac, with an offline library when you can't reach your server - Library Health for duplicates, unmatched items and unplayable files, plus an optimizer for oversized files **Live TV and DVR** - Any M3U/XMLTV source or Dispatcharr - Series recordings, with commercials cut out by Comskip - On-demand films and shows **Music (the part I'm most excited about)** - **Soundprint** analyses your library on your own server, so radios and "sounds like this" follow how songs actually sound - **Muse:** describe a playlist in words - Endless Artist Radio and Track Radio, daily mixes, a DJ, and a thumbs down that keeps a song out of radios and generated playlists - Gapless playback, crossfade, loudness levelling, an equaliser and synced lyrics - Play to Sonos from the Mac app, or send audio to any Mac output on its own - **In Town:** optional concert listings for artists you like, near you, using your own free Ticketmaster key **For the household** - Profiles with optional PINs, and kids' profiles with rating and time limits - Per-person progress, ratings and recommendations - Shared playlists, and optional requests through Seerr - Web themes chosen by an admin, including holiday themes, or upload your own background **Coming from Plex?** You can import watch history, ratings, playlists, markers and accounts, and keep watch state in sync both ways. **Good to know before you try it** - It's a young project and I'm one person, so expect rough edges. Bug reports are very welcome. - The iPhone, iPad and Apple TV apps aren't in the App Store. They're unsigned, so you sideload them or build them in Xcode. The Android APK and the notarised Mac app install directly. - Away from home, Tailscale is the only supported way to connect. - You'll need a free TMDB key for movie and TV details. Music analysis is optional. - Unraid templates are available, and any NAS or server that runs Docker should work. I'd love feedback, especially on what would stop you switching from what you use now.

If you do not know what to answer to non relevant cold messages on LinkedIn, X, email, you can just directly paste this sentence: ---- Hey, could you please star this GitHub project? albumentations-team/AlbumentationsX ---- It is not rude, and person will know that you are not interested in what they are offering.

Office-Word MCP 能让 AI 按作者提取 Word 批注，整理同事的修改意见，MIT 开源。 安装别照旧说明：README 写 Python 3.8 起，当前包要求3.11起；仓库也已归档。想接进长期办公流程，先考虑后续谁来维护。 GongRzhe/Office-Word-MCP-Server

3/ Today: 4 ok, 0 red, 4 not yet. Run the same script, get the same answer. github.com/ogasurfproject-jpg/horizon-shield/blob…

Got tired of git init, git add ., git commit -m "" - so tired I tried making my own version control system, and it actually worked! Heavily inspired by some pre-AI work on the same. Introducing Fendimo. It's built using python, and you can find it here: Andromedan-Organism/Fendimo-Version-Control-System-Inspired-by-Git

What should I break down next? • A profitable wallet's entries and exits • Wallets with matching activity • A Vanguard discovery walkthrough Kaiba's experimental code is public: kryptoxerk-dot/kaiba-ultimate-agent-public All things crypto, automated For engagement, build the next week around this thread: Pin it, then publish one wallet breakdown, one real Vanguard screen recording and one follow-up research finding.

Every forecast, scored against public outcomes: paratros.com Row PT-2026-022 · v1 · 5fbf640c committed: github.com/kpjmd/paratros-ledger/commit/a4ef98c1d… Report: x.com Informational. Provided as-is, without warranty. No reliance for any purpose.

What should I break down next? • A profitable wallet's entries and exits • Wallets with matching activity • A Vanguard discovery walkthrough Kaiba's experimental code is public: kryptoxerk-dot/kaiba-ultimate-agent-public All things crypto, automated For engagement, build the next week around this thread: Pin it, then publish one wallet breakdown, one real Vanguard screen recording and one follow-up research finding.

Fully offline neural TTS running on the ESP32-S3 itself (no server, no PSRAM). Useful for voice satellites? I built a small TTS model running directly on an ESP32-S3: no PSRAM, 122 KB of RAM, nothing in the cloud. Speech starts in about 0.3 s and generates faster than real time. You can make it calmer, louder, whispered, happier and more. It's not a Home Assistant integration yet. Before I build anything, I'd like to know whether on-device speech on a satellite would be useful to you (offline announcements, timers, alerts when the server is down), and what quality bar it would need to meet. Limits: 2 voices, English only, 16 kHz. Free for personal, non-commercial use. Video above: Details: huggingface.co CLI : Code: WhissleAI/steer-tts (esp32/ folder)

What should I break down next? • A profitable wallet's entries and exits • Wallets with matching activity • A Vanguard discovery walkthrough Kaiba's experimental code is public: kryptoxerk-dot/kaiba-ultimate-agent-public All things crypto, automated For engagement, build the next week around this thread: Pin it, then publish one wallet breakdown, one real Vanguard screen recording and one follow-up research finding. Run a poll: “What hurts your copy trades most?” Late signals / fees and slippage / missing exits / correlated wallets. Write five useful replies daily on relevant discussions. Ask specific questions about the profit window, arrival delay or exit evidence. Track substantive replies, bookmarks and profile visits over two weeks to see which topics earn continued interest.

IDA-Ghidra-bridge Having issues reversing? Use this is very effective!

GB Miner v1.0.0 is out ($TAPE integrated) ① 📼 One miner for two coins: Parano1d (parano1d:native) and Retro Boomer ($TAPE). It is the successor of GB Parano1d Miner 3.0.6. github.com/gustlborg/gb-miner/releases/tag/v1.0.0 Highlights - Pick the coin by start script: start-noid.sh / start-tape.sh (Linux), start-noid.bat / start-tape.bat (Windows). - parano1d:native pools: the same as in 3.0.6, without ariabrain (AriaPool paused NOID mining on 18 October). - $TAPE pools by name: gbpool (default), gbpool-ssl, gbpool-farm, gbpool-solo, suprnova, suprnova-ssl, suprnova-solo, rplant (rplant-na, rplant-asia), pqpool, pqpool-ssl, or any pool address. - 0 % pool fee on the GB Pool of both coins with this miner, on all four ports. - Dashboard in the coin's colour: GB Miner · parano1d:native (green) or GB Miner · $TAPE (orange). - Same kernels and hashrate as GB Parano1d Miner 3.0.6, for both coins and still the most fastest and most efficient. Everything else Mining modes (balanced default, performance, efficiency), autotune, power limit and clock settings work as in 3.0.6. Linux · Windows · HiveOS GTX 16, RTX 20/30/40/50, Volta, GA100, Hopper · NVIDIA driver 580 or newer, no CUDA Toolkit needed. Pascal (GTX 10-series) and older cards are not supported. 0 % pool fee on the GB Pool of both coins with this miner, on all four ports. pool.noidexplorer.org tape.pool.army

🤖 Meet DeepSeek-Bot! A DeepSeek harness with agent memory and a message gateway, designed to run as an active AI agent like GrokBot or Hermes. Worth exploring for anyone building persistent AI agent workflows. ⚡ 🔗 rlaope/deepseek-bot t.co

Online .DV to .mp4 convert tool Free DV to MP4 converter. Runs in your browser. Your files never get uploaded. Try it: dv-to-mp4-converter.pages.dev Code: swooshboots/lovable-dv-to-mp4-converter I wanted a way to convert footage during a session to be able to clip out different people’s tricks and send them the .mp4 all from mobile during the session. I couldn’t find another way to do this on mobile so I vibe coded this little app. Works great, please share any feedback any anything else you’d like to see added. Thanks for looking!

我再也不会错过邮件通知了 sb-child/print-my-mail sb-child/dz-print

Pixel-guessing computer use is getting old. 👀 Vercel Labs open-sourced agent-desktop: a Rust CLI that lets AI agents control native apps on macOS, Windows & Linux through OS accessibility trees. Stable element refs, actions safe to retry. #AIAgents vercel-labs/agent-desktop

英文论文想整篇译，在线工具按页收费，复制到网页翻译表格排版全乱、公式丢失。 docutranslate 直接处理 PDF/Word/Excel/EPUB/字幕，保留段落与表格结构，导出仍是原格式。 xunbu/docutranslate

Best app (paid or not) for motion sickness? So apparently since I turned 30 I lost the super power of using my phone, read or play videogames while in a car to the point where I get sick just by going to my phone and send a whatsapp text, it's ridiculous. I currently have an android 12 Redmi Note 9 pro and an android 13 AYN Thor so I don't have the latest motion sickness assist included in the latest android version and I need something that works in both screen orientations. I tried MotionEase and frankly this app is excellent; it supports vertical and horizontal screen orientations, is not very invasive and for the first time in years I was able to watch a tv show in my way to the work, it was amazing. Sadly it has a ton, and I really mean a lot of ads and the premium version costs like 10 dollars a week, it's very expensive. Then I installed Motion Cues which is not only free but also is open source which is great!.... but only works if the screen is in vertical orientation so I can't use this on my AYN thor or my phone while watching movies as the dots go the wrong way. I was searching for another app and then I found Perusa, which not only seem to be actively updating and have a lot of options but also it seems like they were looking for testers here. I can't test it at this time as the app only works when on a vehicle so I'd have to test it tomorrow, it looks very complete and the price is a little more than 2 dollars on my local currency so it's pretty accesible.. but I've never hear about those devs or their app before. is there another app that I should test?, is Perusa the app I'm looking for?, should I just avoid using my phone on my way to the work? I appreciate any suggestions.

I built an alert-only AI log-monitoring agent (Python + Gemini + n8n) and tested it on real SSH logs. Here are the honest results, including the failures Hi all, I'm an automation engineer and I built a small project called AI Cyber Patrol Agent. What it does: - Reads login logs, validates and sanitises them - Scores suspicious activity with transparent rules (repeated failures, success after failures, new IP/country, odd hours, many failures from one IP) - Groups events into incidents and asks Gemini for a short summary (retry + rule-based fallback when the API returns 503) - Sends an alert-only email through an n8n webhook, with duplicate prevention - Streamlit dashboard to review everything Design choices I cared about: the risk score is computed by code, never by the LLM; it never auto-blocks anything; inputs are hardened against CSV, header and prompt injection. Testing (the part I'd like feedback on): - 25 development cases from the public Loghub OpenSSH dataset: 25/25 pass. But I fixed one failing case (slow attack) by adding a per-IP 24h rule, so that number is partly circular. - 10 hidden cases from a different slice of the log, run once after freezing the code: 6/10. It caught 4/5 attacks, but gave 3 false alarms because the per-user rule counts failures from other IPs during a flood on root/admin. One slow attack spread over 3 days was missed. I documented the causes and a roadmap (count failures per user+IP pair, longer window for slow attacks) in the README. Live demo (synthetic data): cyber-petrolling1319.streamlit.app Code: ztsprofessionalaiid7-alt/cyber-petrolling I'd love feedback on the detection logic and on how you'd evaluate something like this more fairly. Thanks!

Hyrule Warriors: Age of Imprisonment, Switch 2 Emu I added Hyrule Warriors: Age of Imprisonment to steam. Running steam in big picture mode. Game runs very well. No real issues so far. I skipped through the cutscenes so I don't spoil for anyone, and I've already watched them. Resolution 2X (1440/2160) with around 60 FPS Cover and art from steamgriddb, [steamgriddb.com Emulator, [github.com/Djihads80/Eglantine](https://github.co…

@bufferfrom @chasmmmmmmmmmmm it's ok brother I'm using rust. and in fact one game I made is like 700kb web assembly check it out it's like Roblox slop, so the graphics are not good. I've got a real project but it's a Rust(the game) clone in Rust lol AnthonE/Fireplace t.co

Fully offline neural TTS running on the ESP32-S3 itself (no server, no PSRAM). Useful for voice satellites? I built a small TTS model running directly on an ESP32-S3: no PSRAM, 122 KB of RAM, nothing in the cloud. Speech starts in about 0.3 s and generates faster than real time. You can make it calmer, louder, whispered, happier and more. It's not a Home Assistant integration yet. Before I build anything, I'd like to know whether on-device speech on a satellite would be useful to you (offline announcements, timers, alerts when the server is down), and what quality bar it would need to meet. Limits: 2 voices, English only, 16 kHz. Free for personal, non-commercial use. Video above. Details: huggingface.co CLI : github.com/WhissleAI/steer-tts/tree/main/esp32

EARTHBOUND PC PORT COOP VER. 4.5 IS LIVE! Hey everyone, how's it going? EarthBound PC PORT COOP UPDATE 4.5 IS NOW LIVE! 4 PLAYERS: That's right! We can now play with up to 4 players!! You can play as Ness/Mess, Paula, Jeff, or Poo! Players can join at any point in the game and will automatically adapt to the story's progression, items, and levels, keeping things as balanced as possible! Fight Together: A new optional battle system! When players are close to each other as an enemy encounter begins, their friends will automatically join the fight! Otherwise, any player who wants to participate has 5 seconds to get there! LAN MULTIPLAYER: THAT'S RIGHT, GUYS! IT'S NOW POSSIBLE TO PLAY CO-OP WITH DIFFERENT DEVICES IN THE SAME SESSION THROUGH WI-FI LAN!! It's also possible to play online across different networks, but this requires some additional setup involving port forwarding, so you'll need to know what you're doing! You can also try using Hamachi or a shared VPN! (Performance in this mode may vary considerably. I've tried my best to optimize it and keep things running smoothly, but performance also depends on the network adapter of the device hosting the session, whether it's a phone or a computer.) Cross-platform play between different devices is also possible! There are also plenty of tweaks here and there, implementation revisions, and a few extra little things I've added. Feel free to explore and create your own way of playing, guys!! INCLUDING PLAYING WITH FRIENDS! Link to GitHub: https://github.com/edzito69/Earthbound-PC-Port-COOP/releases/tag/EarthboundCOOPv4.5 THANKS!!and DOINGG!!

오늘 오전에 Copilot, Cursor, GPT-4o 막 섞어 썼음. 어느 놈이 뭘 해줬는지 기억이 안 나서 막막했음. 근데 '/AI업무보고' 치니까 싹 다 한 보고서에 정리됨. 신기함. jyoung9154/AIAgent_Report #AI개발 #업무보고 #광고

New project i wrote: Nimformers: a Transformer inference backend written in Nim, running on Metal/Vulkan I'm the author of Nimformers, a personal project of mine. I recently ran some benchmarks on two devices and wanted to share the numbers and the issues I ran into, and get some feedback on the design. Some design points: - Metal and Vulkan backends — on Mac it uses Metal directly - int8/int4/fp8/fp4 quantization support, plus NIMQ state dicts Issues I ran into (being honest): - Vulkan experience isn't very stable — performance on Iris Plus is lower than Metal, still trying to figure out why - APIs are still changing, some interfaces aren't stable yet - Small ecosystem, docs can't compete with PyTorch Repo: bobbyshop-vui/nimformers-framework

@DeadlyLaw csns1/bypass-paywalls-chrome-clean-magnolia1234 Pretty good paywall bypass extension for chrome

Ghidra / IDA Reverse Servers For those having issues reversing / rebuilding games with claude: oscarsantanaarias/IDA-Ghidra-bridge This is a direct bridge to ghidra/ida that automatically gives any AI full access to binary code in IDA/GHIDRA

桌面 Agent 报 STALE_REF 先别再点一次 Vercel 的 agent-desktop 把按钮编号绑在那次快照上 界面变了 就重新 snapshot 拿新编号 这个小规矩我挺认 尤其是删文件 发消息这种动作 点错一次已经够麻烦了 在接桌面自动化的 可以把这条写进失败处理 vercel-labs/agent-desktop

Part 03: Vector Addition in CUDA is out. I've covered how to perform vector addition on the GPU using CUDA threads, including device memory allocation, host to device transfers, global thread indexing, kernel launches and bounds checking. Repo: navaneethkrishnansuresh/CUDA-and-Kernels t.co

用 Cloudflare 免费额度搭域名邮箱后，我顺手给它写了个原生 iOS 客户端 ---  大家好，分享一个自己做了挺久的东西：CF Mail，一个原生 SwiftUI 写的 iPhone / iPad 邮件客户端。 起因 之前用 [maillab/cloud-mail]( maillab/cloud-mail) 在 Cloudflare 上搭了域名邮箱。Worker + D1 + Email Routing ，基本零成本，一个域名可以开无限个地址，注册各种服务时一个站一个邮箱，很爽。 问题是手机上只能开网页版，没有推送，也没法和 Gmail 、QQ 这些常用邮箱放在一起看。系统自带的「邮件」 App 又连不上它（它没有 IMAP ）。所以就自己写了一个客户端，顺带把服务端也改了一版。 能干什么 两种账户放在一个 App 里： - 普通邮箱：走 IMAP/SMTP 直连，已经内置了 Gmail 、iCloud 、Outlook 、Yahoo 、Fastmail 、QQ/Foxmail 、163/126/yeah 等服务商的预设。QQ 和网易会提示你用「授权码」，不用自己去查服务器地址。「收件箱、已发送、草稿箱、已删除」这些中文文件夹名会自动识别。 - 自建 CloudMail：填上你的 Worker 地址就能用。如果你有管理员权限，App 里还能直接管用户、角色权限、邀请码、全部邮件和系统设置，不用再开电脑。  其他功能： - 统一收件箱，以及服务端搜索：可以按主题、正文、发件人、收件人来搜 - 分开发送：群发时每个收件人各自收到一封，互相看不到对方，发通知、发邀请很好用 - 富文本编辑、附件、草稿、联系人，回复和转发 - iPad 分栏布局，大屏上列表和正文可以同时看 - 新邮件推送（ CloudMail 账户，可选） 隐私这块我比较在意 - 没有广告，没有第三方统计或跟踪 SDK ，也不需要注册 CF Mail 账号 - 账号凭据只存在 iOS 钥匙串里，可以用 Face ID / Touch ID 锁住 App - 远程图片可以设成默认拦截、每封邮件询问或始终加载，防追踪像素 - IMAP 邮箱是手机直连邮件服务器，中间不经过我的任何服务器 - 推送走一个独立的网关。自建的 Worker 只发「有新邮件」这类脱敏事件，不直连 APNs ，也不保存设备 Token - 诊断日志只存在本地，并且做了脱敏，只有你手动导出时才会离开手机 服务端也开源了 我在上游基础上维护了一个分支：[CF Mail cloud-mail]( readori/cloud-mail) 主要改动有：安全加固（初始化保护、Webhook 签名、密钥加密存储、会话加固）、CC/BCC 、更多邮件 API 、通知钩子、数据库迁移保护，以及一套部署前的契约测试。 用一个域名加一个 Cloudflare 账号就能跑起来，个人用基本在免费额度内。 收费说明 先说清楚，App 有内购： - 免费版：1 个 CloudMail 服务器（每个服务器 1 个身份），加最多 10 个 IMAP/SMTP 邮箱，所有功能都能用 - Pro：不限服务器和账户数量，有订阅版和买断版，价格以 App Store 显示为准 [可在这里写具体价格] 个人用的话，免费版应该就够了。 下载 - App Store： [ apps.apple.com ] - 目前没上架国区，只能用外区 ID 下载 - 要求 iOS 16 及以上 还在持续迭代，欢迎大家提 bug 和需求，帖子里回复或者去 GitHub 提 issue 都可以。在 V2EX 回复里留言想要的功能，我会优先考虑 🙏

I made a torrent client that searches multiple sites without leaving the terminal hey guys, i made tork, a torrent client where you can search across multiple sites and download stuff without leaving your terminal just type what youre looking for, pick a result and hit enter. no jumping between sites or copying magnet links around. it puts everything in one place and merges duplicate results too got 200+ stars on github now which is kinda crazy lol. still working on it but pretty happy with how its coming along its free and open source, try it out and lmk what you think :) melqtx/tork

🚀 Taikai 2.8.0 is here! 👉 Pluggable exports: SARIF & Markdown export of architecture rules for AI agents 👉 Metrics: inheritance, size & coupling 👉 File content checks More insights. Stronger architecture. 🔥 github.com/enofex/taikai/releases/tag/v2.8.0

Stack (in no particular order): Zephyr RTOS, C, Rust, Go, Java, Kotlin, Spring Boot, Node.js, TypeScript, NestJS, Python, FreeSWITCH, Kamailio, AWS (CDK), Cloudflare, Terraform, LLM/agentic/MCP & more CV: parisera.ro Linkedin: linkedin.com Email: victor@parisera.ro Github: github.com/3Nigma Web: parisera.ro Location: EU-registered company, open to other time zones (ie. KST) Remote: yes Type: Part/Full/Contract, B2B Currently looking for: embedded/firmware work on Zephyr RTOS, plus backend, infrastructure and real-time/VoIP roles. I've been programming professionally since I was 14 and leading teams for over a decade. Most recently I've been leading an engineering team and architecting an AI voice initiative on top of FreeSWITCH/Kamailio call infrastructure (SIP/VoIP, microservices migration, containerized NestJS). Before that, I co-founded a Web3 company where I led the firmware for a secure hardware wallet (Zephyr RTOS, C), built an open-source library for Hashgraph, and designed the backend and infrastructure for an IDO/INO platform on AWS CDK. Earlier: technical lead roles in Java/Spring microservices and Alexa/Google Home voice assistants, plus on-prem infra, CI/CD and Android. 20+ years, from silicon to cloud, server and edge, startups to enterprise. I'm a first-principles thinker who likes to keep my mind sharp and my hands busy building. Let's connect. I'd love to see what you're building and how

Eject yourself from work with a little help from Codex