Bulk classification and decision inference on one GPU: the content is read once, every question about it is scored in the same pass. OpenAI-compatible, Apache 2.0
All the ideas came from the boot sector scene where people had been writing chess and BASIC games in 512 bytes, even a C compiler. I wondered what one can do using a neural network instead of a program and generating boot logs for non-existent hardware was something fun to try. Entire inference engine is 510 bytes of MBR, weights, vocab, and strings reside in the sectors following MBR. Small char-level neural network was trained on dmesg generated by machine, so machine produces Linux boot messages infinitely. train.py performs precisely the same integer arithmetic operations as the asm code does and thus produces an output identical character-by-character to the one produced by QEMU. The difference analysis revealed my favorite bug. The characters at the beginning matched those generated by the simulator, but then the output diverged. I retrieved the hidden layer via QEMU’s debug port and compared the neurons' values with those of the simulator, one neuron at a time. Only the value of the first neuron was correct. The activation involves idiv, and idiv stores its remainder in edx. The pointer to the biases table was stored in edx, which resulted in the first neuron being biased correctly and the following ones using garbage. Now the bias' address is determined from the loop counter, and no pointers have to survive the division. Another bug I found simply by looking at the code before I did my first run-through: the BIOS print routine zeros out bx, but bx contains the predicted character number. This explains the push and pop of bx surrounding the print. sectorllm is the most similar to my project, as it runs an actual 5-layer llama2 with attention mechanism in just 1.2K of assembly code. It is the trade-off in the opposite direction, as my engine needs to fit into one sector and therefore my network has just one hidden layer of int8 operations for dreaming the dmesg.
Instead of trying to run DOOM on things, I have instead decided to use DOOM (many, many, many doom games actually) to create a transformer. For a lot of the math required I essentially had to use DOOM game mechanics and voodoo dolls (DOOM is turing complete btw) and exchange them with the actual operations, essentially creating an alternative to NumPy (a 65000x slower one). Thus, I am proud to present DOOMFORMER. Here are a list of things DOOMFORMER can do : It can classify MNIST digits, even if you write these digits on the map! It can talk(kinda) You can use it to make decisions, so, it can play DOOM. Yes! You can play DOOM with thousands of DOOM games now! The marine's mind can be controlled by DOOM, playing on a map generated by DOOMFORMER. Truly, the possibilities are endless, and even larger-scale training of DOOMFORMER could be done on a measly T4. If anyone would like to help me scale this up and make crazier things, I'd love to work together.
Project URL Category Python Project Title Solaris Dev Shop: a local Python router that picks one installed skill for a task Project Solaris Dev Shop is a public MIT skill collection and local Python router. It picks one installed skill for a task and runs that skill command. It does not start an agent or call an API. Highlights It routes one local task to one installed skill file. It does not call an API, start an agent, or host a model. Example Code markdown or Videos
Project URL Category Python Project Title Anki Miner: automatic language vocabulary cards Project Desktop app for automated Anki language vocabulary card creation from video, audio, manga and books, in 32 languages. A video gives cards with the sentence, audio, a screenshot and a dictionary definition. Highlights Creates cards for a whole episode, season, book or manga at once. 32 mining languages, including Chinese, Japanese, Korean and English. The UI supports Simplified Chinese. Example Code bash pipx install anki-miner ankiminergui or Videos
用 Python 自动操作 Windows 桌面应用。这是一个 Windows 图形界面自动化 Python 库,能够模拟鼠标和键盘操作窗口、对话框和控件,还可以读取界面上的文本数据。它支持 Win32 API 和 MS UI Automation 两种后端,可以通过窗口标题、控件名称等文本属性定位元素,几行代码就能完成打开菜单、点击按钮、输入文字等操作,适用于桌面软件的自动化测试和重复操作。
Project URL Category Other Project Title Resume Claim Guard: an agent skill that stops hallucinated resume claims Project An open-source agent skill (SKILL.md) for Claude Code and AI coding agents that enforces a strict 4-step resume workflow: anchor every claim to a verified fact base, audit metrics against exaggeration, and generate bilingual role-aligned bullets. MIT licensed. Homepage: Highlights Prevents hallucinated resume claims by binding each bullet to a verified fact base, and shows changes as yellow-highlight diffs so you can review every edit. Bilingual output, usable for interview prep too. Example Code markdown or Videos
I’ve spent the last year building context for AI analysts in the mid-market. I’ve open sourced a library from my learnings, along the same principles of the Matt Pocock skills. Principles: * Claude / Codex / Cursor etc are great data agents - you just need context to enable them to answer business questions correctly. * Much of the important context lives in analysts heads' * Context must be maintainable (short, human readable - markdown and yaml) * Added context should be load bearing, not repeat of information that is available in another source of truth - and evaluations prove the value of the context * Companies should own their own context, and must be shareable with any model * Context should be version controlled (and github works as a context store). What the library does: * Performs an interview with your analysts (context-interview skill), one domain at a time, similar to what I have done as a consultant. The interview is informed by your query logs. * The AI develops evaluation questions at the same time as building the context - and at the end of each domain interview asks you to run an evaluation with and without the context. The evaluation will actually check the results against your dashboards if you allow it. We have shown increases in accuracy from 87% to 95% by adding this context in our imaginary open-source company, Shorelane. * Provides a set of skills to use the context that follow best practices for self service analytics: * Write a plain english plan for the business user to approve * Execute the plan * Verify that the execution followed the plan, and provide a simple uncertainty score. * Enable a adversarial review of the answer if their are doubts. What you get: Context written in markdown and yaml after you finish the interview. We use a single github repo, divided into domains. The interview can be shared and restarted by checking in/out of github. A set of skills to share with your business team Optional: We offer a fast MCP server to serve the Github-based context to your team.
Contenido:LingTube es una extensión para traducir videos de YouTube, convertir voz a texto y generar doblaje con IA. Al abrir un video y activarla con un clic, traduce los subtítulos o la voz de videos en otros idiomas (incluso los que no tienen subtítulos), y los lee en voz alta con una voz natural. Así puedes entender videos de YouTube sin estar pendiente de los subtítulos.Empieza al instante a usar el doblaje con IA y la traducción de subtítulos en YouTube, Coursera, Udemy, Bilibili y TikTok: solo reproduce cualquier video y comienza de inmediato.Ideal para aprender idiomas, cursos online, tutoriales técnicos, noticias, pódcast, anime, fragmentos de películas y ver contenido mientras viajas o haces otras cosas.Funciones principales Traducción de videos de YouTube: traduce subtítulos, voz y contenido de videos sin subtítulos. Subtítulos bilingües: muestra traducción + original al mismo tiempo, o solo la traducción o solo el original. Doblaje con IA / lectura de subtítulos: usa Edge TTS o la voz del navegador para leer la traducción en voz alta. Videos sin subtítulos: cuando el video no tiene subtítulos, transcribe automáticamente la voz, la traduce y la lee. Traducción con contexto: compatible con Google Translate e interfaces OpenAI para entender mejor frases largas y contenido técnico o especializado. Aprendizaje con IA: genera vocabulario, gramática, ejercicios, evaluaciones y material de repaso. Control de audio: puedes escuchar el audio original y la voz de la traducción al mismo tiempo, y ajustar el volumen de cada uno por separado. Activación rápida: un clic o el atajo de teclado Alt+Shift+V. Resumen de videos con IA: una vez iniciada, puede conversar con los subtítulos traducidos completos para resumir el video, extraer las ideas principales, explicar términos técnicos y localizar momentos clave. Perfecto para videos largos, cursos, tutoriales técnicos, noticias y pódcast. Privacidad Solo funciona en páginas de YouTube, Coursera, Udemy, Bilibili.tv y TikTok. La configuración se guarda localmente en el navegador. Los subtítulos, traducciones y solicitudes de voz solo se envían a los servicios que tú habilites. Extensión / plugin de ChromeChrome Web Store: oficial: 亮点 Contenido:LingTube es una extensión para traducir videos de YouTube, convertir voz a texto y generar doblaje con IA. Al abrir un video y activarla con un clic, traduce los subtítulos o la voz de videos en otros idiomas
Security scanner for AI agent skills. Detect vulnerabilities, malicious patterns, security risks, prompt injection, data exfiltration, and supply-chain risks in Claude Code, Codex, and MCP skills before you install them.