Convierte claude code y cualquier herramienta de IA en un equipo completo de investigación académica
Con esta guía gratuita puedes montar:
Deep Research → 13 agentes + 8 modos
Academic Paper → 12 agentes + 11 modos
Paper Reviewer → 7 agentes
Academic Pipeline → Orquestador de 10 etapas con verificación de citas
Diseñado para que la IA haga el trabajo pesado (búsqueda, redacción, revisión y control de calidad) mientras tú mantienes el control humano.
Gratis para uso no comercial
Enfoque fuerte en integridad, citas verificadas y estilo personalizado
Repo en los comentarios, esta increible
Introducing autoresearch for GitHub repos
Change 'Github' to 'ARGithub' in any repo URL
Research artifacts extend beyond papers. Autoresearch is especially useful for experimenting on existing codebases that move fast and outpace their own publications.
With one URL change you can now deploy an agent to orient itself on the codebase, resolve setup issues, and iterate on experiments.
A lawyer in Manhattan gets a 500-page contract. Every clause needs to be searchable. By hand: one week.
An accountant in Chicago gets 200 scanned invoices. Every number needs to land in a spreadsheet. By hand: four days.
A researcher at Stanford has 50 academic papers. Tables, formulas, charts locked inside PDFs. By hand: two weeks.
Every one of them is losing days of their life to copy-paste.
Now meet MinerU.
A free and open source tool that reads any PDF, Word doc, PowerPoint, Excel sheet, or scanned image. It pulls out the text in reading order. Tables become clean HTML. Equations become LaTeX. Handwriting handled. 109 languages.
You give it a 200-page PDF. You get clean Markdown back in 90 seconds.
What makes it different from every other PDF tool:
- Multi-column layouts. It reads top to bottom within each column. Not left to right across the page. Like a human reads.
- Scanned documents. OCR built in. Point it at a photo of a printed page from 1995. Get clean text back.
- Math formulas. LaTeX-quality recognition. Every equation renders correctly.
- Tables. Merged cells, multi-row headers, tables that span three pages. All preserved.
- Ten-thousand-page documents. Sliding window processing. No manual splitting.
- Batch mode. Point it at a folder of 500 documents. Walk away.
Three ways to use it:
- CLI. One command per document.
- Python SDK. Five lines of code.
- Web app at https://t.co/AIC2NNey41. Upload, click, download. No install.
Plugs into Claude Desktop, Cursor, Windsurf, LangChain, LlamaIndex, RAGFlow, Dify, and FastGPT. Feed extracted documents straight to your AI agent.
The story:
The OpenDataLab team at Shanghai AI Laboratory needed to extract clean text from millions of scientific documents to train a language model. Existing tools failed. They built their own. Then they open sourced it.
68,551 stars. MinerU Open Source License, built on Apache 2.0. Free for personal and commercial use. Three technical reports on arXiv.
Adobe Acrobat Pro charges $239.88 a year. It still loses your tables.
ABBYY FineReader Corporate charges $165 a year. It still cannot do equations.
Mistral OCR charges $2 per 1,000 pages. Your bill never stops.
MinerU costs $0. Runs on your laptop. Your documents never leave your machine.
Here is the wild part.
The lawyer got her contract back in 4 minutes. Every clause searchable.
The accountant fed 200 invoices in. Every number landed in a spreadsheet in 12 minutes.
The researcher fed his 50 papers in. He wrote his literature review on a Sunday afternoon.
The document your company has been processing by hand for years takes MinerU minutes.
Your documents become text. Your text becomes data. Your data becomes answers.
The week you used to lose to paperwork is back in your hands.
These 10 GitHub repos can scrape almost the entire internet.
And companies charge $2,000+/month for access to the same capabilities.
Bookmark this thread 👇
1. Firecrawl
https://t.co/NkSY0uKdmk
Point it at any website and it crawls every page, renders JavaScript, and returns clean structured data ready for AI.
130K+ stars and one of the fastest-growing open-source projects ever.
The crawling engine quietly powering a huge number of AI startups.
2. Crawl4AI
https://t.co/IXCu9v2dna
The #1 open-source crawler for AI workflows.
Converts messy websites into clean LLM-ready markdown with no API keys, no accounts, and no per-page fees.
Built after a developer got tired of paying for expensive scraping APIs.
3. Browser Use
https://t.co/HvSZcFydHr
An AI agent that uses websites like a human.
It clicks, scrolls, logs in, fills forms, and extracts data from pages traditional crawlers can't reach.
Created by ETH Zurich researchers and exploded to 95K+ stars in record time.
4. Crawlee
https://t.co/Bj9CYuWrva
A professional-grade crawling framework with rotating proxies, retries, browser automation, queue management, and anti-blocking systems built in.
The same infrastructure many scraping companies sell as a service.
5. Scrapy
https://t.co/YtBORRdDbA
The industrial powerhouse of web scraping.
Battle-tested for more than a decade and capable of crawling millions of pages efficiently.
Still one of the most trusted tools in data engineering.
6. MarkItDown
https://t.co/nBksZhBf4U
Microsoft's open-source tool that converts websites, PDFs, Office documents, images, and other files into clean markdown for AI pipelines.
Becoming a core building block of modern RAG systems.
7. Scrapling
https://t.co/57rhAI5NHI
A stealth scraping toolkit designed to survive website changes and evade common anti-bot systems.
Adapts when page structures shift, reducing maintenance headaches.
8. scrcpy
https://t.co/eGAEXiPWuQ
Control Android devices directly from your computer.
Extract data and automate mobile apps that don't even have websites.
The gateway to scraping mobile-only ecosystems.
9. AutoScraper
https://t.co/EcJIM2S3Km
Show it one example and it figures out the pattern automatically.
No complex selectors.
No endless maintenance.
Just tell it what data you want and let it learn the rest.
10. curl-impersonate
https://t.co/KAGgJyrdSx
Makes your requests look exactly like real Chrome, Safari, or Firefox traffic.
One of the most important low-level techniques behind many premium scraping APIs.
Most people think the internet is locked behind APIs.
The reality?
The source code is already on GitHub.
And it's free.
A senior Google engineer just dropped a 19-page PDF on "Loop Engineering" for LLM and agentic systems.
Act → Observe → Learn → Repeat
• Act: the LLM proposes a code transformation (tile this loop, parallelize that one).
• Observe: a compiler runs it and reports back - is it valid? faster? slower? by how much?
• Learn: the LLM reads that feedback and adjusts its next move.
• Repeat until it stops finding improvements.
The agent gets smarter purely from grounded feedback inside its own context window.
This 19-page PDF totally changed the way I’m building agentic systems today.
Read it now, then explore the article below.
OMG.. this is wild...
I FOUND A FREE TOOL THAT TURNS ANY REAL CITY ON EARTH INTO A FULL 3D MAP.
Every building and every road, delivered at scale.
You can export it as a file and do whatever you want with it.
It is called map3d. 100% FREE. Open source.
Built on real OpenStreetMap data, enter any city name to generate a 3D version of that city in seconds.
You can then export the whole thing as a GLB file and use it in games, videos, presentations, digital twins, or just to stare at your city from above.
↳ 3D buildings with real heights
↳ Roads and street layouts
↳ Export as GLB file
↳ Works for any city in the world
↳ Completely free
Before this you needed expensive GIS software and a geography degree to do this. Now you need a browser.
Visting Fellow Programme in Finland - FULLY FUNDED
The University of Jyväskylä (JYU) is accepting applications for the JYU Visiting Fellow Programme 2027. The programme offers researchers holding a doctoral degree and residing outside Finland the opportunity to conduct an in-person research visit at JYU for either one month (30 days) or two months during 2027.
📚 Who can apply?
Researchers with a doctoral degree based outside Finland.
Host University: University of Jyväskylä (JYU)
📅 Visit period:
• Earliest start: January 2027
• Latest end date: 31 December 2027
🎓 What Does the JYU Visiting Fellow Programme Offer?
💶 Grant of €4,000 per month (paid directly to the fellow)
📅 Funding for a 1- or 2-month research visit in 2027
✈️ Flexible use of the grant to cover travel, accommodation, and living expenses
🔬 Access to JYU's vibrant research community, comprising more than 1,600 researchers
🌍 Opportunities for international collaboration and interdisciplinary research
🎤 Possibility to present your research to the university community
🏛️ Full integration into the University of Jyväskylä's international academic environment
Grant Decisions
The decisions of the Visiting Fellow Programme will be announced by October 9, 2026. Applicants will receive an email notification after the funding decision has been made.
📅 Application Deadline: 14 August 2026
Credit: University of Jyväskylä (JYU)
One of the better agentic AI courses I've seen
Nearly 10 hours of great content. Covers LangChain, LangGraph, RAG, deepagents, guardrails, and more
Any other good Lang* resources out there for folks who are interested in learning?
https://t.co/OXNPMeGiyd
10 GitHub tools that are ridiculously hard to believe are free:
1. yt-dlp
Download videos and audio from 1,800+ websites in the highest quality possible. Supports subtitles, metadata extraction, playlists, and more. It keeps adapting to platform changes and has earned 172K+ stars.
https://t.co/qw4B5ZTd8X
2. Stirling PDF
An all-in-one PDF toolkit that replaces most Adobe Acrobat features. Merge, split, sign, OCR, compress, redact, and convert files completely locally. 81K+ stars.
https://t.co/FSVlrMKwvV
3. Homepage
A beautiful self-hosted dashboard that puts all your apps, services, and servers in one place. Perfect for homelabs and power users.
https://t.co/jHrxQLKUiy
4. LocalSend
The AirDrop alternative that works everywhere. Send files between Windows, macOS, Linux, Android, and iPhone without accounts, cloud storage, or size limits.
https://t.co/FOW5RXjtEm
5. AppFlowy
An open-source Notion alternative with documents, databases, and offline-first storage. Your data stays under your control. 70K+ stars.
https://t.co/VTgGJhUnMK
6. Immich
A self-hosted Google Photos replacement with automatic backups, facial recognition, AI search, and shared albums.
https://t.co/mQ40T1Ua84
7. Reactive Resume
Create, manage, and publish unlimited resumes for free. No hidden paywalls when it's time to export.
https://t.co/KkZYdfpbRD
8. Whisper
OpenAI's open-source speech-to-text model supporting 99 languages. The same technology many transcription services charge money for.
https://t.co/jBCrYA7dXm
9. n8n
A powerful workflow automation platform that lets you connect apps and automate tasks without paying per execution. 190K+ stars.
https://t.co/Es1RfxwEqt
10. Firecrawl
Give it a URL and it turns an entire website into structured, AI-ready data. A favorite tool for AI builders and developers.
https://t.co/LhUHaLR0Il
Open source is quietly replacing software people pay thousands of dollars for every year.
Most people haven't discovered these tools yet. 🤯
Have you thought about the future of Agentic AI?
We have too, so AWS and @OpenAI leaders recently came together to discuss where agents are working today, and what your organization needs next to lead with AI.
Check out the full stream below.
These 5 powerful storytelling frameworks will help you communicate more effectively and advance your career. Perfect for interviews, presentations, LinkedIn posts, and personal branding.
SOMEONE JUST DESTROYED THE REAL ESTATE SECTOR AND NO ONE IS TALKING ABOUT IT
Someone scanned an entire house with their phone. They uploaded it.
Now anyone on earth can walk through it from their browser. No app. No VR. No agent. No appointment.
Click → you're inside. Every room. Every angle. Every shadow. Photorealistic.
The numbers don't make sense:
- Agent commission on a $500k apartment: $15,000
- Scan cost: ~$200
- Time to "visit" 50 apartments: one afternoon
- File size: smaller than a TikTok
The science is also insane:
It's called 3D Gaussian Splatting, instead of polygons (like games render), it uses millions of tiny glowing "clouds" of color and depth.
AI reconstructs reality from your photos. The result loads on a phone and it feels like you're THERE.
The business opportunity is even crazier:
Freelancers are already charging between $300 and $800 per scan to real estate agencies, Airbnbs, shops, dealerships, museums.
One person + one phone + one weekend = a business.
Open source. Built on PlayCanvas.
GitHub Link Below
https://t.co/aY5bD3iDKC
Asked Claude Fable 5 to build a Windows OS and it built a fully functional web-based Windows OS clone in the browser - sign-in, notifications, Edge, Solitaire, the works. Plus, even created Copilot lol, A Minecraft clones, vision-based gameplay, insane 3D worlds & more.
Full demo: https://t.co/KLedfqInZn
Everyone is trying to make AI generate videos.
HeyGen just taught AI how to edit them.
Using HTML.
That's a much bigger deal than it sounds.
HyperFrames is an open-source framework that lets AI agents create, edit, preview, and render videos using HTML, CSS, and JavaScript. Think of it as "video-as-code" for the AI era.
«Write HTML → Render Video»
→ Build videos the same way developers build websites
«Built for AI agents»
→ Claude Code, Cursor, Gemini CLI, and other coding agents can generate and modify videos directly from prompts
«Browser preview + MP4 rendering»
→ Edit visually, then render production-ready videos locally
«Full video stack»
→ Includes a studio editor, rendering engine, player, and production pipeline in one ecosystem
Most creators work like this:
Open editor
→ Drag clips
→ Edit timeline
→ Export video
HyperFrames looks like:
Write code
→ Generate composition
→ Preview instantly
→ Render video
That's the shift.
Everyone thinks AI will help create content.
The bigger opportunity is letting AI operate creative tools directly.
Because once videos become code,
they become:
Version controlled
Automatable
Reusable
Agent-friendly
The future of content creation may not happen inside Premiere Pro.
It may happen inside VS Code.
And the people learning "video as code" today
are getting an early look at where AI-powered media production is heading.
Reminder: every Hugging Face Space is an API your agents can call :)
I asked mine to build a website about the flowers of France 🌸 and it used VAST AI's TripoSplat Space to turn photos it found into real 3D Gaussian splats, live on the page!
All on my HF Pro daily ZeroGPU credits (40 min/day renewed daily for only $9/month)
Andrej Karpathy's advice for beginners getting into AI:
"Put in 10,000 hours of work."
He's right.
But most builders waste the first 1,000 hours on the wrong things.
They write code before understanding context windows.
They build agents before understanding token limits.
They ship products before understanding what models can't do.
The builders who compound fastest aren't the ones who code the most.
They're the ones who understood the fundamentals before touching a single line.
These are the 10 concepts that make the first 1,000 hours count ↓
Bookmark this before you start.