If we had to put it very simply..

A new deep research agent just beat OpenAI on every benchmark. > 3x more citations > Generates PDF reports from 500+ sources > Fully traceable, auditable output > Connects to Crunchbase, Drive, and more Can be used for M&A, due diligence, market and tech landscapes analysis.

Fully local Manus AI. No APIs, no $200 monthly bills. An autonomous agent that thinks, browses the web, writes code, and plans. Keeps all data on your device.

Wow. Sequoia just backed a startup going after Devin, Cursor, and OpenAI Codex. Their agents can read your entire codebase, write, test, fix, and merge PRs automatically It’s like having a fully autonomous software engineer by your side, 24/7. Examples 👇

This is a huge upgrade to traditional 3D reconstruction methods. It is more flexible and practical for real-world applications. This could be a game-changer for drones, AR, robotics—or basically anything that needs fast, accurate 3D mapping from video.

It doesn't need fixed camera setups or calibrations and can handle different lenses or focal lengths. When camera calibration is known, it reaches state-of-the-art accuracy, but even without it, the performance is still impressive.

It includes: 🧠 LLM basics: Models, price tiers, choosing the right one 🔍 Tool Use: Search, deep research, file uploads, Python execution 🎨 Multimodal AI: Working with images, OCR, video, voice AI 💻 Coding & Writing: Cursor, Claude Artifacts, NotebookLM

4/ DeepSeek has open-sourced many of its libraries in the past week and it's setting the community on fire. These tools (such as FlashMLA, DeepEP, etc.) make training and running large-scale LLMs faster and more efficient. https://t.co/FWCnr3BQqJ

3/ Amazon just introduced its next generation of AI assistant: Alexa+. It's more conversational, smarter, personalized, and equipped with agentic tools to get things done for you. It costs $19.99/month, but is free for all Prime members. https://t.co/euipXB6Lcd

2/ Anthropic released Claude 3.7 Sonnet, and it’s a game-changer for coding. It ranks #1 on real-world software benchmarks like SWE-bench and TAU-bench. https://t.co/yGpDtXWaje

1/ OpenAI just announced the release of GPT-4.5 for Pro users and developers. It's not a reasoning models, but is reported to have higher "EQ" and better capabilities than GPT-4 and GPT-4o. https://t.co/YX9ZEscSWl

Are you an AI developer? Check out https://t.co/LcabaCQ80b to get a daily summary of breakthrough models, repos and papers in AI. Read by 200,000+ devs.

Mercury is now available for all users for free. Try it now: https://t.co/N78K6vVs4h

Why it’s a game-changer: ▸ Runs on standard GPUs—no special hardware required ▸ Generates full code structures, not just one token at a time ▸ Optimized for RAG, tool use, and autonomous agents ▸ Ranks #1 in speed, #2 in preferred completions on Copilot Arena

It completes entire code blocks in a few steps, making it significantly faster than traditional models.

The fastest code LLM just dropped. It runs 10x faster than GPT-4o Mini at 1000+ tokens/sec on standard GPUs. Mercury is the first large-scale diffusion large language model (dLLM), generating text in parallel instead of one token at a time.

Are you an AI developer? Check out https://t.co/LcabaCQ80b to get a daily summary of breakthrough models, repos and papers in AI. Read by 200,000+ devs.

Project:

Impressive. Two AI agents on a phone call realize they’re both AI and instantly switch to encoded audio signal communication. They use a data-over-sound library called "ggwave" to send messages faster than speech. The project won 1st place in a hackathon.

Are you an AI developer? Check out https://t.co/LcabaCQ80b to get a daily summary of breakthrough models, repos and papers in AI. Read by 200,000+ devs.

GitHub: https://t.co/ZByVzHEqnL