Gemini 2.5 Pro is now available to everyone for free.

DeepSeek surpasses ChatGPT in monthly traffic

DeepSeek has become the fastest-growing AI tool globally, with its monthly new site visits exceeding OpenAI's ChatGPT, according to analytics platform aitools. xyz.

The report shows DeepSeek reached 525 million visits in February, surpassing ChatGPT's 500 million. Currently, DeepSeek holds 6.58% market share, ranking third behind ChatGPT (43.16%) and Canva (8.27%).

MoshiVis is the first open-source real-time speech model that can talk about images

It sees, understands, and talks about images — naturally, and out loud.

Voice interaction with a compact model endowed with visual understanding opens up new applications, from audio description for the visual impaired to visual access to information.

Runway introduced Gen-4 a new SOTA AI models for media generation and world consistency.

Amazon AGI Lab unveiled Nova Act, an AI agent system that can control browsers to perform tasks

— Beats Claude 3.7 Sonnet and OpenAI’s CUA on reliability
—Will power Amazon's next-gen Alexa+
— Also includes an SDK to build custom browser agents

AutoEval: Revolutionizing Robotic Policy Testing

Evaluating robotic manipulation policies has traditionally been a bottleneck in robotics research - requiring extensive human supervision to reset scenes, monitor tests, and record results.

The UC Berkeley team has developed AutoEval, a system that transforms this process.

Key Innovations:

Autonomous Evaluation: Robots test themselves with minimal human intervention (>99% reduction in human time)

High Correlation: AutoEval results match human evaluations with r=0.99 correlation

24/7 Testing: The system can run continuously, completing ~500 evaluation episodes daily

Public Access: Two AutoEval stations with four tasks are now available to researchers worldwide.

What makes this significant is that AutoEval combines the reliability of real-world testing with the scalability previously only possible in simulation. Unlike simulated environments (which often poorly correlate with actual performance), AutoEval provides trustworthy results on physical robots.

Researchers can now remotely submit their algorithms via a web interface and receive comprehensive evaluation reports - accelerating development cycles and enabling fair, standardized comparisons across different approaches.

Deloitte has released a new report looking at the current and future landscape of payment stablecoins

All hands released 2 big things

1. OpenHands LM: The strongest 32B coding agent model, resolving 37.4% of issues on SWE-bench Verified. It’s an open-weights 32B model based on Qwen.

OpenHands LM punches above its weight, resolving 37.4% of real-world issues, accuracy competitive to DeepSeek, a model 20x the size!

It can be run locally on a machine with a single 3090 GPU for instance. It also has 128k context for long agent trajectories and large code bases

2. OpenHands Cloud: SOTA open-source coding agents from your computer, phone, github, with $50 in free credits

OpenAI Academy just launched. Learn AI for free

academy.openai.com

New report. The $400B BCI Market. Who’s Leading, What They Own, and How IP Will Decide the Future

The Brain-Computer Interface (BCI) market is experiencing explosive growth, transitioning from research laboratories to commercial reality.

The industry is divided into 2 fundamentally different segments:

Enabling BCIs: Devices that translate neural intent into digital or mechanical outputs, helping patients with severe motor impairments (ALS, spinal cord injuries, stroke) control interfaces, robotic limbs, or communication systems.

Preventative BCIs: Systems that monitor neural activity for abnormal patterns that predict adverse events, such as seizures or depressive episodes, shifting neurological care from reactive to proactive.

Synchron The Quiet Winner of the BCI Race

Despite the heightened media attention on Neuralink, Synchron demonstrates more significant progress.

The company:

1. Was the first to receive FDA IDE approval for clinical trials of a permanently implanted system

2. Developed the revolutionary minimally invasive endovascular Stentrode™ technology

3. Built solid global patent protection (118 documents across 16 families)

4. Formed a strategic partnership with Nvidia in 2025

Investors recognized Synchron's potential - the company raised over $200M from ARCH, Khosla Ventures, Gates Frontier, and Bezos Expeditions.

Neuralink is developing a fully implanted high-bandwidth interface.

Key factors:
- The N1 implant with 1,024 electrodes and a robotic implantation system
- Valuation of $7.8 billion - the highest among private BCI companies
- Commencement of PRIME clinical trials with the first human implant in January 2024
- 87 patent documents across 31 families

Other Significant Players

Precision Neuroscience: Ultra-thin, high-density electrode arrays
Blackrock Neurotech: Leader in long-term implant stability
INBRAIN Neuroelectronics: World's first graphene-based neural interfaces
Paradromics: Leader in high-bandwidth neural data capture
Axoft: Advanced soft polymer implants

The BCI patent landscape is already actively forming:

1. Over 2,160 patent families from 664 organizations
2. Universities (Tianjin University, University of California, Stanford) own most of the foundational patents
3. Corporate IP leaders: Kernel, NeuroPace, Neurolutions, Cognixion, Panasonic

The winners in this race will be companies that:

- Integrate IP into their business strategy, not just accumulate patents

- Build defensible market positions in the early stages

- Align technology with the right regulatory pathway and reimbursement model

For investors, founders, and strategists, this rapidly evolving market presents a rare opportunity to shape the future of neurotechnology. The window of opportunity to define strategic leadership in this field is rapidly closing.

Simular introduced open-source Agent S2

Agent S2 blends generalist reasoning with specialist grounding for precise, long-horizon computer use tasks:

1. Mixture-of-Grounding
2. Proactive Hierarchical Planning
3. SOTA on OSWorld, AndroidWorld, and WindowsAgentArena

Paper.

The most important paper by Google about AGI. A few assumptions behind Google DeepMind’s paper on preparing for AGI:

1. no ceiling at human-level capabilities,

2. powerful systems by 2030,

3. positive loops via AI R&D,

4. a reliance on relatively continuous progress.

Anthropic Introduced Claude for Education.

They 're partnering with universities to bring AI to higher education, alongside a new learning mode for students.

Claude for Education is available today in London School of Economics and Political Science, Northeastern University, and Champlain College. It's also available for all Pro users with an .edu email.

Claude for Education gives academic institutions secure, reliable AI access for their entire community. For example:

1. Students can draft literature reviews with proper citations, work through calculus problems with step-by-step guidance, and get feedback on thesis statements before final submission

2. Faculty can create rubrics aligned to specific learning outcomes, provide individualized feedback on student essays efficiently, and generate chemistry equations with varying difficulty levels

3. Administrative staff can analyze enrollment trends across departments, automate repetitive email responses to common inquiries, and convert dense policy documents into accessible FAQ formats—all from a familiar chat interface with enterprise-grade security and privacy controls.

General Agents introduced Ace: The First Realtime Computer Autopilot

Ace is not a chatbot. Ace performs tasks for you.

On your computer. Using your mouse and keyboard.

At superhuman speeds.

Google DeepMind have achieved a notable milestone in AI with the Dreamer V3 algorithm.

Published in Nature, "Mastering diverse control tasks through world models" introduces a reinforcement learning algorithm that performs well across more than 150 diverse tasks with a single configuration.

Key Achievements

Versatility: Dreamer works effectively across continuous and discrete action spaces, visual and vector inputs, dense and sparse rewards, and various application domains.

Minecraft Diamond Challenge: Applied with its default parameters, Dreamer successfully collects diamonds in Minecraft without human data or curricula - a challenging task requiring farsighted strategies and learning from sparse rewards in an open world.

Consistent Learning: Through techniques based on normalization, balancing, and transformations, Dreamer enables more stable learning across domains that traditionally required extensive hyperparameter tuning.

Scaling Properties: The research shows that larger models achieve higher performance while requiring less interaction with environments, offering a predictable relationship between computational resources and efficiency.

While Dreamer represents an evolutionary advancement rather than a revolutionary breakthrough, it addresses a significant challenge in reinforcement learning: the brittleness of algorithms when applied to new domains. Traditional approaches require substantial human expertise and experimentation for each new application, limiting practical utility.

CAMEL-AI's Trifecta: Loong, OWL, and CRAB - The Future of AI Agent Systems

Loong: Self-Improving AI in Specialized Domains
Project Loong tackles the fundamental challenge of training LLMs to reason effectively in specialized domains without expensive labeled data. Instead of manually creating massive datasets, Loong generates and verifies synthetic data automatically.

GitHub. HF.

Key features:
Generates domain-specific synthetic Q&A pairs from small seed datasets
Verifies correctness through dual validation: code execution and Chain-of-Thought reasoning
Trains models through reinforcement learning on verified synthetic data
Currently supports 8 specialized domains including advanced mathematics, physics, computational biology, and finance
Loong enables LLMs to develop expertise in domains where curated data is scarce, unlocking new potential for specialized AI.

OWL: The Digital Task Automator
OWL (Optimized Workforce Learning) addresses real-world task automation with an impressive track record - ranking #1 among open-source submissions on the GAIA benchmark.
Key capabilities:
Browser automation via Playwright
Multi-search engine support
Python code execution
Document parsing across formats
Multimodal processing (video, images, audio)
Integration with numerous specialized toolkits
OWL's multi-agent architecture uses a UserAgent to break down tasks and an AssistantAgent to execute them using various tools, making it effective for complex workflow automation.

CRAB: Breaking the Environment Barrier
CRAB (CRoss-environment Agent Benchmark) is the first framework enabling agents to perform tasks across multiple environments - from smartphones to desktops and beyond.

The Integrated Ecosystem: MCP as the Universal Connector
All 3 projects integrate with the Model Context Protocol (MCP), introduced by Anthropic and rapidly becoming the "USB interface" of the LLM world. MCP provides standardized connections between AI assistants and data systems, enabling seamless operation across tools and environments.

Together, these projects represent a comprehensive approach to autonomous AI:

Loong trains specialized domain knowledge
OWL executes complex tasks
CRAB enables operation across diverse environments.

This trifecta addresses the "last mile" challenge in agent automation: long-term decision-making and adaptation. Where traditional agents follow instructions but don't truly evolve, CAMEL-AI's ecosystem creates environments where agents can perceive, act, and learn from experience.