Hey — Here’s last week AI roundup.

This week, Meta grabbed the spotlight with an AI system that can turn brain signals into words without surgery. It's one of the biggest steps yet toward practical brain-computer interfaces and shows how quickly AI is moving into healthcare.

The rest of the week was just as busy. We cover Anthropic's new Claude Science, OpenAI's first AI chip, China's latest AI advances, and more. lets dive in:

In today’s AI researches:

• Meta’s AI Reads Brain Signals Into Words Without Surgery

• Anthropic just launched Claude Science

• China Says Its AI Can Now Match Mythos at Finding Bugs

• Learn How to Give Your AI Agent a Credit Card Safely

• AI Researches case studies

• AI Guides

• wild predictions from Anthropic CEO, Cloudflare CEO

• Best performing Ai tools of the week

• More Ai news and developments

Read time: 5 minutes

 NEW 

🧠 Meta’s AI Reads Brain Signals Into Words Without Surgery

AI Researches: Meta introduced Brain2Qwerty v2, an AI brain-to-text system that reads full words and meaning from brain signals while people type, bringing no-implant brain scanning much closer to surgery-level results.

The Details

• Nine volunteers spent around 10 hours in a brain scanner, typing while Meta’s AI collected nearly 22,000 sentences of brain-signal data.

• Brain2Qwerty v2 uses one AI model to read raw brain signals and another to understand the meaning behind the words.

• The system reached 61% average word accuracy, while the best volunteer hit 78%, far ahead of older no-surgery methods.

• Meta also released the code for v1 and v2, so other researchers can test the AI system and improve it faster.

What if AI handled your job search tonight?

Job hunting is exhausting.

AIApply makes it automatic.

Your AI Career Agent works 24/7 to:

• Find relevant jobs online

• Tailor your resume instantly

• Generate personalized cover letters

• Auto-apply while you sleep

• Help you land more interviews

No more endless tabs.
No more repetitive applications.
No more wasting hours every week.

AIApply helps you apply faster, smarter, and at scale so you can focus on what actually matters: getting hired.

Start landing interviews faster.

 iMajor AI Developments that Happened in this week i

Anthropic just launched Claude Science, an AI workspace built for scientific research. The app connects to scientific databases, lab tools, and HPC systems, helping researchers analyze data, write code, and draft papers in one environment.

OpenAI is rolling out GPT-5.6 Sol, Terra, and Luna first to trusted partners at the U.S. government’s request, with wider access planned after safety checks around coding, science, and cybersecurity.

OpenAI just unveiled Jalapeño, its first custom AI chip. Built with Broadcom in just nine months, the chip is designed to run ChatGPT, Codex, and future AI agents more efficiently while reducing OpenAI's reliance on Nvidia.

Anthropic launched Claude Tag, turning Claude into an AI teammate inside Slack. Teams can simply tag @Claude to delegate tasks, with the AI working across connected tools, channels, and codebases before replying when the job is done.

 TRENDING 

🛡️ China Says Its AI Can Now Match Mythos at Finding Bugs

AI Researches: Chinese AI systems are reportedly catching up to Anthropic’s Mythos in cybersecurity, with Z.ai’s GLM-2 and 360’s new tools claiming strong bug-finding power while the U.S. limits access to its own top models.

The Details

• Researchers found Z.ai’s new GLM-2 model can find software bugs about as well as Mythos, according to a Wall Street Journal report.

• China’s 360 also says its Tulongfeng tool found 3,432 software flaws, including 105 confirmed by Chinese officials, though Reuters could not fully verify the claims.

• 360 founder Zhou Hongyi said China cannot let only the U.S. hold AI tools that could change cyber offense and defense.

• Instead of only chasing bigger models and better chips, 360 says it is using an agent setup that mixes AI, security knowledge, bug databases, and automated tools.

 AI GUIDE FINDINGS 

Cut Claude Code Costs With Smarter Model Routing this guide breaks down how developers are using Kimi K2.7 for cheap, routine coding work and Claude Opus 4.8 for high-stakes reasoning, showing how to reduce AI coding costs while keeping quality for architecture, debugging, security, and final code reviews.

Build AI Agents That Work in Loops
Learn how to design agentic systems that discover tasks, hand off work, verify results, save progress, and repeat on a schedule instead of depending on one-time prompts.

Run AI Agents Locally on Mac
Learn how to set up multiple local AI agents on a Mac that can read code, check GitHub, write reports, test changes, fix bugs, and run without cloud costs or API keys.

Researchers at Boston Children's Hospital and Harvard used OpenAI’s o3 Deep Research to analyze 376 previously unsolved pediatric genetic cases, helping doctors confirm 18 new rare disease diagnoses that had been missed despite earlier specialist reviews. The AI searched symptoms, genes, and the latest research, giving families with long-standing unanswered cases another chance at finding a diagnosis.

New research from Exponential View estimates the generative AI industry made $110 billion in revenue last year and could grow to $175 billion this year, expanding faster than any previous technology, including the internet. Lower AI costs are driving much higher usage, businesses are adopting AI more quickly, and the rapid growth is even increasing electricity demand as more data centers are built.

💳 How to Give Your AI Agent a Credit Card Safely

1. Set up AgentCard
Install the AgentCard CLI, sign in, complete the official Stripe setup, then create a low-limit virtual card for a small test purchase.

2. Let your AI shop safely
Give your AI agent a specific task with one merchant, one product, a maximum budget, and tell it to stop before placing the order.

3. Review before paying
Approve any verification prompts, check the final checkout page yourself, complete the purchase if everything looks correct, then review the transaction and close the virtual card.

🔗 Learn more→

Anthropic CEO Dario Amodei told lawmakers that open-source AI is moving down a “very dangerous path,” warning that once powerful models are released publicly, companies can no longer monitor misuse, revoke access, or update safety guardrails.

Cloudflare CEO Matthew Prince said AI could “destroy small businesses” as buying decisions shift from humans to agents that may not easily discover or trust smaller brands.

• 🧩 HydraDB — Runs a graph database on object storage for AI workloads, powering agent memory, ontologies, and company-wide knowledge graphs.

• 🖥️ Hostinger — Offers AI-managed VPS hosting with AMD EPYC, NVMe storage, API access, and free weekly backups.

• 💡 IdeaProof — Validates startup ideas in 120 seconds with market size, competitor SWOT, and financial projections from 50+ live sources.

• 📚 Automateed — Generates 150+ page books with AI images, designed covers, and publishing support through its own marketplace.

• 💬 Reply.io — Automates outreach across email, LinkedIn, calls, and SMS so sales teams can spend less time on manual tasks.

• 💘 MGAI — Writes dating-app replies using lessons from The Message Game and real messaging examples.

Anthropic upgraded Claude Design with brand-aware editing and Claude Code sync. The beta now imports design systems from repos, files, or codebases, edits directly on canvas, and exports work to PDF or PowerPoint.

Cursor just launched an iPhone app for AI-powered coding. The new iOS app lets users launch coding agents by voice or text, switch between cloud and local machines, and review pull requests on the go.

Anthropic says Alibaba carried out the largest AI distillation attack on record.
The company claims nearly 25,000 fake accounts generated 28.8 million Claude conversations in just 45 days to extract advanced AI capabilities

The AI boom is making Apple products more expensive. Apple has raised prices on Macs and iPads worldwide, citing surging memory costs caused by massive demand from AI infrastructure.

Anthropic, OpenAI, and Stripe just launched a $500M initiative to fight the common cold. The nonprofit, Intercept, will fund early-stage research into vaccines, antiviral treatments, and cleaner indoor air technologies to prevent respiratory illnesses.

Meta is making AI smart glasses cheaper and more accessible.
Its new Meta Glasses replace the Ray-Ban branding, starting at $299 while adding improved AI features powered by Muse Spark.

SpaceX’s Colossus data centers are becoming an AI moneymaker.
The company signed a $6.3B compute deal with Reflection AI, adding another tenant alongside Anthropic and Google.

Sakana’s new Fugu model is turning AI orchestration into a frontier strategy.
It uses multiple AI models behind one API to handle coding, research, and security tasks