- AI Toast
- Posts
- Anthropic launches Claude Sonnet 4.5
Anthropic launches Claude Sonnet 4.5
Plus: How to use the world’s first Agentic Browser
Poonam Soni
September 30, 2025
Hey there, and welcome to AI Toast!
Today’s menu:
Claude Sonnet 4.5 is here 🔥
SB 53 was signed into law 👨⚖️
Learn how to use the world’s first Agentic Browser 😱
ChatGPT Parental Controls 👨👩👧
AI Chrome Extensions 🕶️
Quick AI Toasts: Microsoft’s new AI Agents 💬
Total read time: About 6 minutes, perfect for a quick coffee break.
Anthropic just dropped Claude Sonnet 4.5, promising a genuine leap in AI coding and agent capabilities. This release isn’t just faster, it's a rethink of what smart, safe, and truly useful AI means for real-world work.
Best in Coding: Sonnet 4.5 now tops global coding tests (SWE-bench Verified), excelling in long, complex, real-world software tasks.
Agent and Tool Upgrades: New features like checkpoints, a streamlined terminal, VS Code extension, smarter memory, and even direct code execution and file creation in conversations.
Deep Reasoning, Strong Math: Marked improvements show up in logic, math, and handling huge, multifaceted projects.
Real-World User Praise: Developers at Canva, GitHub Copilot, Figma, and more report massive speed and accuracy gains, with Sonnet 4.5 reducing errors, supporting agentic coding, and transforming workflows.
Safety and Alignment: It’s their safest model yet, aggressively tackling risky behaviors and prompt injections, released under strict AI Safety Level 3 standards.
Build Your Own Agents: The Claude Agent SDK is now public, giving builders the same toolkit Anthropic uses internally.
A Glimpse of the Future: “Imagine with Claude” experimental preview shows AI generating and evolving entire software on-the-fly, pointing to creative, collaborative AI design.
Claude Sonnet 4.5 signals a step-change for developers and teams using real AI, with a strong emphasis on reliability and safety.
Personal Take: This is a turning point. Soon, AI won’t just support creative work, it’ll genuinely co-create and automate deep tasks, drawing us closer to seamless human-machine teamwork. The future? Think less about “using AI,” and more about what you’ll build with it.
Small Budget, Big Impact: Outsmart Your Larger Competitors
Being outspent doesn't mean being outmarketed. Our latest resource showcases 15 small businesses that leveraged creativity instead of cash to achieve remarkable marketing wins against much larger competitors.
Proven techniques for standing out in crowded markets without massive budgets
Tactical approaches that turn resource constraints into competitive advantages
Real-world examples of small teams creating outsized market impact
Ready to level the playing field? Download now to discover the exact frameworks these brands used to compete and win.
Image Source: ca.gov
In a bold move, California has enacted SB 53, a landmark bill aimed at reigning in AI’s wild west. It’s the first US law demanding greater transparency and responsibility from the world’s leading AI labs.
SB 53 was signed into law by Governor Gavin Newsom on September 29, 2025.
The new rules mandate transparency about safety protocols for major AI labs like OpenAI, Meta, Google DeepMind, and Anthropic.
Companies must now report critical AI safety incidents, including criminal acts or manipulative AI behavior to the state’s emergency services.
Whistleblower protections are strengthened for employees speaking out on safety concerns.
While Anthropic has supported the bill, OpenAI and Meta pushed back, warning state laws could splinter regulation and slow innovation.
The law may inspire other states to adopt stricter AI guardrails, New York has a similar bill waiting for approval.
Newsom’s statement emphasizes balancing innovation with public trust and community protection.
California just took the lead on AI accountability, and the ripple effect could redraw the US tech landscape.
Personal Take: AI’s future doesn’t stop at innovation, it’s also about trust and, safety. California’s move makes the conversation real, and the world is watching for what’s next.
Learn how to use the world’s first Agentic Browser
Go to the Fellou website and download. No invite code. Unlimited access. Limited Time.
Open Fellou CE. You’ll see a dashboard where you can launch AI-powered tasks or browse as usual.
Want to research a topic? Use “Deep Search” to search across multiple sites, Fellou gathers information from various sources, even from your logged-in accounts.
Example: Research “travel trends 2025” and receive a compilation of insights from travel blogs, booking sites, and reports, all in one visual summary.
Deep Search and Visual Report is 100% Free.Paste long articles or reports, and use Fellou’s summarizer. The browser will create easy-to-read visual insights so you can grasp key details quickly.
Sometimes, you don’t even need to paste. Use @Web in Fellou, just type @ in chat to pull exactly what you need from any webpage.
Example: “Summarize the latest product announcements from this page {@Web}” → you get a ready-to-use summary instantly.Fellou can fill forms, scrape data from web pages, and organize your downloads or files automatically. For example, automate job searches on multiple platforms, Fellou fills out forms and sorts results for you.
Activate local file management, it will sort your desktop files by type or date, color-code them, and make everything searchable.
A few other features:
Multisite Deep search and Deep action
Visual and summarized insights
Automated web and local tasks
Bulk job application and travel planning agents
File sorting on your computer
Personal website builder (AI-powered)
Sign up now and get 1000 Sparks as a welcome gift
No Invite code. Unlimited Access. Limited Time.
Fellou CE is an Agentic browser that automates, organizes, and simplifies your digital life!
Choose the Right AI Tools
With thousands of AI tools available, how do you know which ones are worth your money? Subscribe to Mindstream and get our expert guide comparing 40+ popular AI tools. Discover which free options rival paid versions and when upgrading is essential. Stop overspending on tools you don't need and find the perfect AI stack for your workflow.
OpenAI has officially launched parental controls for ChatGPT, giving parents more ways to manage how teens use the AI platform. As AI tools become part of everyday life, this update is designed to create a safer and more balanced experience for families—without locking teens out entirely.
How to get started:
Link your teen’s account, then choose the settings that fit your family. You can change these anytime.
Open Settings via your profile icon.
Select Parental controls.
Click + Add family member (see Invite teen steps below).
Once your teen accepts, select their name under Family members.
Adjust settings like:
Content Filters: “Reduce sensitive content” is automatically enabled once accounts are linked. This filters out graphic material, harmful viral challenges, and other age-inappropriate responses.
Quiet Hours: Set blackout times when ChatGPT is off-limits.
Memory: Choose whether ChatGPT remembers past chats.
Model Training: Control if chats contribute to AI improvements.
Voice Mode: Turn off voice chat if needed.
Image Generation: Disable AI images entirely.
Invite Teens: Link your teen's account to choose settings together. Invite by phone (web only) or email, and change or remove the link anytime. One parent can link with multiple teens; however, each teen links with one parent.
After acceptance, you'll see them under Family members and can adjust settings anytime. Note: Teens can unlink accounts, and parents will be notified.
Additional Features:
Safety Notifications: Receive alerts if ChatGPT detects potential harm or distress in your teen's conversations.
Account Management: You can unlink accounts at any time, and your teen can also choose to disconnect their account.
Personal Take: Overall, I think this is a step in the right direction. As AI tools become more deeply woven into everyday life, it would be irresponsible not to build in safety features oriented toward minors. These parental controls show OpenAI is thinking about real-world impact and not just tech capabilities.
AI Chrome Extensions
Perplexity AI Chrome Extension: Get lightning-fast AI-powered answers and research in your browser, right where you’re working.
AI Blaze: Instantly trigger AI (like GPT-5 or Claude Sonnet) to write, summarize, or edit text anywhere on the web using shortcuts and custom prompts.
Monica: An all-in-one AI assistant that helps you chat, summarize content, write, and code faster, available as a browser extension and app.
QuillBot: An AI writing tool to paraphrase, check grammar, summarize content, and improve any text for effortless writing.
Fireflies: Automatically transcribe, summarize, and analyze meetings, making it easy to capture key points and actions without manual note-taking.
Quick AI News Bites
Microsoft 365 Copilot introduces Agent Mode and Office Agent, giving users smart AI help for tasks in Excel, Word, and PowerPoint. This new feature lets you create and analyze spreadsheets and documents with simple prompts inside Microsoft 365 Copilot. You can collaborate with Copilot to make your work faster, easier, and more polished.
Brave’s search engine now has a 'detailed answers' feature, so you can get longer, report-style responses complete with useful links, videos, and product info. You don’t need to do anything special — just ask your question and Brave Search will handle the rest. The company encrypts and deletes chats after 24 hours to keep your privacy safe.
Ex-Google CEO Eric Schmidt warns the US could fall behind in the AI race, as China focuses on practical applications and favors open-source models. Schmidt explains that China’s accessible AI and open training data make it easy for other countries to adopt their tech, posing a challenge for the US.
DeepSeek has released a new ‘sparse attention’ AI model, designed to significantly lower API server costs for long-context tasks. The model uses smart prioritization of important data, so developers might pay half as much for inference when using DeepSeek’s new model. Since it’s open-weight and free to test, other providers could soon adopt similar tricks.
Boost revenue and gain new customers by partnering with us
Reach over 35K AI enthusiasts with your product.
Join our newsletter to connect with tech professionals, investors, engineers, managers, and business owners worldwide. DM now!
Got feedback, a story worth toasting, or a wild tech question? I’d love to hear from you; just reply, or find me on X.
Cheers,
— Poonam Soni