👾 AI Nuggets #17: The Battle of AI Assistants & Browser Agents

Welcome to edition #17 of the AI Nuggets. In this issue:

• Amazon unveils Alexa+

• Claude 3.7 Sonnet: breakthrough in software engineering

• GPT-4.5 research preview

• AI investments: are the magnificent 7 overvalued?

• Sesame crosses the voice uncanny valley

• Convergence's Proxy outsmarts OpenAI's Operator

• Web scraping got easier with ScrapegraphAI

Latest News

🅰️ Amazon Unveils Alexa+: The Next Generation AI Assistant

Amazon has introduced Alexa+, a completely reimagined AI assistant powered by generative AI. This next-generation Alexa is designed to be more conversational, smarter, and capable of taking action across thousands of services and devices. $19.99 per month and free for Amazon Prime members, Alexa+ will begin rolling out in the coming weeks, starting with Echo Show devices.

What's New

Alexa+ is built on Claude 3.7 and LLMs available through Amazon Bedrock but extends far beyond simple queries and responses. The system introduces a concept called "experts" – specialized groups of capabilities that help Alexa+ accomplish specific tasks. These experts allow Alexa+ to control smart home devices from brands like Philips Hue, make reservations through OpenTable, play music across multiple streaming services, order groceries, and more. Most impressive are the new agentic capabilities that enable Alexa+ to navigate the internet independently to complete tasks without supervision. You could ask Alexa+ to arrange an oven repair, and it would find a service provider through Thumbtack, authenticate, schedule the appointment, and report back when complete. The assistant also understands personal context, remembering your preferences, purchases, and important information you share.

Business Impact

The system will be available across multiple touchpoints – in homes, cars, a new mobile app, and a browser-based version at Alexa.com – creating a seamless ecosystem for users. Businesses should consider how to integrate with Alexa+ through the Alexa Skills Kit to engage customers through voice. The platform's partnership with services like Grubhub, Uber Eats, and Thumbtack shows how businesses can leverage Alexa+ as a new customer acquisition channel. For technical professionals, Amazon's breakthrough in getting LLMs to reliably orchestrate APIs at scale provides valuable insights for their own AI implementations.

👨🏼‍💻 Claude 3.7 Sonnet with Thinking: Developers Love It

Anthropic has released Claude 3.7 Sonnet, their most advanced AI model to date. It introduces a groundbreaking "extended thinking" capability that allows the model to show its reasoning process. This first-of-its-kind hybrid reasoning model works in two modes: standard for quick answers and extended thinking for complex problems. Alongside this, they launched Claude Code, a command line tool that helps developers with coding tasks.

What's New

The model can now share its thinking process with users. This means you see how it reaches conclusions. API users can control the thinking budget of up to 128,000 tokens. The system excels at coding tasks with major improvements in handling complex codebases and full-stack updates. Anthropic also expanded GitHub integration to all Claude plans. This lets developers connect code repositories directly to Claude.

Business Impact

The model especially excels at software engineering tasks. Developers can bring their productivity to another level.

💬 GPT-4.5: OpenAI's Most Advanced Chat Model Released

OpenAI has released GPT-4.5, their largest and most capable chat model to date. Available now as a research preview to Pro users and developers, this model represents a major advancement in scaling unsupervised learning. Early testing shows interactions feel more natural with GPT-4.5, which demonstrates improved knowledge, better understanding of user intent, and greater emotional intelligence. OpenAI designed it to be more useful for writing tasks, programming, and solving practical problems while reducing hallucinations compared to previous models.

What's New

GPT-4.5 focuses on scaling unsupervised learning to improve world model accuracy and intuition. This differs from reasoning models like OpenAI o1 and o3-mini that rely on chain-of-thought processing. The model shows significant improvements in factual accuracy and reduced hallucination rates on knowledge questions. Human testers preferred GPT-4.5 over GPT-4o across creative, professional, and everyday queries. While hallucinations are reduced, it’s not a major step up on benchmarks.

Business Impact

The model is available immediately to ChatGPT Pro users, with rollout to Plus and Team users next week, followed by Enterprise and Edu users. Developers on paid API tiers can access GPT-4.5 through Chat Completions API, Assistants API, and Batch API. The model supports key features like function calling, Structured Outputs, streaming, and system messages. Due to its size, GPT-4.5 is very expensive and compute-intensive ($75/$150 per million input/output tokens). It’s best suited for applications requiring superior emotional intelligence and creativity.

AI Investments - What to Watch¹

🇺🇸 Are the Magnificent 7 Overvalued?

The US stock market now represents 60.5% of global market capitalization. This concentration raises questions about sustainability. The "Magnificent 7" tech giants – Apple, Microsoft, Alphabet, Amazon, Meta, Tesla, and Nvidia – drive much of this dominance. These companies command premium valuations based on future AI-driven growth projections. Analyst expectations show varied profit growth trajectories over the next five years. Nvidia leads with 193% expected growth, fueled by AI chip demand. Amazon follows at 137%, with Netflix at 128% and Microsoft at 107%. Alphabet projects 76% growth while Apple shows a more modest 45%. But Tesla stands alone with a staggering 258% projected growth rate.

Tesla: Overvalued or Underappreciated?

Tesla deserves special attention. The company achieved a remarkable milestone in 2024, with Model Y becoming the world's best-selling car at 1.09 million units. This real-world market penetration supports the bull case. Elon Musk himself believes analyst projections underestimate Tesla's potential. He recently responded to growth forecasts stating "It will require outstanding execution, but I think more like 1000% gain for Tesla in 5 years is possible." This would represent a 10x return on investment. Tesla bulls point to the company's advantages in AI, autonomous driving, energy storage, and manufacturing efficiency. Bears counter that competition intensifies daily and Tesla's valuation assumes near-perfect execution across multiple ambitious initiatives.

Investment Considerations

Smart money sends mixed signals. Warren Buffett now holds a record cash reserve of $325 billion at Berkshire Hathaway. JP Morgan CEO Jamie Dimon sold personal holdings worth $230 million. These moves from experienced market veterans suggest caution. US markets have reached historically high valuations across multiple metrics. The fundamentals appear strong, with even the US government cutting expenses through the newly formed DOGE department. However, concentration risk remains significant. Prudent investors might consider diversification beyond the Magnificent 7. Alternative options include commodities, which often perform well during inflationary periods, or international exposure to markets like China and Latin America. These regions may offer better value propositions as capital rotates toward higher potential returns. The AI revolution creates transformative opportunities, but not all current valuations will be justified by future performance.

AI Use Cases and Tools

🗣️ Beyond Robotic Responses: The Future of Voice AI

Sesame has unveiled groundbreaking research aimed at crossing the "uncanny valley" of conversational voice technology. Their Conversational Speech Model (CSM) represents a significant leap forward in creating AI voices that feel genuinely human. Traditional voice assistants speak in neutral tones that lack emotional depth. This flatness makes them useful for simple tasks but exhausting for extended interactions. Sesame aims to achieve "voice presence" – the quality that makes spoken interactions feel authentic and valued.

This advancement opens new possibilities for customer engagement. Imagine voice assistants that truly understand customer frustration and respond with appropriate empathy. Support lines could maintain a consistent tone across interactions without sounding robotic. Sales and marketing tools could engage potential clients with warmth that builds trust. Businesses that prepare for this shift will have significant advantages in customer experience and operational efficiency as these technologies mature.

🕵️‍♂️ Convergence Proxy: A Practical AI Tool for Web Tasks

Convergence's Proxy is making waves in the browser-use agent market by outperforming OpenAI's much-hyped Operator. This UK startup offers a more affordable and often more effective solution for autonomous web navigation. Proxy launched in December 2024 and already demonstrates superior reasoning capabilities across a range of real-world tasks. The pricing model makes it accessible to businesses of all sizes – offering five free sessions daily or unlimited access for just $20 per month. This presents a dramatic cost advantage over Operator, which requires a $200 monthly ChatGPT Pro subscription.

Proxy excels through its "Generative Tree Search" technology. This approach creates models that predict web states after actions are taken, allowing it to explore possible outcomes before choosing the optimal path. In practical testing, this translates to more flexible problem-solving. When asked to book a romantic restaurant at noon in Napa, Proxy showed advanced reasoning by starting with availability first, then filtering for romantic options. Operator approached linearly and reached a dead end when its first choice had no availability. Proxy also performs better at product searches, finding items on Amazon more efficiently than its competitors.

You can leverage Proxy for numerous time-saving tasks. Try prompts like "Find the three most affordable accounting software options for small businesses and summarize their features" or "Book a business hotel in Chicago near the convention center for next Tuesday through Thursday." For e-commerce research, "Compare prices for [your product] across major retailers and identify the best deal with shipping included" delivers actionable insights. Customer research becomes simple with "Find and summarize recent reviews for [competitor product] focusing on complaints and limitations." The free tier lets you experiment without commitment, making this an ideal entry point for businesses looking to automate repetitive web tasks and focus human talent on higher-value activities.

MadebyAgents Updates

👾 What if Web Scraping Could be Smarter, Faster, and Easier?

Join me and the founder of ScrapeGraphAI as we explore the cutting-edge of AI-driven data extraction!

This video is packed with a tutorial and an insightful interview.

That’s It for This Week

✨ Before You Go:

We’d love to hear what you think. Please share your opinion.

See you next time!

Tobias from MadebyAgents