Claude Found the Internet, OpenAI Found Its Voice

Good morning AI entrepreneurs and enthusiasts,

Claude is no longer living in the past — Anthropic has officially upgraded its powerful AI assistant with the ability to search and cite live web content.

Now that one of the most capable AI models has access to real-time information, will this overdue but vital update push Anthropic into the lead in the race for AI assistant dominance?

In today’s AI news:

• Claude now supports live web search
• OpenAI’s voice AI gets a personality injection
• Breathe new life into black-and-white photos
• Apple shakes up its AI team amid Siri issues
• Top Tools & Quick News

Claude gets real-time web access

The News: Claude now supports live web searches, closing a major gap between Anthropic and competitors like ChatGPT and Gemini.

The details:

• Claude 3.7 Sonnet can now determine when it needs fresh information from the internet.
• It cites sources directly, helping users verify and fact-check in real time.
• The feature is live for U.S. paid users, with plans to expand globally and to free users soon.
• To enable it, go to settings and switch on the Web Search tool.

Why this matters: It’s surprising Claude only just got web access, considering how long competitors have had it. But now, with real-time capabilities layered on one of the strongest models available, Anthropic could threaten more search-native tools like Perplexity.

Voice AI gets a major personality upgrade

The News: OpenAI just dropped new API-powered models for both text-to-speech and speech-to-text, marking a significant advancement in AI-powered voice technology.

The Details:

• The new gpt-4o-mini-tts enables users to generate speech using styles like “talk like a pirate” or “use a bedtime voice,” via simple text instructions — a major step in customized voice generation.
• gpt-4o-transcribe surpasses previous Whisper versions, offering state-of-the-art transcription across multiple languages and significantly improving accuracy in noisy environments.
• The public demo site, openai.fm, lets users experiment with various voices and styles in real time — a move that makes the tech more tangible and widely accessible (Tom's Hardware).
• Both models are accessible via API, and integrate with the Agents SDK for developers building voice-first apps.

Why it matters: These advancements move OpenAI closer to a future dominated by real-time, natural-sounding AI agents. However, rivals like ElevenLabs and Sesame still maintain an edge in hyper-realistic voice replication. Expect broader adoption of voice interfaces across entertainment, customer support, accessibility, and productivity platforms.

Apple retools Siri leadership amid delays

The News: Apple is restructuring its AI leadership, placing Mike Rockwell — creator of Vision Pro — in charge of Siri to help get long-promised upgrades back on track.

What’s changing:

• Rockwell now reports to software SVP Craig Federighi, shifting Siri oversight away from John Giannandrea, who reportedly lost Tim Cook’s confidence.
• Paul Meade, formerly leading hardware for Vision Pro, will now oversee the Vision Products Group.
• Siri’s most anticipated personalization features, previewed at WWDC, are still delayed without a clear release date.
• The Apple Intelligence platform, central to iPhone 16, has underperformed and was late to market.
• Internal tests exposed inconsistencies and technical issues, which fueled the leadership shakeup.

Why it matters: The shakeup followed urgent discussions at Apple’s Top 100 leadership summit, where AI delays were a central topic. While this restructuring signals a commitment to reviving AI innovation, Apple’s overall roadmap remains shaky. Whether this leadership pivot accelerates progress or simply reshuffles the same problems remains to be seen.

Top Tools

• Llama Nemotron – Nvidia’s new open-source reasoning model series.
• EXAONE Deep – LG’s advanced model for math and coding tasks.
• Epiphany – AI voice app for note-to-task conversion.

Quick News

• OpenAI’s o1-pro model launches via API, priced at $150–$600 per million tokens — 10x more than o1.
• Perplexity nears $1B raise at an $18B valuation, closing in on $100M ARR.
• NotebookLM now has interactive mind maps for visualizing notes.
• Meta expands AI assistant across 41 European countries after regulatory hurdles.
• Hugging Face issues statement supporting open-source AI over closed commercial systems.