In partnership with

OpenAI released next-generation audio models in the API

Key Takeaway: OpenAI has released advanced speech-to-text and text-to-speech models, enabling developers to create smarter, more customizable voice agents with human-like interaction capabilities.

More Insights:

• Next-gen transcription: New models like gpt-4o-transcribe offer industry-leading accuracy, especially in noisy or accented environments.

• Expressive voices: The gpt-4o-mini-tts model can now be instructed on how to speak—e.g., empathetically or dramatically.

• Smarter training: Innovations in reinforcement learning and audio-specific datasets drive better performance and understanding.

• Easy integration: Developers can now plug these models into apps via OpenAI’s API and Agents SDK for fast, low-latency voice interactions.

Why it matters: As AI becomes more embedded in daily life, voice will be the most natural interface—these models mark a major step toward intuitive, lifelike AI communication that feels less like talking to a robot and more like talking to a human.

An entirely new way to present ideas

Gamma’s AI creates beautiful presentations, websites, and more. No design or coding skills required. Try it free today.

Claude can now search the web

Key Takeaway: Anthropic’s Claude chatbot can now search the web, offering real-time, cited information to compete with AI rivals like ChatGPT and Gemini.

More Insights:

• Web search is in preview for paid U.S. users, with wider access coming soon.

• Only the newest model, Claude 3.7 Sonnet, supports this feature.

• Responses include direct citations from sources like NPR, Reuters, and social media.

• Claude’s shift from being “self-contained” shows rising pressure from AI competitors.

Why it matters:
This move signals the evolving expectations for AI assistants—not just to be smart, but up-to-date. As AI tools become trusted sources of truth, their ability to pull in real-time, verifiable data could redefine how we interact with knowledge itself.

Cloudflare turns AI against itself with the endless maze of irrelevant facts

Key Takeaway: Cloudflare’s new “AI Labyrinth” tool tricks unauthorized AI crawlers into wasting resources on fake, irrelevant content instead of blocking them outright.

More Insights:

• Rather than blocking bots, AI Labyrinth feeds them convincing but meaningless AI-generated pages.

• The fake content uses factual info to avoid spreading misinformation while staying unrelated to the actual website.

• Human users won’t encounter the maze—it's hidden and only accessible to crawlers.

• Data collected from trapped bots helps improve future bot detection using machine learning.

Why it matters: As AI scrapers grow more aggressive and evasive, Cloudflare’s approach flips the script—using AI to defend against AI—raising ethical, technical, and environmental questions in the escalating arms race over web data.

Other stuff

• A high schooler built a website that lets you challenge AI models to a Minecraft build-off

• Man files complaint after ChatGPT said he killed his children

• How I force LLMs to generate the correct code

• Your A.I. Lover Will Change You

• OpenAI has released its first research into how using ChatGPT affects people’s emotional wellbeing

• AI start-up Perplexity confirms interest to buy TikTok

• Inside Google’s Two-Year Frenzy to Catch Up With OpenAI

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.

OR