Perplexity Launches Deep Research

Hello, AI Enthusiasts!

Welcome to FavTutor’s AI Recap! We’ve gathered all the latest and important AI developments for the past 24 hours in one place, just for you.

In Today’s Newsletter: 😀

• Perplexity Launches Deep Research

• Boomer prompts' don't work with its new reasoning models, OpenAI says

• Anthropic prepares next major Claude model

Perplexity
😮 Perplexity Launches Deep Research

Perplexity has introduced Deep Research, a new AI research tool that can analyze hundreds of sources and create detailed reports in minutes.

Insights for you:

• Perplexity has launched Deep Research, a free AI research tool that can analyze hundreds of sources and generate detailed reports quickly. Unlike OpenAI's similar tool which costs $200/month, Perplexity offers this feature for free with 5 daily queries for free users and 500 for Pro users.

• The system works in three steps: comprehensive source searching, intelligent analysis and follow-up research determination, and clear report generation.

• Deep Research has demonstrated strong performance in AI testing, scoring 20.5% on Humanity's Last Exam (outperforming models like Gemini Thinking and DeepSeek-R1) and achieving 93.9% accuracy on SimpleQA.

OpenAI
⚡️ 'Boomer prompts' don't work with its new reasoning models, OpenAI says

OpenAI has published guidelines for effective use of its o-series models, emphasizing direct instruction over complex prompting techniques.

The o-series models, which OpenAI calls "the planners," specialize in tasks related to mathematics, engineering, law and finance. The GPT models serve as "workhorses" for quick, basic tasks. "One model family isn't better than the other—they're just different," OpenAI writes.

Insights for you:

• OpenAI has released new guidelines for effectively using its AI models, advising users to provide simple, direct instructions rather than relying on complex prompt engineering techniques.

• The company discourages the use of overly detailed "boomer prompts" and instructions like "think step-by-step" for its o-models, instead recommending concise, well-structured instructions using separators such as XML tags, along with clearly defined success criteria and constraints.

• OpenAI suggests that the o-series models, described as "the planners," are particularly well-suited for complex tasks in fields like mathematics, engineering, law, and finance, while GPT models serve as "workhorses" for quick, cost-effective standard tasks.

Anthropic
🧠 Anthropic prepares next major Claude model

Anthropic is set to release a new AI model in the coming weeks, according to a new report from The Information — which will combine traditional language capabilities with advanced reasoning features, excel at coding, and offer developers more control over balancing speed and compute.

Insights for you:

• The hybrid approach will allow the new model to function as either a standard LLM or a deep reasoning engine, adapting to different use cases on demand.

• The model will also introduce a sliding scale system that lets developers precisely control how much reasoning power to allocate to each query.

• At maximum reasoning, the model reportedly shows particular strength in real-world programming tasks and can handle large-scale codebases.

• Recent rumors had suggested that Anthropic already internally had a model better than OpenAI’s o3, but it hadn’t been released due to safety concerns.