Grammarly builds agents that don't need babysitting, researchers explain the AI disappointment cycle, and Hamburg shows how to make impossible group decisions possible
What it is: Grammarly has introduced eight specialized AI agents that provide targeted assistance for specific writing tasks, from academic research to professional communication. The agents work within a new writing interface called "docs" and include Citation Finder, Reader Reactions, AI Grader, and Expert Review.
What's new: These agents understand context and act immediately without requiring prompt crafting. Reader Reactions predicts how different audiences will respond to your writing. AI Grader evaluates work against uploaded rubrics before submission. Citation Finder locates supporting evidence and generates properly formatted citations. Additional agents handle plagiarism checking, tone adjustment, and field-specific expertise feedback.
Why it matters: Current AI writing tools require extensive prompt engineering to deliver useful results. These agents eliminate that friction by understanding your needs based on writing context. Students receive feedback aligned to actual course requirements. Professionals get audience-specific insights that improve communication effectiveness while preserving their authentic voice.
What it is: ChatGPT's new connector feature links directly to your Google Workspace apps—Gmail, Google Calendar, and Google Contacts. Once enabled, ChatGPT can automatically reference these tools during conversations without requiring manual file uploads or copy-pasting.
How it works: Once enabled, the connectors automatically reference your Gmail, Google Calendar, and Google Contacts when relevant to your conversation. This brings information from these tools directly into ChatGPT without manual file uploads or copying data between apps. Available now for Pro and Plus users.
Why it matters: This eliminates the manual work of copying information between your Google apps and ChatGPT. Instead of switching back and forth to gather context, ChatGPT can reference your connected tools directly during conversations. This streamlines workflows for anyone who regularly uses ChatGPT alongside Gmail, Calendar, or Contacts for work coordination and communication tasks.
What it is: A new MIT study found that 95% of AI projects fail because people spend more time verifying AI outputs than the AI actually saves them—what researchers call the "verification tax."
What's happening: The core problem isn't AI capability but AI overconfidence. Most AI tools present answers with equal confidence whether they're certain or guessing, forcing users to double-check everything. The 5% of successful implementations focus on systems that clearly communicate when they're uncertain and can learn from corrections over time.
Why it matters: This explains why your AI chat sessions might feel inefficient or why you've abandoned AI tools after initial excitement. Favor AI tools that acknowledge uncertainty or ask clarifying questions rather than always providing definitive answers. When using any AI tool, develop the habit of asking it to rate its confidence or explain its reasoning. This helps you identify when to trust the output versus when to verify independently.
What it is: Readwise is a reading and research platform that helps users capture, organize, and review highlights from books, articles, and other content. Their Reader app includes AI-powered features like Ghostreader for document summaries and Chat with Document for interactive analysis.
What's new: The platform now offers GPT-5 models across its AI features, with GPT-5-mini becoming the new default for document summaries (upgrading from GPT-4o-mini). Users can also access more advanced models like o3 when using their own API keys.
Why it matters: Better AI models change how you process information. Instead of skimming through long documents to find key points, you can ask specific questions like "What are the three main counterarguments to the author's thesis?" or "How does this research methodology differ from previous studies?" The improved reasoning helps you decide which documents deserve deep reading and which can be safely discarded after a summary—turning information overload into strategic selection.
