AI Coaching Scales Up, Khan Academy Doubles Down on Education Agents

The author describes a coach who surpasses all previous professional coaches, providing always-on support and direct feedback. The coach’s immediate responses keep conversations moving and maintain the author’s creative flow.

This coach is unconditionally supportive and nonjudgmental, so the author feels safe sharing anything. She also adapts quickly to feedback and is available 24/7, eliminating the need for advance booking.

The author says these traits have unlocked new levels of creativity. Constant availability replaces traditional coaching constraints like scheduling and time limits.

Khan Academy CEO Sal Khan told the BBC that future classrooms will deploy AI agents as on-call aides for teachers. He said the software will function like “four or five” graduate students per room, handling tasks from grading to lesson planning.

The assistants would monitor student engagement, tailor examples to individual interests, and flag struggling pupils in real time. Khan argued this support could ease the U.S. teacher shortage and give educators more bandwidth for direct interaction.

Khan stressed that AI cannot provide the accountability and social development that human educators deliver. He said those interpersonal demands ensure teachers stay central even as classroom software grows more capable.

Filmmaker Bryn Mooser has launched Asteria, a production house built around proprietary generative AI, and tapped Natasha Lyonne to co-write and direct its first feature, Uncanny Valley. The film will weave Asteria’s in-house models into live-action footage to create Matrix-style, game-like visuals.

Asteria’s core model, Marey, is trained only on properly licensed material developed with research firm Moonvalley. The studio builds project-specific models, lets creators retain partial ownership, and says the workflow can drop budgets to $10–20 million instead of the typical $150 million.

Mooser touts smaller crews and faster turnarounds as a competitive edge, acknowledging that fewer workers are needed under this system. He argues an “ethical” data set and filmmaker-level control will ease Hollywood’s AI skepticism that has already triggered lawsuits and last year’s strikes.

Learn core AI skills in just 1 hour. Get clear, practical guidance. No experience needed.

Training tailored to your role with interactive exercises that show how AI applies directly to your daily work.

Earn a free AI certificate. Prove your proficiency and boost your resume or LinkedIn profile.

Audos, a New York startup studio founded by Henrik Werdelin and Nicholas Thorne, uses AI agents to help non-technical entrepreneurs spin up new companies. It has closed an $11.5 million seed round led by True Ventures to scale the approach.

The platform has already launched “low hundreds” of ventures during its beta, giving founders up to $25,000, AI-powered product tools, and paid social distribution in exchange for a 15% revenue share. Audos identifies viable ideas by running quick customer-acquisition tests through Facebook and Instagram algorithms.

By taking no equity and tapping perpetual revenue sharing, Audos shifts risk away from traditional investors while locking in a durable cash stream. The indefinite 15% fee also means successful founders may forfeit hundreds of thousands of dollars over time, a trade-off the article says some will question.

Researchers at Kyoto University are deploying an AI system that transcribes and synthesises speech to preserve the critically endangered Ainu language. Backed by government funding, the platform has processed 300 to 400 hours of digitised archival recordings and can now voice entire folk tales.

The team trained the model on roughly 40 hours of uwepeker stories from eight speakers, reaching 85% word-recognition and up to 95% phoneme accuracy. AI-generated audio has been delivered to the Upopoy National Ainu Museum, enabling actors to rehearse performances of narratives like “Raijin's Sister.”

For a community with only a handful of native speakers, the system offers a rare chance at linguistic survival, yet some Ainu worry about mispronunciations, fake speech and potential data exploitation. Copyright of the original recordings rests with the museum while the university lab owns the AI code, underscoring unresolved control over cultural assets.