In partnership with

Good morning!

DeepSeek shares new research that could make AI models cheaper and more stable.
OpenAI is also gearing up for smarter voice AI as it prepares a new voice-first device.

In todays email:

  • Daily Update

  • Social Media

  • Today’s Highlights

  • YouTube

  • Today Trend

  • Social Media

Read time: 5 min

Learn AI in 5 minutes a day

This is the easiest way for a busy person wanting to learn AI in as little time as possible:

  1. Sign up for The Rundown AI newsletter

  2. They send you 5-minute email updates on the latest AI news and how to use it

  3. You learn how to become 2x more productive by leveraging AI

DAILY UPDATE

DeepSeek shared new research that hints at cheaper and more stable AI models.

DeepSeek released a new paper showing a fresh way to design neural networks that can cut costs and improve training stability. The work may point to big gains in its next major AI release.

  • The paper introduces mHC, a method that makes large AI training more stable with very little extra compute cost.

  • CEO Liang Wenfeng helped write the paper and uploaded it himself, showing strong personal involvement in the research.

  • Tests on 3B, 9B, and 27B models showed better results, especially on reasoning tasks.

This shows DeepSeek is still finding ways to build stronger AI at lower cost. After the success of R1, this research suggests future models could be even more efficient. With better chip access and steady research progress, China’s AI models could become far more competitive by 2026.
Read more…

SOCIAL MEDIA

TODAY’S HIGHLIGHT

OpenAI is reshaping its teams to build better voice AI for a new device coming soon.

OpenAI is bringing several teams together to improve its audio AI, as it prepares for a voice-first personal device designed by Jony Ive’s team, expected in about a year.

  • OpenAI’s voice models are still slower and less accurate than text-based ChatGPT, which pushed the company to reorganize its teams.

  • A new upgraded voice model planned for early 2026 will let users speak while the AI is talking, making conversations feel more natural and smooth.

  • The first hardware device is still about a year away and will focus on voice over screens, with ideas like smart glasses and a smart speaker also explored.

This move shows OpenAI is serious about going beyond software and into real-world devices. With Jony Ive’s design leadership and a focus on avoiding phone addiction, the project has strong attention.
Read more…

YOUTUBE

TODAY TREND

Clear for Slack
Clear messages get answered quicker

Community Figma MCP server
Allow AI Agents to help you with Figma designs!

Griply 2026
Achieve your goals with a goal-oriented task manager

vibecoder.date
Find who you vibe with, git commit to love

HappyMail
We made email simple again

SOCIAL MEDIA

That’s it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve the experience for you.

How was today’s email?

(Tell us what you liked or what could be better)

Login or Subscribe to participate

Keep Reading