implementation
On top of that, V3 embraced multi-token prediction (MTP). Rather than predicting text one word at a time and inspired by Meta’s FAIR (Fundamental AI Research) team’s ideas toward building "Better & Faster Large Language Models via Multi-token Prediction," it predicts multiple words simultaneously. Finally, a trick called FP8 training helps V3 run
... See moreEvan Armstrong • What Actually Matters (And What Doesn’t) for DeepSeek
Category design involves educating the market about a new, often-ignored problem as well as a solution that you can provide.
Category Pirates, Christopher Lochhead, Eddie Yoon, Katrina Kirsch, • The 22 Laws of Category Design
.psychology .implementation
Connection: Vulnerability, impartiality, empathy, and wonder
The majority of us have been taught that connection is earned through achievement. We believe that once we become successful, smart, or generous enough, we’ll be worthy of connection.
But people don’t want you to be perfect. They want to be connected to you.
Joe Hudson • Knowledge Work Is Dying—Here’s What Comes Next
The metrics featured most prominently in early product literature of hydraulic backhoes were shovel width (contractors wanted to dig narrow, shallow trenches) and the speed and maneuverability of the tractor.
Clayton M. Christensen • The Innovator's Dilemma: When New Technologies Cause Great Firms to Fail (Management of Innovation and Change)
If you want to improve your memory and concentration, you need to create a belief system that supports them.
Kevin Horsley • Unlimited Memory: How to Use Advanced Learning Strategies to Learn Faster, Remember More and be More Productive
Perhaps R1’s biggest breakthrough is the confirmation that you no longer need enormous data centers or thousands of labelers to push the limits of LLMs. If you can define what “correctness” means in your domain —whether it’s coding, finance, medical diagnostics, or creative writing— you can apply reasoning-oriented RL to train or fine-tune your own
... See moreEvan Armstrong • What Actually Matters (And What Doesn’t) for DeepSeek
The key lesson from this distribution should be that using the average as a comparison measure can be dangerous with any multiple. It makes far more sense to focus on the median.
Aswath Damodaran • The Little Book of Valuation
Or maybe you’re running a creative team. In a pre-AI world, in order to create consistency, you needed to reduce your work down to principles and systems so that writers and designers could consistently capture your brand's voice and style.
In a post-AI world, you don't necessarily need to reduce your work in such a way. Instead, you can find
... See moreDan Shipper • Five New Thinking Styles for Working With Thinking Machines
Since we can’t improve our people-reading skills that much, we have to focus our efforts on making others more readable.