My Daily AI & Markets Briefing

About My Daily AI & Markets Briefing

Your 15-minute daily briefing for people who build with AI. Two hosts cover the market movers that matter, the AI story of the day, a staff-level ML engineering interview question worked through step by step, and a fun Did You Know from tech history or a weekend build idea. New episode every morning. Questions or feedback: mahdi@brightmind-ai.com

Publisher

Mahdi Naser Moghadasi

Category

technology

Language

en

Explicit

No

First episode

16 Jun 2026

Latest episode

28 Oct 2026

Latest episodes

53 of 134 episodes have a transcript.

1. 

   28 Oct 2026TRANSCRIPT

   SOXL tops $100, DeepMind's Genie 4, and the KV-cache interview question

   Host Mahdi explores SOXL's rise, DeepMind's Genie 4, and interviews an expert on the KV-cache ML interview question.

2. 

   27 Oct 2026TRANSCRIPT

   SOXL pops 6%, OpenAI's ETH claim, and asking metrics first in system design

   Host Mahdi discusses SOXL's market surge, OpenAI's potential ETH disproof, and the crucial first question for system design interviews.

3. 

   26 Oct 2026TRANSCRIPT

   SOXL bounces 2.4%, Anthropic files for an IPO, and the LSTM vs. transformer question

   Host Mahdi discusses SOXL's market bounce, Anthropic's IPO filing, and explores the LSTM versus transformer debate in an interview.

Inside a recent episode

SOXL tops $100, DeepMind's Genie 4, and the KV-cache interview question

Published 28 Oct 2026 · Transcript excerpt

[…] Don't fail to give hard numbers. Don't ignore pre-fill versus decode. And don't confuse it with API-level prompt caching that just saves text strings. Okay. Give it to us. What does the perfect 30-second summary sound like? It sounds like this. Because past keys and values don't change, we cache them per layer to trade quadratic compute for linear memory footprint. This makes the decode phase completely memory bandwidth bound. Mathematically, it requires two times layers times KV heads times dimension times bytes, which is roughly 320 kilobytes per token for a 70B model in FP16. This cache lives dynamically in GPU memory managed in pages by engines like VLLM. […]

Reach and audience

Public platform figures. Ratings count people who left a rating, not total listeners.

Apple Podcasts (US)

5.0 / 5

1 ratings

Contact My Daily AI & Markets Briefing

Guest appearances

Does not typically book guests

Based on episode analysis; this does not confirm that the show is currently accepting guests.

Questions about My Daily AI & Markets Briefing