Rami's Readings #149 - DeepSeek V4 & Mahjong

The latest on AI, LLMs, AI in Action Conference in Montréal, DeepSeek V4, More Headless Browsers, Luxury Malls, Mahjong, and more.

May 03, 2026

Welcome to Rami’s Readings #149 - a weekly digest of interesting articles, papers, videos, and X threads from my various sources across the Internet. Expect a list of reads covering AI, technology, business, culture, fashion, travel, and more. Learn about what I do at ramisayar.com/about.

👋🏼 Welcome New Subscribers

Hello! A hearty thank you for subscribing to Rami's Readings! There are quite a few new subscribers this week, thanks to recommendations from The AI Ethics Brief, Product Byte, and The VC Corner. I am thrilled to have you on board! In this newsletter, I curate the best papers, tweets, and articles I have read during the week focusing on LLMs, AI, economics, business, and technology news. You can learn more about me on my website.

👋 I’ll be Speaking at AI in Action in Montréal!

I’m heading back to Montréal on May 5th for the AI in Action conference.

I am super excited for this conference as the AI conversation has moved from what’s possible to what’s actually working. In my talk, I’ll cover the latest trends in LLMs and SLMs, orchestration, and multi-agent systems. We’ll also explore how to prepare for the next major moment in the AI landscape.

I’m especially interested in comparing notes with founders, executives, and builders on what it takes to move from impressive demos to scaled products: distribution, evaluation, cost, UX, and the operational work that turns AI experiments into real workflows and businesses.

If you’ll be at AI in Action in Montréal on May 5th, or nearby that week, reply to this newsletter or on LinkedIn. I’d be glad to connect. If you’d like a discounted ticket, please reach out to me.

trees near city — Photo by Jackie Hutchinson on Unsplash

🤖 AI Reads

DeepSeek V4 — Almost on the Frontier, a Fraction of the Price

Notes: More on DeepSeek V4 which we covered last week. Huawei Ascend was supposedly used to train the model indicating that NVIDIA’s GPU stranglehold is weakening.

Mike - The Open Source Alternative to Harvey and Legora

Notes: I assume the naming is inspired from Suits. Great traction on GitHub for 3 day old release.

Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses

Notes: Autoresearch for the coding agent harness itself.

Recursive Multi-Agent Systems

Notes: A very interesting paper, but not particularly relevant for system builders, more for AI architecture research.

browser-use / browser-harness: Self-Healing Harness That Enables LLMs to Complete Any Task

Notes: Another tool for chrome browser support for agents.

h4ckf0r0day / obscura: Headless Browser for AI Agents and Web Scraping

Notes: Yet another new tool for browser support. This one takes a different approach to make a completely headless experience.

dirac-run / dirac: Coding Agent Singularly Focused Efficiency and Context Curation

Notes: Neat Uses ASTs and Hash Anchored Edits. I would expect other popular coding agents to support these pretty soon.

shanraisshan / claude-code-best-practice

Notes: Neat repository (not mine) collecting much of the Claude Code plugins and workflows I shared in this newsletter over the past few months.