Shout out to @stevekrenzel of @withlogicinc for sponsoring !!

By the end of 2025, most new code will be written by large language models. Is that the end of programming as a career? https://t.co/loqipUihRZ

New post: What comes in the way of great recruiting? Over the years, I’ve concluded that similar to the power law of startups and founders, there's a power law of recruiting. In every team, a few players disproportionately move the needle. If we agree that recruiting these 10x

What if code wasn’t written—but generated and executed on demand? 🤯 In our latest “One-Shot” episode, @stevekrenzel explores the rise of disposable software, where AI dynamically creates and runs code, eliminating traditional static structures. Optimizing for LLM-driven

You can get smarter programmers to work on a project using AI-first development practices than you can on a traditional programming project. https://t.co/aUm1aUaydU

Like everyone else in tech, been trying to obsessively digest all the news and takes this weekend around Deepseek R1. Here are some thoughts and questions swirling in my mind regarding R1 and what it means for the US AI landscape:

Yule LOGIC Day 11 - Dive into how temperature and top-p influence LLM outputs, balancing creativity and coherence for different applications.

https://t.co/k66Cr92zzx

Happy New Year everyone! Our last day of Yule LOGIC is all about controlling LLMs with their temperature and top-p parameters. Stay warm (or cold?) out there!

Yule LOGIC Day 10 - Learn how human-friendly inputs and outputs improve LLM performance, with practical tips for crafting clear, structured instructions and generating reliable results.

I give a lot of people prompt engineering advice, and if I had to summarize 90% of it in one suggestion: Human-friendly is LLM-friendly. @markgolazeski writes about it today https://t.co/NpOfOz14pm

Yule LOGIC Day 9 - Discover how few-shot knowledge distillation bridges the gap between large and smaller LLMs, enabling efficient performance with minimal data.

https://t.co/BVM2kiQh1y

We do some magical things at LOGIC, like getting gpt-4o-mini to perform as well as gpt-4o. Read how:

Yule LOGIC Day 8 - Learn how structured output with LLMs ensures reliable, schema-compliant data, bridging the gap between AI and traditional software systems.

Structured output is involved in every single LLM call we make. It'd be substantially more challenging to build reliable LLM systems without it. Today's Yule LOGIC post is all about it: https://t.co/Guwrb3P7wk

Why do we need embeddings AND rerankers? What do they even do? Yule LOGIC Day 8: https://t.co/fIKKOfDv5S

Have you gotten to play around with Claude's computer use yet? Today's Yule LOGIC post walks through setting it up locally and getting it going. It's the perfect thing to tinker with this weekend. https://t.co/ivPy5ajEOl

Hope everyone had a Merry Christmas! Today's Yule LOGIC post is all about turning a plain ol' business document (a runbook) into a functioning team member: https://t.co/BYijtctqS4

Yule LOGIC Day 5 experiments with transforming your runbooks into Slack agents that guide you through incidents step-by-step—no flipping through pages, just actionable answers when you need them most. https://t.co/fswt8Ql5FY