John Berryman

Context Engineering Requires AI Empathy Transform your AI applications by thinking like an AI intern on their first day. This post reveals how empathy-driven context engineering leads to better LLM performance, covering everything from hand...

𝐓𝐡𝐢𝐧𝐤 𝐨𝐟 𝐲𝐨𝐮𝐫 𝐋𝐋𝐌 𝐚𝐬 𝐚𝐧 𝐢𝐧𝐭𝐞𝐫𝐧 𝐨𝐧 𝐭𝐡𝐞𝐢𝐫 𝐟𝐢𝐫𝐬𝐭 𝐝𝐚𝐲.

The AI intern is:

- 𝐁𝐫𝐢𝐠𝐡𝐭 𝐛𝐮𝐭 𝐧𝐨𝐭 𝐩𝐬𝐲𝐜𝐡𝐢𝐜
- 𝐏𝐫𝐞𝐟𝐞𝐫𝐬 𝐟𝐚𝐦𝐢𝐥𝐢𝐚𝐫𝐢𝐭𝐲
- 𝐍𝐞𝐞𝐝𝐬 𝐭𝐢𝐦𝐞 𝐭𝐨 𝐭𝐡𝐢𝐧𝐤
- 𝐇𝐚𝐬 𝐚𝐭𝐭𝐞𝐧𝐭𝐢𝐨𝐧 𝐢𝐬𝐬𝐮𝐞𝐬
- 𝐂𝐚𝐧 𝐛𝐞 𝐚 𝐜𝐨𝐧𝐯𝐢𝐧𝐜𝐢𝐧𝐠 𝐥𝐢𝐚𝐫
- 𝐇𝐚𝐬 𝐮𝐧𝐮𝐬𝐮𝐚𝐥 𝐯𝐢𝐬𝐢𝐨𝐧
- 𝐂𝐚𝐧'𝐭 𝐟𝐨𝐫𝐦 𝐧𝐞𝐰 𝐦𝐞𝐦𝐨𝐫𝐢𝐞𝐬

Full blog post is here arcturus-labs.com/blog/2025/10...

I had an absolute blast talking about the "7 Deadly Sins of AI Applications" with Hugo Bowne-Anderson, and I'm really pumped to see it in text form now!

(But I do think that neither cartoon John or cartoon Hugo are as handsome as the real deal.)

hugobowne.substack.com/p/patterns-a...

AI Chat Startups - Is the bubble about to pop? Investor money chases chat-style interfaces for every domain. I believe this is wrongheaded: ChatGPT and pals will do your domain better as they address more and more long tail-use cases, and better i...

@jnbrymn.bsky.social and I are going to have a controversial discussion where maybe we don't agree

maven.com/p/3ad9c7/ai-...

In this talk, I'll explore the state-of-the-art in LLM evaluation, covering modern techniques like LLM-as-Judge. I’ll also discuss the essential processes and culture shifts that are required to launch reliable AI products and drive ongoing improvement.

Generative AI has made it easier than ever for companies to build products quickly. However, LLMs are inherently nondeterministic and unpredictable. Integrating LLMs into products demands an unprecedented level of quality assurance – requiring new strategies for continuous evaluation.

Cutting Through AI Hype with John Berryman (Part 2 of 2) · Zoom · Luma Join us for the second in a two-part series that goes beyond AI buzz and into practical insight. This event will feature a dynamic blend of presentation, group…

I'm super excited to present to 7CTOs tomorrow on "Wresting with AI Evaluations"

Register here: lu.ma/u9cp3exb

Supercharging LLM Classifications with Logprobs Turn your LLM into a precision instrument for classification – no fine-tuning required. This post shows how to go beyond simple

Then you can make a classifier that is tuned to whatever threshold matches a training set most accurately. Easy peasy.

Full post here: arcturus-labs.com/blog/2025/03...

The idea is really simple – basically you just make a classifier the returns a single token – good/bad or red/green/blue or 1/2/3/4 – But rather than looking at that token you look at the probability of the tokens.

You probably knew you could turn an LLM into a classifier. But the basic approach returns a hard classification – yes or no; good or bad.

In this post I'll show you a simple technique to make your LLM classifiers more nuanced – they tell you HOW good or HOW bad something is.

The Hidden Simplicity of GenAI Systems At first, GenAI seems neatly split into agents, RAG, and workflows, but up close, it’s messy. This talk clears the fog by uncovering the small set of principles behind all major patterns. Once you gra...

With the staggering amount of AI paradigms – Workflows, RAG, Agents – it seem impossible to get started learning AI and building your own projects.

But once you realize that all of the ideas are built upon the same simple foundations, you become free to let your ideas flow.

How Evals Made GitHub Copilot Happen GitHub Copilot stands as one of the first commercially successful generative AI applications. To make the product work, the copilot team had to invent evaluation methodologies with no existing bluepri...

This should be a really fun talk, Shawn Simister and I have been recalling to ourselves how it all worked. What we realized is that we were doing an early form of LLM-as-judge back in early 2023.

maven.com/p/da8264/how...

I'm looking forward to seeing you live-code a vector database. Lucky for me my only job for the call is to throw tomatoes. :D

MAJOR ANNOUNCEMENT

SoftwareDoug LLC has accepted $2K in seed funding from Turnbull Family Grocery Fund to bring you search/AI/RAG consulting, coaching, and training

Read the full deep-dive on why visual reasoning is the next frontier in AI:
arcturus-labs.com/blog/2025/03...

Follow for more.

Picture this: AI solving physics problems by visualizing objects in motion, or predicting social interactions by imagining body language and facial expressions. That's where we're headed.

The key insight: Just like chain-of-thought reasoning transformed how AI thinks through problems with words, visual reasoning will let AI work through problems by creating and analyzing sequences of images.

Today's AI can put a detective hat on your cat. Tomorrow's AI will help design your garden, rearrange your furniture, and solve complex spatial puzzles by actually visualizing different scenarios and their outcomes.

OpenAI just revolutionized image generation, but that's just the beginning. The real game-changer? Visual reasoning is coming - where AI won't just manipulate images, but will actually think through problems using visual simulation.

@merlehazard.bsky.social well hi! Any new economic ballads recently? Surely you've no lack of inspiration right now. There's got to be some sort of ballad related to tarrifs.

Exciting times ahead! As we enter a new golden age of creation with LLMs, I dive deep into RAG, AI-assisted coding, and the future of dev tools in my latest podcast with
@dmitrykan.bsky.social for Vector Podcast

👋 Want to know how to build AI agency into your application? Do it incrementally.

Vlog here:
youtube.com/watch?v=f5qW... and original post in video description. 🤙

Thanks for reading our book and posting about it. I'm glad you got something from it!

Building Reliable AI Apps? Start by Firing Yourself.

Don't dive straight into complex AI agents. Start manual, document processes, then systematically automate each task.

Full breakdown + case study:
arcturus-labs.com/blog/2025/01...

A perfect reply. Thank you Evgeny.

I think I'll try this once a year until someone bites :)

I was hoping that someone would fall into my trap actually :P – look up Cunningham's Law

Turnbull's law: "The best way to get the right answer on the Internet is not to ask a question; it's to post the wrong answer." I love it! :D

Prompt Engineering for LLMs: The Art and Science of Building Large Language Model–Based Applications Prompt Engineering for LLMs: The Art and Science of Building Large Language Model–Based Applications [Berryman, John, Ziegler, Albert] on Amazon.com. *FREE* shipping on qualifying offers. Prompt Engineering for LLMs: The Art and Science of Building Large Language Model–Based Applications

amzn.to/4fMj2Ef

Home for the holidays means cozy blankets, the scent of pine, and a tech book in the stocking. Prompt Engineering for LLMs—the perfect gift for the developer who finds magic in code and creativity in AI.

How AI Is Built Podcast · Nicolay Gerold · How AI is Built dives into the different building blocks necessary to develop AI applications: how they work, how you can get started, and how you can master them. Build on ...

I recorded 2 podcasts this week all about LLM application development. My recordings will be out soon, but until then make sure to check out these great programs:

How AI is Built w/ @nicolaygerold open.spotify.com/show/3hhSTyH...

Vector Podcast w/ @DmitryKan open.spotify.com/show/13JO3vh...