What happens internally after an OpenAI endpoint receives a user prompt?

I wasn't comfortable with treating AI models as magical black boxes, so I made some sequence diagrams that I'm sharing with you in this post:

milendyankov.com/blog/2026/04...

We used to glorify the app servers?
The WebSpheres, WebLogics, JBosses, Zopes, Zends, ...

We’re doing it again with LLMs.

IMHO "doing one thing well" phase is next, so I'm coining a term now - μAI (Micro-Intelligence)

Remember you heard it here first when it becomes viral in a few years 😜

Anatomy of a Model - the Developer Perspective In the old days, every IT organization had a dedicated, almost sacred role: the DBA (Database Administrator), often informally known as the "Gatekeeper of the Schema." These individuals ensured that t...

The "Anatomy of a Model - the Developer Perspective" blog post from the "AI for Application Developers" series took longer to write and went way beyond my notes. But it was fun, and I even learned new things. I hope you will, too.

milendyankov.com/blog/2026/03... #AIforDevs

A thought of the day: Did AI kill DRY 🤔?

A Trip in the AI/ML Model Formats Jungle As developers, we can immediately recognize the structure of a .csv file and how it differs from a .md one. Want to experience the frustration of someone completely lost when faced with those extensio...

Wrote sort of a tiny survival guide for the AI model formats jungle. If you are looking to run AI/ML models locally or just curious how they are distributed, have a read and let me know if it helps.

milendyankov.com/blog/2026/03...

AI/ML Models Are Not Libraries The first time I wanted to use a model in my own application, I expected the experience to be similar to using a library or a framework. It seemed Hugging Face was the Maven Central or the npm (or the...

Entering the AI world, I expected a familiar landscape of libraries and APIs. I found "serialized Python" and a massive gap between my code and the math.

Let me save you some disappointments if you are on the same track: milendyankov.com/blog/2026/03...

AI for Application Developers

AI is currently polarized between academic math and "Look ma!" demos.

I feel that we, the app developers, lack the middle ground. We need to understand the abstraction layers below the magical API calls.

I’ll be sharing my notes on the subject as blog posts: milendyankov.com/blog/2026/03...

Never ever in my decades-long career have I been so frustrated by the phrase 'You are absolutely right".

Finite and Infinite Startups Three weeks ago, Adam Wathan, the creator of Tailwind CSS, released a podcast containing an honest confession – "We had six months left" – and an explanation of why they had to let some folks go. For ...

I've spent years in companies calling themselves startups. It took me a while to realize the label doesn't imply sameness. It took me even longer to self-diagnose the career friction I experienced. The recent Tailwind story triggered me to write something about it: milendyankov.com/blog/2026/01...

Multi-Hat Disorder - Building Personal Website Like an Enterprise App While I’ve been focused on helping companies grow, my personal brand has taken a hit. I’ve never been interested in being an influencer or a showman, but I’ve come to realize that technical expertise ...

Somehow, revamping my personal website felt awfully similar to designing enterprise software systems. Different scale, same existential dread over choosing the wrong framework. If you find yourself constantly asking "Am I doing this right?", this may help (or not): milendyankov.com/blog/2025/01...

Exact Instructions Challenge - THIS is why my kids hate me. | Josh Darnit YouTube video by Josh Darnit

Do you remember this viral video from about a decade ago?

👉 www.youtube.com/watch?v=cDA3...

It was satire, not a design spec for AI agents.

For me, AI solves one of the two hardest problems in software development - naming things. It still can't really help much with cache invalidation and off-by-one errors 😜

Thank you! I'm happy and grateful I could present it at @DevBcn

Milen Dyankov - AI for Java Developers - From Buzzword to Code Speaker Milen Dyankov Abstract AI is undeniably a buzzword, but it is here to stay. What does this mean for Java developers? Can we design and build applications that harness the power of AI? Most importantly, where do we start? This session focuses on inference—using pre-trained models directly from your code. Instead of relying on proprietary APIs, it'll demystify the "magic" by breaking the AI buzzword down into familiar concepts and components. This will enable you to use AI models as naturally as you would work with databases or workflows. Through live demos, you'll learn to leverage basic embedding models, explore vector databases, and use tools like LangChain4J. By the end, you'll have a practical roadmap to kickstart your journey into Java-powered AI. https://www.devbcn.com

Catch Milen Dyankov's insightful session "AI for Java Developers - From Buzzword to Code" at #devbcn25! He'll dive deep into practical AI applications in Java - don't miss this chance to level up your skills. Check out the session recording: buff.ly/dNQRU2B #DevBcn

Milen Dyankov - AI for Java Developers - From Buzzword to Code YouTube video by DevBcn: The Barcelona Developers Conference

Curious how to start using #AI models from your #Java code? If you have an hour, here is the recording of my "AI for Java developers" talk at @devbcn.bsky.social

youtu.be/UjBUHdfUjVM?...

Let me know if it was helpful.

Once, only elites could read and write—they were the writers. Literacy spread, but few became authors. Writing well still takes skill and craft.

Software dev was once for specialists. Now, no-code lets almost anyone “write code.”

But just like literacy, producing code doesn’t make you a developer.

In tests - time. Definitely time.

I was really worried my talk is too basic for #devbcn25 Imagine my surprise when I saw the reviews. Thank you everyone who attended!

The only session I could attend (partially) is the one by @milendyankov.com and I have to say that I liked it very much !
Very well explained, all the concepts related to #AI, the models behind it, what are #LLM's,
@ollama @langchain4j, ... bravo! 👏 🙂

I can literally listen to @brianvermeer.nl all day long without ever getting bored.

From AI hype to hands-on code! @milendyankov will walk Java developers through real use cases in his #devbcn25 session: AI for Java Developers – From Buzzword to Code. Join us July 8–10 in Barcelona!
Details 👉 buff.ly/YpnATm9

Would you go for 2.10 or 2.9 when you need the latest version of a library/app?
Asking for my math teacher 😉

Debugger in baggy jeans You squash bugs in oversized hoodies with style. The stack trace fears you.

I am Debugger in baggy jeans. Unleash your inner Duke and win a limited-edition T-shirt! #Java30WithIntelliJIDEA www.jetbrains.com/lp/java-30/d...

School of Code students about Pinecone YouTube video by Pinecone

A good #VectorDB is a must-have for AI-powered solutions on top of massive datasets.

But that doesn’t mean this is an experts-only territory. With the right tools and guidance, even newcomers can use #RAG and add genuinely useful #AI to their projects.
www.youtube.com/watch?v=OSB0...

This is where I'll present "AI for Java developers" tomorrow. Today @piotrprz.bsky.social rocks the @jprimeconf.bsky.social scene!

Congratulations! Keep up the great work!

Careers | Pinecone Join us in building the database for machine learning. We're looking for talented individuals to help us push the boundaries of vector search technology.

Hey folks, just wanted to let you know we have a few open roles at Pinecone – specifically for Senior/Staff Software Engineers and a Support Engineer. If you or someone you know is looking for a new opportunity, check them out: www.pinecone.io/careers/?utm...

Tool up: Pinecone’s first MCP servers are here | Pinecone We are releasing our first three MCP servers: Pinecone Assistant MCP (remote), Pinecone Assistant MCP (local), and Pinecone Developer MCP (local). Together, the three support the following use cases: ...

I’ve been exploring the land of MCP servers recently. Meanwhile, my colleagues at Pinecone have been building the first official ones for Pinecone’s Database and Assistant. Now that they’re out, I wrote a blog about what those MCP servers are and how to use them: www.pinecone.io/blog/first-M...

The emerging vibe coding field has come up with a new productivity metric: kilos of code 😜