Artificial Intelligence
How We Use Claude Code at Serokell
AI agents may know Haskell, Git, and CI, but they do not know how your team actually works. At Serokell, we turned our internal engineering guidelines into Claude Code skills so agents could follow the same conventions as our engineers. This post covers how those skills became a reusable plugin and what we learned from testing it on real repositories.
Upstream posts
Building a RAG Investment Assistant
For VC teams, valuable knowledge is everywhere: PDFs, inboxes, call transcripts, news feeds, and internal notes. The problem is getting the right context when you actually need it.
We built a platform that brings all of this information into a single searchable knowledge base, with LLM agents on top that can:
• prepare meeting briefings
• research people and companies
• answer questions across the firm’s internal knowledge
• continuously ingest and structure new information
In the article, we go under the hood and explain how the system works, including its storage architecture, seven-stage ingestion pipeline, Airflow automations, and the agent layer that we substantially redesigned after the first production release.
We also cover a production incident that changed how we approached monitoring and reliability.
If you’re building RAG systems, internal AI assistants, or agentic workflows on top of large amounts of unstructured company data, this one gets pretty technical.
Claude Code: From Agent to Useful Tool
The main point of the article is that Claude Code needs engineering around it just as much as any other development tool. On its own, the agent can write code and run commands, but it quickly runs into the usual problems: missing context, repeated mistakes, unsafe decisions, and fragile fixes. A better setup gives it project rules, current documentation, MCP access, reusable skills, hooks, task trackers, Git safety nets, and clear session workflows. With these pieces in place, Claude Code can handle implementation, investigation, and verification more reliably, while the developer still owns architecture, review, and product judgment. The result is not an autonomous engineer, but a practical tool that works best when its context and workflow are engineered deliberately.
Article by Kirill Dukhanin
June 22nd, 2026
19 min read
Beyond the Hype: Crossing the GenAI Divide in Real-World Business
If you judged the state of AI in business by your LinkedIn feed, you’d think the revolution already happened..
Design Patterns for Long-Term Memory in LLM-Powered Architectures
The explosive growth of large language models (LLMs) has reshaped the AI landscape. Yet their core design is still fundamentally stateless: a drawback often referred to as “conversational amnesia.” …
The Real Limits of AI Agents in 2025
TL;DR: Everyone says 2025 is the year of autonomous AI agents. We’ve built a lot of them in production, and that’s exactly why we think most of the current hype just doesn’t add up. In this post, we'll break down the most common misconceptions, talk about what actually works in the real world, and explain why the math and economics behind the hype don’t hold up yet.
A Bit Late but Ultimate Analysis: DeepSeek
tldr: if you want to find out everything about deepseek you can read this blogpost that has 3 separate sections with increasing technical details and difficulty from one to another, so you can stop …
Editor’s pick
A Guide to Kalman Filter
In this article, we’ll make a few notes on how Kalman filter works, and will show how to overcome most common problems that might appear while using it.
Upstream posts
How to Evaluate an ML Model Performance?
Every data scientist knows that one of the most exciting parts of machine learning is picking the right algorithm to crack a problem. But how do you make sure that your chosen algorithm actually delivers the results you're aiming for? In this article, we discuss several methods allowing you to measure the performance of your ML model.
Top 11 AI Tools for Developers
The rapid advancement of AI has transformed the way developers approach software creation, problem-solving, and collaboration. AI tools are no longer just experimental—they’ve become essential components of modern development environments. This article will be especially useful for developers who want to automate repetitive tasks, enhance code accuracy, and improve project management. We offer an overview of some of the best AI tools currently available, including pricing and key features, so you can choose the right tools to fit your needs and projects.
What Is AutoML?
Today it is hard to imagine a business that doesn’t use machine learning in analytics and decision-making. However, the development and deployment of ML models can be very complicated and resource-consuming. Even experienced data scientists and ML engineers could benefit from effective automation. AutoML is a way to automate machine learning in your enterprise. In this article, we will talk about the benefits of AutoML, its operating principles, and ways to use it in your business.
How Top Companies Use AI
In the last few years, artificial intelligence has become an omnipresent technology. Many businesses, including world leading companies, apply AI for various purposes: business intelligence, performance analytics, enhancing user experience, optimization, and more. In this article, we will look at 9 companies that use artificial intelligence. Hopefully, they can inspire you and give some suggestions about how you can use AI creatively in your business.
A Comprehensive Guide to BERT
In this article, we’ll explore how Bidirectional Encoder Representations from Transformers (BERT) works, and why it’s called a universal tool for solving dozens of natural language processing tasks.
LLMs for Business: Fine-Tuning and Risks
This is the second article in our series “LLMs for Business.” In it, we explore how to fine-tune large language models and use them safely so they don’t harm your business.