Engineering Blog

Technical insights from Grid Dynamics engineers

Causal Inference in Practice: Geo-Experiments vs. Synthetic Control

Causal Inference in Practice: Geo-Experiments vs. Synthetic Control

By Djordje Grozdic · Aug 12, 2026

How to measure causal effects when A/B testing is impossible: geo-experiments vs. Synthetic Control, with a real retail pricing pilot.

Agentic Harness Engineering: Taming Autonomous AI in Production

Agentic Harness Engineering: Taming Autonomous AI in Production

By Riya Kumari · Jul 10, 2026

Raw autonomous agents on production networks are ungoverned root access. Agentic Harness Engineering isolates the reasoning loop from your infra.

Terraform Drift Detection: When Auto-Sync Is the Wrong Fix

Terraform Drift Detection: When Auto-Sync Is the Wrong Fix

By Dharshan Madhavan · Jul 2, 2026

Auto-generated drift PRs launder emergency console changes into your source of truth. A triage model for which drift to codify, revert, or ignore.

AI Isn't Just ChatGPT, Claude, Gemini — Here's the Rest of the Map

AI Isn't Just ChatGPT, Claude, Gemini — Here's the Rest of the Map

By Deepak Bhuvaneswaran · Jul 1, 2026

ChatGPT is one branch of AI. Here's the rest of the map — four types that fail four different ways, and one bug behind all of them: context.

Staged Firmware Rollouts at the Edge: Designing for Rollback

Staged Firmware Rollouts at the Edge: Designing for Rollback

By Gokul Ganesh · Jun 28, 2026

Why edge firmware rollouts fail after passing every gate — and how dual-bank partitioning plus staged telemetry limits the blast radius.

How Quality Engineering Must Change as AI Stops Being Deterministic

How Quality Engineering Must Change as AI Stops Being Deterministic

By Sanjay Kumar · Jun 24, 2026

Testing rule-based AI verifies logic. Testing ML and generative AI audits behavior — here's how QE practice must change at each stage.

Beyond Vibe Coding: Objective-Validation Protocols for AI Code

Beyond Vibe Coding: Objective-Validation Protocols for AI Code

By Vishnuram G · Jun 22, 2026

From vibe coding to adversarial multi-agent TDD: a three-level framework for validating AI-generated code before it reaches human review.

Monitor your agent with Langfuse

Monitor your agent with Langfuse

By Adam Wrobel · May 12, 2026

Trajectory observability for coding agents: which skills and tools actually run, in what order, and at what cost—wired to Langfuse from Claude Code.

From REST to WebRTC: Choosing the Right Protocol for Voice Apps

From REST to WebRTC: Choosing the Right Protocol for Voice Apps

By Djordje Grozdic · May 4, 2026

A comprehensive guide to selecting communication protocols for real-time audio, comparing WebRTC, WebSocket, REST/HTTP, gRPC, SIP, etc. for optimal performance and latency

Beyond ReAct: Advanced Reasoning Patterns for Production AI Agents

Beyond ReAct: Advanced Reasoning Patterns for Production AI Agents

By Tigran Apoyan · Apr 21, 2026

From self-reflection to meta-cognition—why foundational patterns fail at enterprise scale and what to use instead

Vasa Syndrome: We've Been Sinking the Same Ship for 400 Years

Vasa Syndrome: We've Been Sinking the Same Ship for 400 Years

By Sviatoslav Shcherbatiuk · Apr 16, 2026

The article is about bad product/delivery decisions wearing “AI digital transformation” as a costume

Design Is a Coding Job Now. Adapt, or Fade Out.

Design Is a Coding Job Now. Adapt, or Fade Out.

By Anastasiia Voievidka · Apr 3, 2026

AI is reshaping design — not by replacing designers, but by exposing every gap in how teams work. The designers who don't shift toward coding with AI won't survive the transformation. Here's what's actually changing, and what to do about it.

From Failure to Skill: Designing a Universal /learn Command for Coding Agents

From Failure to Skill: Designing a Universal /learn Command for Coding Agents

By G. Montenegro · Mar 27, 2026

How to implement a portable /learn command that turns runtime failures into reusable skill updates or new skills.

Methodology as a Discipline in AI-Assisted Development: an OOA Example

Methodology as a Discipline in AI-Assisted Development: an OOA Example

By Aleksandr Kondratskii · Mar 22, 2026

How methodology can make intermediate reasoning more visible, reviewable, and reusable in AI-driven workflows.

TDD in Agentic Workflows: Specs as Executable Guardrails

TDD in Agentic Workflows: Specs as Executable Guardrails

By Adam Wrobel · Mar 22, 2026

Test-Driven Development becomes practical when delegated to AI agents. Tests written before implementation act as executable specs that keep agents on track.

AI Workflow Frameworks for Software Development: A Layered Comparison

AI Workflow Frameworks for Software Development: A Layered Comparison

By Aleksandr Kondratskii · Mar 18, 2026

A practical taxonomy of AI software-development workflows that compares methodologies, control layers, skills systems, host-native agent capabilities, and integrated products.

Persistent Memory for AI Agents: Mem0 vs. Vertex AI Memory Bank

Persistent Memory for AI Agents: Mem0 vs. Vertex AI Memory Bank

By Andrei Kopeikin · Mar 2, 2026

A practical comparison of Mem0 and Vertex AI Memory Bank — two frameworks for giving AI agents persistent memory across sessions.

A Practical Commit Standardization Pattern for Teams and AI Coding Tools

A Practical Commit Standardization Pattern for Teams and AI Coding Tools

By Sahil Satralkar · Feb 26, 2026

A practical guide to standardizing commit messages using a shell script and documenting the workflow across AI coding tool instruction files.

Token Cache - a practical introduction

Token Cache - a practical introduction

By Adam Wrobel · Jan 12, 2026

In this post we cover the definition of the KV cache, benefits and use cases, and examples of code for major LLM providers

Building a Bitemporal Index (part 3): Storage

Building a Bitemporal Index (part 3): Storage

By James Henderson · Jan 8, 2026

How XTDB avoids scanning all history for current-state queries by using HAMTs, LSM trees, and a 'recency' heuristic to index bitemporal events in remote object storage.

Building a Bitemporal Index (part 2): "Bitemporal Resolution"

Building a Bitemporal Index (part 2): "Bitemporal Resolution"

By James Henderson · Jan 7, 2026

How XTDB avoids the storage amplification and complexity of traditional bitemporal approaches by reconstructing history on-the-fly, playing events backwards through system-time.

No sci-fi, no philosophy: Capabilities and Limits of Neural Networks

No sci-fi, no philosophy: Capabilities and Limits of Neural Networks

By G. Montenegro · Dec 24, 2025

Understanding neural networks as function approximators and their inherent limits.

Building a Bitemporal Index (part 1): A Taxonomy of Bitemporal Data

Building a Bitemporal Index (part 1): A Taxonomy of Bitemporal Data

By James Henderson · Nov 25, 2025

A technical deep-dive trilogy explaining how a bitemporal-aware database can handle and optimise for temporal data differently to general-purpose databases.

Offline AI Agent: OpenCode + LM Studio

Offline AI Agent: OpenCode + LM Studio

By Aleksandr Kondratskii · Nov 22, 2025

Guide to running an offline AI coding agent using OpenCode with LM Studio and the Devstral model.

Self-Service Analytics: How I Automated Daily Usage Reports in Under 3 Hours

Self-Service Analytics: How I Automated Daily Usage Reports in Under 3 Hours

By Volodymyr Voyevidka · Nov 18, 2025

Building production automation for Cerebra's usage analytics using N8N and Gemini—without asking anyone for help

Comparative practical analysis of Agentic AI Frameworks capabilities in 2025

Comparative practical analysis of Agentic AI Frameworks capabilities in 2025

By Egor Borisov & Pavel Marchenko · Nov 6, 2025

Overview of modern agentic frameworks and comprehensive comparison

A Tensor-Based Approach to Product Configuration

A Tensor-Based Approach to Product Configuration

By Illia Krokhmalov · Oct 18, 2025

Revised version incorporating co-innovation and business impact emphasis

Tools for LLMs step by step

Tools for LLMs step by step

By G. Montenegro · Sep 29, 2025

Step‑by‑step guide on how system prompts define tool usage in large language models, with three difficulty levels covering basic calls, streaming inference, and token‑level fine‑tuning.

Determinism in LLMs

Determinism in LLMs

By G. Montenegro · Sep 19, 2025

How to improve reproducibility in large language models.

From Zero to Agentic AI: My Journey with the Jetson Orin Nano Super Developer Kit

From Zero to Agentic AI: My Journey with the Jetson Orin Nano Super Developer Kit

By Ihor Syrov · Sep 14, 2025

A software engineer's story of building an autonomous AI agent that thinks, sees, and acts - all running on a device smaller than a smartphone

The idea of AGI is broken, hi SROW

The idea of AGI is broken, hi SROW

By G. Montenegro · Sep 13, 2025

SROW(Same Result in its Own Way) is the real way to measure AI impact

Running LLMs Locally: Read This Before You Try at Home

Running LLMs Locally: Read This Before You Try at Home

By G. Montenegro · Aug 31, 2025

A basic practical guide of concepts to understand open language models.

Sketching the New Coding Paradigm from the Roots

Sketching the New Coding Paradigm from the Roots

By G. Montenegro · Aug 27, 2025

A brief sketch of the basic rules for development in the world of AI agents

Agent-to-Agent (A2A) vs Model Context Protocol (MCP) for the Agentic Platform

Agent-to-Agent (A2A) vs Model Context Protocol (MCP) for the Agentic Platform

By Maciej Blaszke · Jul 30, 2025

A Technical Analysis and Comparison of the Agent-to-Agent (A2A) and Model Context Protocol (MCP)

WebSocket vs Server Sent Events

WebSocket vs Server Sent Events

By Maciej Blaszke · Jul 30, 2025

A Technical Analysis and Comparison of the WebSocets and Server Send Events (SSE) in context of Agentic Platforms