Skip to content Skip to sidebar Skip to footer
Let's Talk
Let's Talk
Close

A case study of self-hosted, locally running coding LLM chatbots

Locally running LLMs, Agentic AI, AI tools, Claude Code, Coding models, CUDA, GPU, Large Language Models, Qwen 2.5, Roo CodeAugust 8, 2025676Views0Likes0Comments
Explore flexible AI-coding workflows without vendor lock-in. This hands-on deep-dive weighs options such as bargain GPU clouds – Runpod, TensorDock, Kaggle T4s, and Paperspace – and then illustrates our experiment in self-hosting a Qwen 2.5 Coder on a single RTX 4080. We lay out where dollars, VRAM, and latency stack up, touch on how quantization…

Read More

AI Toolset Studies: Developer experience, usefulness & output code-quality evaluation

Agentic AI, AI tools, Analysis, Claude Code, Claude Sonnet, Creative, Figma, Generative AI, MCP Servers, Prompt Engineering, Software EngineeringJuly 24, 2025788Views0Likes4Comments
This report dives into real-world experiments evaluating the developer experience, usefulness, and code quality output of modern AI tools. Focusing on Claude Code and Figma MCP integration in a React Native project, the study uncovers practical insights, highlighting both the promise and quirks of using agentic AI for frontend tasks. From managing language selection screens…

Read More

The power developer framework from Project Reboot.

OpenRouter experiment Phase 1: Power Users – How Our Top AI Developers Really Work

Cohort Analysis, Agentic AI, AI tools, Claude Code, Claude Opus, Claude Sonnet, CLine, Context Engineering, Gemini 2.5, GitHub CoPilot, Large Language Models, OpenRouter, OpenRouter, Prompt Engineering, Research and Development, Roo Code, Software EngineeringJuly 8, 20253KViews0Likes3Comments
What happens when engineers get total freedom to pick and choose their own tools from a credits-based, model-agnostic AI stack? In Project Reboot’s latest experiment, we allocated OpenRouter credits – matching the value of a monthly Copilot seat – to a randomly selected group of participants, with Copilot left in place as a safety net.…

Read More

Mapping the AI terrain - A Deeper Dive with Cohort-based Developer AI Usage Analysis

Mapping the AI Terrain: A Deeper Dive with Cohort-based Developer AI usage analysis

Research and Development, Agentic AI, AI, Analysis, Cohort Analysis, Large Language ModelsJune 23, 2025622Views0Likes2Comments
In our first post, we shared high-level insights from our AI usage survey across the engineering department at Shuru. But averages can be misleading. To truly understand how AI is impacting developer workflows, we’re diving deeper — by slicing the data into cohorts. This post outlines our approach to cohort-based analysis, how we plan to…

Read More

Project reboot - a new default

Project reboot – A new default

AI, Agentic AI, AI tools, Analysis, Creative, Generative AI, Large Language Models, Research and Development, Software EngineeringJune 16, 20252KViews0Likes3Comments
Project Reboot – A new default is Shuru’s internal initiative to reshape how teams engage with AI in software development. As AI tools become ubiquitous, their adoption across teams remains uneven. With this series, we’re not just trying to change that – we’re committing to doing it in public. From tool usage data across our…

Read More