@terran0925

Group Manager - Solution Architecture https://xitter.cf/t.co/OkhW8SBrbK

 Toronto
terenceluk.blogspot.com
Joined May 2010
  • Tweets1.4K
  • Following134
  • Followers436
What is Multi-Token Prediction (MTP) and How Does It Compare to a Draft Model? blog.terenceluk.com/2026/10/…

Multi-Token Prediction (MTP) vs Draft Models on LM Studio

Multi-token prediction (MTP) explained, with LM Studio tests on a ZBook comparing Qwen3.5 and Qwen3.6 MTP builds against a draft model.

blog.terenceluk.com
26
Running Jev-Style Decision Models Locally with llama.cpp: Laya and OpenJev blog.terenceluk.com/2026/10/…

llama.cpp Decision Models: Testing Laya and OpenJev

Run llama.cpp decision models like Laya and OpenJev locally. Step-by-step setup, PowerShell requests, and what I learned testing both on Windows.

blog.terenceluk.com
56
Turning a Purpose-Built Claude Code Agent Roster Into a Generic One That Builds Any Application blog.terenceluk.com/2026/10/…

Claude Code Agent Roster: A Generic Version That Builds Any Application

How I turned a purpose-built Claude Code Agent Roster into a generic one that builds any application, with real guardrails and specs.

blog.terenceluk.com
1
91
Test-Driving TypeSafe AI’s Jev, the System One Model blog.terenceluk.com/2026/09/…

TypeSafe Jev: The System One Model

TypeSafe Jev returns typed decisions with calibrated probabilities. I test-drive it through Vercel AI Gateway to triage Azure Monitor alerts.

blog.terenceluk.com
2
86
What is Speculative Decoding and How Does a Draft Model Work? blog.terenceluk.com/2026/09/…

Speculative Decoding: 2 Draft Model Pairings Tested

Speculative decoding uses a small draft model to predict tokens ahead of time, cutting LLM inference latency. Learn how it works and why it matters.

blog.terenceluk.com
48
Pointing LM Studio Bionic at a Codebase With Defects I Documented in a Previous Post blog.terenceluk.com/2026/09/…

LM Studio Bionic Graded Against Eight Known Defects in a Real Repo

When I sat down to test the coding functionality of LM Studio Bionic, I realized I already had a ready-made benchmark so rather than asking a model to build

blog.terenceluk.com
79
What Claude’s Memory Actually Is (And Why the Model Itself Has None) blog.terenceluk.com/2026/08/…

Claude Memory: Truths About How It Really Works

Claude Memory is not the model remembering you, it is context resent on every call. See how MEMORY.md, Claude Code, and Claude.ai actually simulate it.

blog.terenceluk.com
72
Test driving LM Studio Bionic – the next generation of LM Studio designed for agentic work with open models blog.terenceluk.com/2026/08/…

LM Studio Bionic: Testing the Agentic Successor to LM Studio

Testing LM Studio Bionic: document grounding, a self-referential style audit, and a web search gate that turned out to be a setting, not a bug.

blog.terenceluk.com
1
133
The Gap Between Best (Frontier Models) and Useful (Local LLM) Is Shrinking blog.terenceluk.com/2026/08/…

Local LLMs: The gap from Frontier is quickly closing

Why the Future of Local LLMs Doesn't Depend on Being the Best

blog.terenceluk.com
77
Subagent Relay vs. Agent Teams in Claude Code: Building the Same Login Portal Twice With the Same Five Agents blog.terenceluk.com/2026/07/…
55
Building an AI News Digest with Claude Code: CLAUDE.md, Skills, Tools, and Workflows in One Project blog.terenceluk.com/2026/07/…

AI News Digest: CLAUDE.md, Skills, Tools and Workflows

Building an AI News Digest with Claude Code: a CLAUDE.md file, a skill, built-in tools, and a parallel workflow that researches AI news automatically.

blog.terenceluk.com
1
143
Building an Azure Function App for Network Connectivity Testing Across VNets and Hybrid Environments blog.terenceluk.com/2026/07/…

Network Connectivity Testing with an Azure Function App

Learn how to build an Azure Function App for network connectivity testing across VNets and hybrid environments using diagnostic REST APIs.

blog.terenceluk.com
1
170
Vibe Coding a Local Browser Agent with GitHub Copilot blog.terenceluk.com/2026/05/…

Local Browser Agent: Vibe Coding with GitHub Copilot

Learn how I built a local browser agent using GitHub Copilot and a local LLM to browse the web, click, and complete tasks automatically.

blog.terenceluk.com
1
2
331
Building an AI-Powered Email Draft Reply Generator with Power Automate and Claude blog.terenceluk.com/2026/05/…

Email Draft Reply Generator with Power Automate and Claude

See how I built an email draft reply generator using Power Automate and Claude to draft smart, ready-to-review replies automatically.

blog.terenceluk.com
42
Vibe Coding a Stock Dashboard with GitHub Copilot blog.terenceluk.com/2026/05/…

Stock Dashboard: Vibe Coding with GitHub Copilot

This is the story of how I vibe coded a personal stock dashboard with GitHub Copilot over a single Sunday afternoon.

blog.terenceluk.com
43
Why the X-Forwarded-Host Header Matters When Running App Service Authentication Behind Application Gateway blog.terenceluk.com/2026/05/…
30
How to Publish an Authenticated Azure App Service Through Application Gateway Without Breaking Health Probes blog.terenceluk.com/2026/04/…

App Service Health Probe: Publishing Through App Gateway

Fix Application Gateway 502 errors: how an App Service Health Probe breaks when App Service Authentication is enabled, and the file-based fix that solves it.

blog.terenceluk.com
41
Vibe Coding a Smart Travel Planner with GitHub Copilot blog.terenceluk.com/2026/04/…
46
How to Configure a Logic App that adds Users to a Distribution List when native “Add user to group” would fail blog.terenceluk.com/2026/04/…

Logic App Distribution List: When Add User to Group Fails

How to work around Microsoft Graph and Exchange limits with a Logic App Distribution List flow that adds mailbox-less users once licensing catches up.

blog.terenceluk.com
39
Using a locally ran LLM to power Visual Studio Code Insiders Chat Feature + Creating a Space Invaders Game blog.terenceluk.com/2026/03/…

Local LLM in VS Code Insiders Chat: Space Invaders Game

How to power the VS Code Insiders chat feature with a Local LLM VS Code Insiders setup in LM Studio, then use it to build a Space Invaders game.

blog.terenceluk.com
32
Load more