Independent AI media & tools
Practical AI,
tested hands-on.
An independent project exploring AI through hands-on experiments, plain-language writing and small tools.
Field notes from the workbenchTiG /
The newsletter
Selected from the newsletter
01 /Notes on local models, prompting, token efficiency and agents.

AI workflows · Tech in GeneralA practical walkthrough of building three Claude Code mods to track turn duration, tool calls and command failures.
Read on Substack: I Built Mods for Claude. Here’s What Happened (opens a new tab)By Hari Krishna · 
Prompting · Tech in GeneralExplore 15 prompting patterns, how they relate to each other, and how to choose one for a task.
Read on Substack: Prompt Engineering Patterns: Complete guide 2026 (opens a new tab)By Hari Krishna · 
LLM architecture · Tech in GeneralA guide to the architectural choices behind modern open models, including attention, positional encoding, inference efficiency and training.
Read on Substack: Modern Open-Source LLM Architecture Explained: Transformer, RoPE, SwiGLU, GQA, GRPO and Beyond (opens a new tab)By Hari Krishna ·
Learn
Make sense of the technology.
03 /Guides and practical experiments from Tech in General on language models, prompting and AI workflows.
Read the full articles on our Substack. Article links open in a new tab.
AaLLM architecture · Tech in GeneralA guide to the architectural choices behind modern open models, including attention, positional encoding, inference efficiency and training.
Read on Substack: Modern Open-Source LLM Architecture Explained: Transformer, RoPE, SwiGLU, GQA, GRPO and Beyond (opens a new tab)By Hari Krishna ·
>_Prompting · Tech in GeneralExplore 15 prompting patterns, how they relate to each other, and how to choose one for a task.
Read on Substack: Prompt Engineering Patterns: Complete guide 2026 (opens a new tab)By Hari Krishna ·
↗AI workflows · Tech in GeneralA practical walkthrough of building three Claude Code mods to track turn duration, tool calls and command failures.
Read on Substack: I Built Mods for Claude. Here’s What Happened (opens a new tab)By Hari Krishna ·
How we work
From experiment to explanation.
04 /01
Test it hands-on
We try the models, prompts and tools ourselves.
Find a tool to try →02
Check the results
A human checks the evidence and the final output.
Send feedback →03
Share what works
We turn what we learn into plain-language notes.
Read the newsletter ↗
About
An independent point of view.
Tech in General is an independent project covering practical AI through a newsletter, small tools and educational content.
We use the Claude API and other AI tools for research and drafting, with a human reviewing what we publish.