灯里/iku

I'm a curious person who really enjoys AI. I write for fun as a personal hobby. Getting your reactions and follows really motivates me. The AI era has its challenges, but it's super fun!

Joined on Sep 19, 2025

Work

AI Engineer (Provisional)

Cover image for Increase my familiarity with BASE64.

灯里/iku

Nov 16

Increase my familiarity with BASE64.

#api #tutorial

12 min read

Want to connect with 灯里/iku?

Create an account to connect with 灯里/iku. You can also sign in below to proceed if you already have an account.

Create Account

Already have an account? Sign in

Cover image for In 2025, AI regulations were established in four major global regions.

灯里/iku

Nov 12

In 2025, AI regulations were established in four major global regions.

#ai #aigovernance #datasovereignty #techpolicy

8 min read

Cover image for 【A Cynic's Guide】The Paradox of Selecting Diagramming Tools: Which Tool Transcends Technical Elegance and the "Approval Hurdle"?

灯里/iku

Nov 6

【A Cynic's Guide】The Paradox of Selecting Diagramming Tools: Which Tool Transcends Technical Elegance and the "Approval Hurdle"?

#discuss #diagrams #documentation #ai

7 min read

Cover image for RAG Architecture Design Theory and Conceptual Organization in the Age of AI Agents: 7 Patterns

灯里/iku

Oct 27

RAG Architecture Design Theory and Conceptual Organization in the Age of AI Agents: 7 Patterns

#rag #llm #architecture #agents

20 min read

灯里/iku

Oct 25

LLMs Learn from "Pseudoscientific Papers" Too - Quality Control for AI Developers

#ai #machinelearning #llm #developers

9 min read

Cover image for Analyzing the Best Diagramming Tools for the LLM Age Based on Token Efficiency

灯里/iku

Oct 21

Analyzing the Best Diagramming Tools for the LLM Age Based on Token Efficiency

#mermaid #tokenefficiency #devops #llm

7 min read

Cover image for AI Chatbot Developers: What's the "Other Safety" We Should Be Thinking About Now? User Protection.

灯里/iku

Oct 17

AI Chatbot Developers: What's the "Other Safety" We Should Be Thinking About Now? User Protection.

#ai #security #llm #safety

24 min read

Cover image for Google Opal is not a “degraded Dify”. Its strategic positioning and optimal utilisation methods revealed through actual use

灯里/iku

Oct 12

Google Opal is not a “degraded Dify”. Its strategic positioning and optimal utilisation methods revealed through actual use

#ai #google #nocode #productivity

10 min read

灯里/iku

Oct 4

"AI makes it cheaper, right?" - Is this happening in your country too?

3 min read

灯里/iku

Sep 30

Beyond YAML: Logic Compression for 50%+ LLM Cost & Latency Reduction

#promptengineering #llm #costreduction #tokenoptimization

18 min read

灯里/iku

Sep 20

When Your CEO Says 'Let's Use AI': A Technology Selection Survival Guide

#ai #career #management

13 min read

灯里/iku

Sep 19

The Era of Choosing RAG — Learning Cognitive Load and Architecture Design from GPT-5’s Failures

#ai #rag #machinelearning #architecture

13 min read

Forem

灯里/iku

Badges

Writing Debut

Currently learning

Available for

Increase my familiarity with BASE64.

Want to connect with 灯里/iku?

In 2025, AI regulations were established in four major global regions.

【A Cynic's Guide】The Paradox of Selecting Diagramming Tools: Which Tool Transcends Technical Elegance and the "Approval Hurdle"?

RAG Architecture Design Theory and Conceptual Organization in the Age of AI Agents: 7 Patterns

LLMs Learn from "Pseudoscientific Papers" Too - Quality Control for AI Developers

Analyzing the Best Diagramming Tools for the LLM Age Based on Token Efficiency

AI Chatbot Developers: What's the "Other Safety" We Should Be Thinking About Now? User Protection.

Google Opal is not a “degraded Dify”. Its strategic positioning and optimal utilisation methods revealed through actual use

"AI makes it cheaper, right?" - Is this happening in your country too?

Beyond YAML: Logic Compression for 50%+ LLM Cost & Latency Reduction

When Your CEO Says 'Let's Use AI': A Technology Selection Survival Guide

The Era of Choosing RAG — Learning Cognitive Load and Architecture Design from GPT-5’s Failures