Studios216 Articles
Articles
Long-form ideas on AI, engineering, software, data and communication.
EN · 2026-10-01 · 10 min read
DiffDocs in 3.5 Months: What AI Accelerated—and What Still Required Engineering
How DiffDocs became a functioning SaaS in 3.5 months: what AI accelerated and which architecture, security, and validation decisions still required engineering.
EN · 2026-09-30 · 8 min read
A Local LLM Is Not a Lab: How Do You Build an AI System You Can Repeat, Measure, and Audit?
From a local endpoint to RAG, agents, observability, and MCP: lessons from an AI lab designed to make every layer repeatable, measurable, and auditable.
EN · 2026-09-30 · 7 min read
Bidirectional Power Triangle: Why Should Every Calculation Start with the Same Input?
54.1 A or 60.1 A for one motor? See how efficiency and bidirectional power-triangle calculations connect kW, kVA, kVAR, PF and current.
EN · 2026-09-30 · 7 min read
From Spreadsheets to WattsWise in 216 Hours: Where AI Helped—and Engineering Still Mattered
From repeated spreadsheet calculations to the first WattsWise submission: 216 hours of active work, AI-assisted implementation, and engineering decisions that still needed an owner.
EN · 2026-09-29 · 10 min read
After 8,901 Commands, I Changed the Way I Work with AI Agents
After 8,901 commands, the problem shifted from adding more tools to governing context, memory, caching, and redundancy in AI agent workflows.
EN · 2026-09-29 · 12 min read
Are You Delegating Too Much to AI… Just Because a Black Screen Intimidates You?
A black screen can make delegating everything to AI feel easier. How to turn useful results into reproducible, verifiable capabilities you can keep.
EN · 2026-09-29 · 11 min read
More Agents or Better Specifications? What Changed After Trying Both
After trying multi-agent workflows and Spec-Driven Development, one conclusion emerged: specify first, then decide when multiple agents truly add value.
EN · 2026-09-28 · 9 min read
AI Does Not Eliminate Engineering; It Forces Engineering to Become More Explicit
AI accelerates execution but does not eliminate engineering. Why specifications, acceptance criteria, audit, and evidence matter more than ever.
EN · 2026-09-28 · 11 min read
AI Said “Done”. The Product Had a Different Opinion
An agent can complete its plan and pass tests while a feature is still incomplete. How to separate autonomy, validation, and evidence-based acceptance.
EN · 2026-09-28 · 10 min read
From an Idea to a Functional Product: What I Have Learned Building Software with AI
AI can turn an idea into a first implementation remarkably fast. What still has to happen before that speed becomes a reliable, maintainable product.
EN · 2026-09-27 · 8 min read
Do We Really Need Another AI Agent—or Should the System Know This Already?
When should an AI agent keep reasoning, and when should known decisions become deterministic software? A practical framework for rules, tests and durable system knowledge.
EN · 2026-09-25 · 9 min read
How to Work Better with AI: From Better Prompts to Better Collaboration
A practical introduction to working better with AI through context, iteration, examples, verification and human judgment—not magic prompt formulas.