About KC Tech

A personal workshop for disciplined AI engineering.

KC Tech is my (KC’s) private lab for building with AI inside hard rails.
I’m a chemical engineer from DTU in the 1980s who never stopped writing tools.
My master’s thesis was on something like “Simulation of complex systems with neural networks” back when backpropagation was mostly theory, not practice.

After decades of scripting factories, process plants, and network ops — from Quick C and MS Fortran onwards — I built AiStudio so an LLM could act as a serious engineering partner: fast, tireless, and constrained by structure.
If you live in .NET, WinForms/WebView2, MEAI, and DI, this will feel like home.

Why this exists

• Coding through a generic chatbot is fundamentally the wrong shape for serious engineering work.
• It is exhausting to keep correcting the same mistakes: diffs instead of full files, stale context, lost state, and “helpful” suggestions that slowly damage a codebase.
• I do not want to argue about versions, fight for recent information, or watch a model forget what it just did — I want a system that treats code, state, and history as first‑class citizens.
• When OpenAI removed the O1 Pro model I relied on, that was simply the trigger: I decided to patch the whole approach instead of chasing models.
• AiStudio’s job is to wrap and repair an otherwise unsuitable product (chat‑based coding) so it behaves like a disciplined engineering environment: full‑file edits, explicit tools, real state, and green builds — or nothing lands.

What lives here

• Documentation‑driven methods for deterministic AI work.
• Two‑mind orchestration (Designer + Programmer) with Observer transparency.
• Safe but powerful limbs: harness, SSH terminal, MacTelnet, vision — always through tools, logs, and gates.
• Notes on memory, grounding, and the rails that keep AI trustworthy.
• Concrete platform choices: .NET, WinForms + WebView2, DI, MEAI as the stable AI contract.

How I work now

• Small files, small commits, zero warnings — always.
• Read‑before‑write, atomic writes, and green gates before anything lands.
• Tools as a language: code_tree, read_file, staged_edit_file, staged_commit, harness, terminal.
• When something is risky, I add a view so I can watch it.
• AiStudio runs as a high‑leverage engineering partner under strict supervision:
  it is often more consistent than a junior engineer, but every terminal session (SSH/MacTelnet), health check, and file change still goes through explicit rails and my approval.

If you want the long view

Read I Believe — AI, Engineering, and the Future for my broader stance on why AI, under structure, is an amplifier — not a threat.

I’m almost 70, not chasing a startup. AiStudio is built by KC, for KC.
The ideas — rails over vibes, safety over swagger, boring tools over fragile magic — are here so other engineers can build their own disciplined ecosystems.