IT's ABOUT TECH #13 · Basel · 10 June 2026
I Stopped Writing Most of My Code
The year my backlog came alive.
Raffael Schneider Enterprise Solution Architect AI Platforms ·
raskell.io
01
It came down to two subscriptions.
May → November 2025
- Before
Mostly using Zed's AI offering. Solid editor, fine integrations. Nothing that changed how I worked.
- May 2025
Claude Code drops. I take the $20 Anthropic subscription on day one. Early adopter, on a hunch.
- November 2025
Opus 4.5 ships. I upgrade to the $200 Max 20× plan right before Christmas vacation. A better model meets three weeks of free time on my hands.
I am an engineer. I know how to build things.
The bottleneck was never ideas. The bottleneck was always time.
02
Then something changed.
I expected a better autocomplete.
What I got was something closer to a junior engineer that never sleeps.
03
The new workflow
Before
- Decide what to build
- Translate thoughts into syntax
- Review and ship
After
- Decide what to build
- Translate thoughts into syntax
- Decide whether the result is acceptable
You no longer spend most of your time translating thoughts into syntax. You spend most of your time deciding.
04
What I actually built
Archipelag.io
Distributed AI compute. Inference across idle GPUs and mining rigs.
Cyanea
Open community platform for life-science research.
Humankind
A long-running creative initiative, finally with a home of its own.
Arcanist.sh
Haskell ecosystem in Rust. Home of hx and the Basel Haskell Compiler.
Zentinel
Security-first reverse proxy on Pingora.
Die Zukunft
A Swiss political party. Past left and right, into UBI, digital sovereignty, and a serious technology agenda.
Experiments and tooling
Terrarium- Shiioo
- Conflux
- Refrakt
Kurumi- Vela
- …and the daily-driver utilities
None of these were started because AI generated ideas. AI allowed existing ideas to escape my notebook.
05
What AI is still bad at
- Requirements
- Architecture tradeoffs
- Long-term maintainability
- Taste
- Product decisions
- Knowing when not to build something
The human did not disappear. The human moved up the stack.
06
The unexpected consequence
I thought AI would make me write software faster.
Instead, it made me start projects I would never have started before.
The biggest change was not productivity. The biggest change was willingness.
07
What happened next
- March 2026 · San Francisco
Co-presented an AI Web Security tool at RSAC 2026 with Milan Duric. Same trip: a private downtown dinner hosted by Maverick Capital, room of AI Platform leaders.
- Spring 2026
Shipped a Temporal-based self-service that lets users order web-service resources while AI quietly handles the operational work. Role at the time: Senior Platform Engineer & Solution Architect, Web Security / Application Delivery.
- Two weeks ago
New role: Enterprise Solution Architect for Group-wide AI Platform and AI-assisted development.
- Last week
Joined J floor. More talks already on the calendar.
None of this was on my calendar a year ago.
Thank you
The most important thing Claude Code gave me was not faster software development. It gave me the ability to turn years of accumulated ideas into working systems before I lost interest in them.