Thinking out loud

Reliable RAG and tool calling usually need one stable working language. Keep English inside the model loop, then localize for users at the edges.

Anthropic just shipped design, prototyping, and presentation tools inside Claude. Combined with Code and Cowork, one person now has a full creative agency on their laptop.

How to use a single DeepL API key across many tenants without glossaries, style rules, or cached translations leaking between customers. Here's the approach I use for Rasepi.

Measuring AI adoption by token spend is the same mistake we made with lines of code in the 90s. Same flaw, new dashboard, much higher stakes.

Half your team is building the future with AI. The other half thinks it's a fad. The gap between them is becoming the biggest competitive risk most companies don't see.

The cost of building just collapsed. So what does that mean for every SaaS company betting their business on 'you don't have to build it yourself'?

I built a full SaaS product, marketing site, developer docs, and blog in three weeks with Claude. Here's the honest breakdown of where AI shines and where you're completely on your own.

One person with AI can do the work of ten. But did anyone stop to ask what happens to that one person? Or what happens if you keep the ten?

Documentation platforms force readers, writers, and AI into one interface. But consuming knowledge and creating it are cognitively different tasks. this platform separates them.

AI readership is up 500%. Notion shipped 21,000 agents. Confluence got Rovo. GitBook published the State of Docs. Five trends from across the industry that tell us where documentation is heading.

The person integrating your API no longer reads your docs. They sit in Claude and describe what they want. Developer relations, API documentation, and the whole getting-started funnel need to be rethought for this new reality.

this platform doesn't just translate your documentation into other languages. It learns your terminology, matches your tone, and lets every language version live its own life. Here's how.

A deep technical walkthrough of how this platform's translation pipeline actually works: glossary resolution, DeepL style rules and custom instructions, content hashing, and the integration that ties it all together.

Most companies have onboarding_germany, onboarding_japan, onboarding_brazil. In this platform, it's just 'Onboarding'. One document. Shared steps translated, local steps per language. No more copies drifting apart.

Global teams don't just need translations. They need knowledge that works in every market, with each language carrying its own structure. Block-level localisation makes that practical.

Expiry dates solve accountability. But a document can go stale in a hundred ways between reviews. Part 2 explains how continuous freshness monitoring fills the gap.

Your documentation might be technically correct today. But in six months, who checks? Freshness is about to become the most important signal in your knowledge base.

Your AI assistant treats a document reviewed last week the same as one nobody has touched in two years. Content governance fixes that.

Reading is powerful, but effortful. Conversation is older, faster, and more natural. Speaking to information often feels mentally lighter than scanning pages of text.

Confluence and Notion were built for a pre-AI model of documentation. They can evolve, but established platforms carry structural baggage. Newer systems can design for AI from day one.

A deep technical walkthrough of how this platform's plugin system, action guard pipeline, and block-level translation engine actually work, with real code from the codebase.

Read about some of the new features shipped in Visual Studio 2022 Preview 3

Many people think Adaptive Cards is something Microsoft invented to be used in MS Teams or Power Platform and the alike and while that is totally true it's only half the picture. Adaptive Cards and the technology behind, including the templating engine, can be used for so much more!

Generating fully working CRUD API's in seconds!

Often controllers are really really similar to each other, here's a generic approach to this

Many people think open source is free software and while that is sort of true, its totally not true depending on the point of view

Here's a small getting started guide for my API Generator Project

A developer career is more than just writing code, #SharingIsCaring and giving back supports other developers!

A full guide how this page is made, the stack i'm using and more

I just uploaded a bunch of stuff to Github because #SharingFeelsGood!

How to create fully working ODATA CRUD API's from just classes.

MadeWithCards now has a lot more content and quite a few more updates!

Quite often i see questions about how app settings work localy, in azure and in docker, here's a small explaination

A brand new updated MadeWithCards.io was released

Simply explained templating is data-binding onto JSON strings.

With AdaptiveCards 1.3 (also 2.1.0 ) cards got a relatively simple but really powerfull change.

Recently, a new version of the Teamwork .Net SDK was released and I’d like to give you a few examples on how this can be used to make development for Teamwork Projects a lot easier when working with .Net.