Notes on how we build AI products and systems in practice
A new era for Curiosity Studio: the foundation and the factory
OCR with three engines, and the 6 MB model that chooses between them
Photoshop, email, calendars, PostScript and OneNote: five formats, five origins
Reading LibreOffice's layout code and writing it again in C#
Reading a file safely: the rules every X-Ray package keeps
X-Ray: one family of .NET libraries for reading every file we index
What an extraction model returns that a JSON prompt does not
Release overview, October 2026
What a diffusion LLM is, and what Mercury 3 is like to run in a workspace
StructuredAI can now ask Jev, OpenAI Decisions and Mercury Decide
Build it from open source? Count the seams first
The app is the easy part. The foundation is not
The era of mockups is over
An aviation market intelligence app, built in an evening
Judge an AI admin by its authority, not its model
curio: Sudo's sandbox, in your terminal
How we rebuilt our design language in three days with AI
Release overview, September 2026
Design for the steered prompt: how Sudo's sandbox and approval hold up
Optimizing code with agents on a machine that won't hold still
Types don't make the model smarter. They make the loop cheaper
Give your agent a secure filesystem and a work environment
Sudo: seventy-four tools, replaced by one shell
Computerwelt: how we built a secure sandbox for agentic workflows
Computerwelt: porting a sandboxed shell and Python interpreter to C#
Porting a Rust document-extraction engine to C#, and ending up faster
Shipping agent documentation inside the NuGet package
Say the outcome, read the diff: a day with Sudo
Release overview, August 2026
What open data means when the software is yours to leave
Running Mistral's Shieldstral on CPU with .NET
Where your data goes, step by step, inside Curiosity
Transpose: our C# to JavaScript compiler just got 7x faster
Release overview, July 2026
Make or buy: the RAG stack decision every AI team faces
Permission-aware retrieval: the security question enterprise AI keeps getting wrong
Prompt Library and AI Tools: a better way to manage AI behavior
Curiosity for builders: a new admin experience
What building your own RAG stack actually costs
Data View: see your knowledge graph without a whiteboard
How a knowledge graph is different from a RAG pipeline
Scaling AI in aerospace and rail: takeaways from VivaTech
Release overview, June 2026
Six things to settle before your AI pilot starts
Release overview, May 2026
Release overview, April 2026
You will be locked into something. Decide where it lands
Release overview, March 2026
Which search do you mean? Six layers, and where each one belongs
Release overview, February 2026
Release overview, January 2026
A search score is an order, not a grade
Release overview, December 2025
Embeddings find similar meaning, identifiers find the same thing: how we merge both in one similarity engine
Release overview, November 2025
Measuring search relevance: precision, recall, reciprocal rank and NDCG
Release overview, October 2025
Release overview, September 2025
Approximate nearest neighbor search: how HNSW, IVF, product quantization and hashing trade recall for speed
Release overview, August 2025
Release overview, July 2025
Cosine, dot product and Euclidean distance: how vector search decides what is close
Release overview, June 2025
No posts match this tag.