NicoGIS is back. 🚀
GIS was the start. The map is much bigger now.
🌍 ArcGIS & Geospatial
💻 .NET & Software Architecture
☁️ Azure & Cloud
🤖 AI, LLMs & Agents
🧠 GeoAI & Spatial Data
Building and sharing across GIS, software, cloud & AI.
GIS is still at the core.
Geodatabase schema evolution is a dependency-graph problem, not a table editor.
Domains, subtypes, relationship classes and attribute rules constrain change order.
Diff the desired state, validate dependencies, then apply DDL deliberately.
#ArcGIS#dotnet
🌍 GeoDynamics: exploring physics inside 3D GIS.
Rockfalls, barriers and water overtopping—combining ArcGIS Maps SDK for JavaScript, WebGL and Rapier physics.
Interactive geospatial simulations, built for the browser.
#ArcGIS#WebGL#GeoDynamics
SqlClient Pool V2 uses Channels to open physical SQL connections concurrently.
But OpenAsync still runs blocking network I/O on worker threads: high RTT can move contention to the ThreadPool.
Measure cold-start latency, p99 and worker growth.
#dotnet#SQLServer
Embeddings are not magic black boxes. In ArcGIS Pro, SHAP over 256-D geodemographic embeddings can be aggregated by latent segments to reveal which geographic context drives predictions. Explainability should be part of GeoAI architecture, not an afterthought. #ArcGIS#GeoAI
Agent interoperability is becoming an application-layer concern. AG-UI gives .NET services a typed event stream for state, tool calls and lifecycle, while https://t.co/HM5KywiPmm keeps IChatClient as the boundary. Protocol first beats framework lock-in. #dotnet#AI
Agentic UI changes the Blazor architecture: the agent is no longer a chat box, but a stateful participant in the UI. Streamed content, tool calls, approvals and shared state become first-class components. The hard problem shifts from prompting to orchestration. #dotnet#AI
Spatial embeddings change GIS from “where is it?” to “what places are similar in context?”. In ArcGIS, geodemographic embeddings can feed clustering, prediction and similarity search without hand-crafting dozens of variables. Location becomes a learned feature. #ArcGIS#GeoAI
.NET 11 performance gains increasingly come from below application code: JIT optimizations, runtime-async, GC trimming/compaction and better codegen.
For architects, the rule is simple: benchmark on the target runtime, not on assumptions.
#dotnet#performance
Production LLMs are infrastructure systems, not just models. 🧠
Quantization changes VRAM pressure, throughput and latency. Batching changes GPU efficiency. HTTPS, observability and governance decide whether the deployment is actually production-ready.
#LLM#Azure
.NET performance isn’t just about faster code. ⚡
JIT, GC, allocations, SIMD, syscalls and locking all shape what happens at runtime.
With .NET 11, some of the most interesting improvements happen below your code.
Performance often starts where application code ends.