We help companies to digitalize their businesses. #softwaredevelopment #customsofware #iotdevelopment

Joined May 2014
443 Photos and videos
Agents need permissions, logs, and review paths. Prompts are the easy part. #AgenticAI #EnterpriseSoftware
Cheaper models change product scope. The key question is how AI performs in daily use. #AI #SoftwareDevelopment
2
Voice AI hit production-grade in 2025. Sub-800ms latency, solid interruption handling. Procurement catching up now. #VoiceAI
5
Most AI projects skip observability and find out at month four. Procure it as a line item from day one. #LLMOps
11
Plain RAG breaks on multi-hop questions. Graph RAG handles them. The architectural question for serious AI projects. #RAG
12
Prompt engineering = 2023. Context engineering = 2026. The work moved upstream of the prompt. #AIEngineering #LLM
1
8
Evals on every commit, not just at the testing phase. Anything else isn't AI in production. #AIEngineering #LLMOps
1
16
A2A is to multi-agent what MCP is to tools. Single-agent today, multi-agent within 18 months. Architecture decision now. #A2A #AI
10
AI lock-in is at orchestration and memory, not the model. That layer is rarely evaluated and the hardest to leave. #AIStrategy
5
MCP just passed 10K servers. The standard interface from agents to enterprise systems. Integration tax is dropping fast. #MCP
5