Data systems, AI infrastructure, search and automation, and the real-world architectures that make intelligent applications survive production.
The latest
An AI agent is software that pursues a goal by deciding, acting through tools and checking the result, in a loop. Most of what is sold as agentic AI is not that. Here is…
Everyone upgrades models. Almost no one engineers what agents remember, forget and may know. That gap is where agentic AI fails in…
Five questions to test any system that decides about you: understand it, contest it, leave it, be forgotten, and still surprise it.
From the imitation game to Article 50 of the AI Act: we went from rewarding machines that fool us to forcing them…
Europe is not building one sovereign cloud. Across customs, health, tax and NATO, a repeatable delivery coalition is emerging.
Italy had Europe’s first national AI law. Draft Decree No. 418 is its operational layer: biometrics, evidence, liability, agentic AI.
How MongoDB’s search layer operates, why observability matters, and how to diagnose it in production architectures.
AI now reads and writes legacy COBOL. The impact is economic and architectural, not code replacement. Where it helps and where risk…
Most data models do not break in design reviews. They break in production. The truth about relationships, duplication and performance.
100K writes per second is an architectural decision, not a tuning trick: shard key, clean write paths and disciplined indexes.
Data quality is not a downstream fix. It is a structural property of the system, measurable only when decisions converge into a…
Generative AI’s real gift to legacy modernization is understanding, not code generation: applications, data and dependencies, read at last.
MDM is not about consolidating records. It is about defining how truth is constructed when systems disagree. A production-grade pattern.
A recorded talk from MongoDB.local Milano on how data modeling changes when applications, not tables, drive design. Lessons from production.
ai.mongodb.com is more than a URL. It marks the moment AI stops orbiting the database and becomes a place inside the platform.
AI is no longer the hard part. What breaks systems is everything around it: memory, retrieval, cost, reliability, and the gap after…
Complexity rarely comes from technology itself. It comes from how data, automation and ownership scale without shared architectural intent.
An AI agent is software that pursues a goal by deciding, acting through tools and checking the result, in a loop. Most of what is sold as agentic AI is not that. Here is the test, and what makes agents work in production.
Twelve books, eleven from Wired’s seasonal thirty plus one of my own, on how compute, data and the state became one conversation.
The remake lands on Switch 2 as a rapturous review and a holiday-quarter bet. Nintendo’s real product is a memory, re-embodied and resold.
Ageing populations and a shrinking clinical workforce push care towards one patient story that holds across providers and time.
One database, twenty-six landing zones. Where MongoDB runs, from IBM Z and VMware to sovereign DBaaS and air-gapped clusters.
Everyone upgrades models. Almost no one engineers what agents remember, forget and may know. That gap is where agentic AI fails in production.
Not emulation but re-embodiment: gate-level silicon memory of a 1982 machine, and a lesson in understandable computing.
Five questions to test any system that decides about you: understand it, contest it, leave it, be forgotten, and still surprise it.
I spent a career building the systems that remember on our behalf. Then I asked who owns them. Why I wrote Who Owns Memory Owns the Future.
From the imitation game to Article 50 of the AI Act: we went from rewarding machines that fool us to forcing them to unmask themselves.
End of content
End of content
Search the archive
Search more than 90 articles across data, AI, public-sector transformation, technology and culture.
