Running Cassandra and Scylla on K8ssandra and the Scylla Operator after Rook dropped its own

Object storage that triggers work, a self-hosted answer to S3 plus Lambda on your own cluster

The harness behind two DOI-archived benchmark reports, and why a study premise can be killed in one working day

Self-hosted notebooks for the whole team, on Keycloak SSO, wired to the lakehouse and every data source you own

Lightweight device messaging into the platform, bridged to Kafka, with the Mosquitto 2.0 change the old guides miss

Self-hosted observability for everything you built, on the open standard instead of a per-host APM bill

A benchmark that killed its own hypothesis: strong models re-derive what they can check, and a weak model trusts a stale note over the evidence in front of it

The CNI that replaced Weave Net, kube-proxy, and a hand-rolled WireGuard overlay with a single component

The container runtime the kubelet actually uses, and how to configure and debug it now that the Docker shim is gone

A reproducible benchmark: 42.7% to 98.7% on knowledge-trap questions, and a platform that learns from empty

Argo CD and Flux, the deployment half the 2020 book left out, with manifests reconciled from a repo you own

Rootless container builds in CI with BuildKit, and where supply-chain guardrails belong

Self-hosted source, CI/CD, and a registry, on the lightweight fork that left its for-profit parent

An Apache-2.0 MCP server that gives frontier models access and grounded context to everything you built

Metadata, lineage, and a business glossary over the whole platform

Dashboards and SQL exploration over the lakehouse and Postgres, a Tableau replacement on the platform's SSO

Visual dataflow and ETL on your own cluster, instead of Workato and the iPaaS vendors

One login for every user-facing tool on the platform, instead of renting Okta or Auth0

SQL analytics over open table formats on the object storage you own

S3-compatible object storage for the data lake, backups, and the lakehouse

The index for the platform, on the Apache-licensed fork of Elasticsearch

A real-time event backbone, without the separate ZooKeeper ensemble to operate

I drove my own platform as a real client and told it to find every place a tool said one thing and did another. A case study in MCP SDLC.

High-availability PostgreSQL on your own storage, without the managed bill

An insight an analyst shares mid-session becomes a reviewed, signed-off change to the catalog rather than vanishing when the conversation ends.

Exposing services and automating TLS in 2026, with Cilium and cert-manager

Steering agents from the server middleware instead of the tool descriptions.

Portable block and shared-filesystem persistence for stateful services, no managed disks required

Compose mcp-trino, mcp-datahub, and mcp-s3 as Go libraries behind one endpoint, instead of running three servers that can't see each other.

The vendor-neutral substrate for a self-hosted platform: kubeadm, containerd, and Cilium on generic VMs