iiinova — where your digital self keeps evolving

- observable enough to debug - structured enough that an agent can help safely The goal is not “AI operates Kubernetes.” The goal is: make Kubernetes declarative, observable, and policy-gated enough that an agent can become a useful operator instead of a new source of chaos.

000

It gets constrained inputs, read/write boundaries, policy gates, branch protection, and Git as the actuator. Higher-risk changes stay human-reviewed. That feels like the right shape for small infrastructure: - cheap enough to run on one node - Git-native enough to audit

100

Flux, Grafana, and Forgejo MCP servers give the agent operational context. Alerts can spawn bounded jobs that diagnose, propose a PR, and report back with evidence. The important part: the agent is not a root shell with vibes.

100

Prometheus, Grafana, Loki, Tempo, Alloy, Alertmanager, blackbox probes, and runbooks make the system explainable before an agent touches it. 10. MCP + alert-to-agent loop

100

Secrets stay encrypted in Git and are decrypted by Flux inside the cluster. Shared platform secrets are replicated intentionally, not copied around by hand. 9. Observability

100

One generic Helm chart, explicit app registration, per-project service-account impersonation, scoped RBAC, quotas, limit ranges, PSA labels, and default-deny egress. 8. SOPS + age

100

6. CloudNativePG per project namespace Each project can own its app namespace and database, with WAL/base backups through Barman to S3-compatible object storage. 7. Tenant boundaries

100

CI pushes sortable image tags. Flux writes the promoted tag back into each app’s deploy/values.yaml, so deployments remain Git commits. 5. Gateway API + Traefik + cert-manager + external-dns Apps get HTTPRoutes. The shared gateway carries wildcard TLS. DNS lands in Cloudflare automatically.

100

3. Forgejo Self-hosted Git source of truth, OCI registry, and Actions CI. The build runner is off-cluster so CI spikes don’t fight production workloads. 4. Flux image automation

100

A lean Kubernetes base. Single-node today, with the app and DB model designed so it can move toward HA without changing how apps are deployed. 2. Flux Operator One FluxInstance syncs the cluster root from Git. No app-of-apps sprawl, no imperative drift as the default operating model.

100

I’ve been building a small agentic GitOps infrastructure on k3s, and the main lesson is this: the agent should not be the platform. GitOps is the platform. The agent is an operator that works inside boundaries. Current shape of the stack: 1. k3s on Hetzner

100

Check it out here: pullplane.com/

000

I genuinely don't know yet if this solves real pain or if it's just another idea. that's why I'm looking for testers. bring your own Codex / Claude subscription or API key, I cover sandbox and hosting. all I want back is honest feedback. comment or DM men

100

can non-technical people and teams move fast in a realtime collaborative workspace, with experts able to step in when it matters? - everything runs in isolated sandboxes (generated, tested, previewed there) - separate agents for research, coding, and review, each on the model you choose

100

but the reality is still, a human has to review it in an business / enterprise context. an auto-scan won't tell you your architecture is wrong or your tech-stack decision will hurt in six months. the expert in the loop is the value. so I built pullplane to test one idea:

100

Arthur @arthurliebhardt.com · 15/04/2026

I do a lot of work around agent-assisted coding/vibe coding, and two questions come up every single time: How do I get my app live? How do I make it secure? my first instinct was an agentic cloud that just auto-checks and ships it

320

TIL you can type ! in claude code to execute shell commands directly and then keep working with the results previously I just opened another tab and copy pasted everything like back and forth

120

Arthur @arthurliebhardt.com · 13/04/2026

best part? spin it up from anywhere, let it work while you do other stuff this is where development is heading currently

000

Arthur @arthurliebhardt.com · 13/04/2026

@cursor_ai cloud agents are wild you literally spin up a VM, tell it what to build, and watch it: - open browsers - click around - run tests - record everything - debug itself first agent walks you through setup and gets everything running without you lifting a finger

100

Arthur @arthurliebhardt.com · 11/04/2026

what are your experiences with @convex? How fast did you hit the limits of the free tier and what is possible with the 25 dollar plan?

000

GitHub - arthurliebhardt/today-md

github.com/arthurliebh... install/update via (export/sync data before): curl -sL raw.githubusercontent.com/arthurliebh... | bash

github.com

Contribute to arthurliebhardt/today-md development by creating an account on GitHub.

000

This release tightens the planning loop on macOS: a reliable Cmd-_ quick add flow, Monday-based week views, cleaner rescheduling, better calendar onboarding, preserved subtasks on reload, and clearer list color picking.

100

I introduced some bugs with the latests changes and focusing on fixing them today-md v1.7.6.

100