Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag · Browse the archive.
Longhorn V2's Kernel Bypass: What SPDK Storage Actually Buys Your Self-Hosted Volumes
Longhorn's V2 engine moves block storage into userspace with SPDK for lower latency than the iSCSI-based V1 path — at the cost of a CPU core, 2 GiB of hugepages, and raw block disks per node. Measured gains, hardware caveats, and when to pilot it on your own metal.
Leapcell's Serverless Escape Hatch: the Traffic Level Where Per-Millisecond Billing Stops Being Cheaper
Leapcell bills serverless per millisecond but lets any service flip to fixed persistent-server pricing. A worked bill across three traffic profiles puts the crossover near one million 60-ms requests a month — and shows how response time moves it 10x.
Keep Secrets Out of Your AI Agents: Credential Gateways, Short-Lived Tokens, and the Leak That Should Be Impossible
A prompt-injected agent leaks whatever secrets it can see — so stop letting it see them. How credential gateways, short-lived OIDC tokens, and Vault-backed brokers make agent access provable and revocable.
KEDA vs Knative vs Sablier: Three Ways to Make Idle Kubernetes Services Cost Nothing
KEDA, Knative, and Sablier put the scale-to-zero interception layer in three different places. A concrete comparison of cold starts, config cost, and probe hygiene — plus the health-check audit that decides whether any of them saves money.
Karpenter Still Doesn't Run on Bare Metal: What vCluster's Auto Nodes Workaround Really Buys You
Karpenter provisions right-sized cloud nodes in under a minute, but it still has no bare-metal provider in 2026. A stage-by-stage latency budget shows what vCluster's Terraform-backed Auto Nodes actually recovers — and why Cluster Autoscaler remains the sane default for a Cluster API fleet on owned hardware.
The Hybrid-Platform Trap: Why Bolting AI Agents Onto Your Dev Platform Breaks
35% of organizations run AI tooling bolted onto a developer platform designed for humans. A five-row audit of what breaks when agents operate the sidecar — auth, status, rollback, guardrails — and the API-first contract that fixes it.
From $87 in Heroku Dynos to $7 on Hetzner: a Real 2026 Migration Bill, Line by Line
A line-by-line recompute of a Node.js API plus worker plus Postgres on Heroku Standard dynos ($87/month) versus the same containers on a Hetzner CX23 (~$7/month) at September 2026 prices, with honest egress math, a hobby-to-growing-prod sensitivity ladder, and where Heroku's zero-ops premium still wins.
Fly.io's Two New 2026 Billing Lines: Read the Invoice Before You Migrate Off It
On January 1, 2026, Fly.io's invoice grew two lines — metered volume snapshots and app-scoped egress IPs. A line-by-line read of what they meter, a worked workload total against a flat Hetzner box, and what the 2024-2026 margin-tightening pattern signals for your stay-or-go decision.
Fly.io's Two New 2026 Line Items, Priced Line by Line: Snapshots and Cross-Region Traffic vs a Flat Hetzner Fleet
Fly.io started metering volume snapshots in January 2026 and inter-region private traffic in February. A line-by-line bill for a typical two-region app totals ~$23 vs ~$11 on flat Hetzner hardware — and ~$45 with an Africa/India region.
Subscribe
New posts land in your reader as soon as they publish. Pick a format — all three carry the same posts.
Current feeds keep roughly two days of posts so daily polling does not miss a burst. Older entries stay reachable from the feed's next-page link in readers that follow it, or from the blog archive.
Following one topic instead? Browse tags