Self-Hosting vs SaaS in 2026: 37signals Saved $2M a Year. Should You?
A balanced look at self-hosting vs SaaS and cloud repatriation in 2026: 37signals' published cloud exit numbers, Ahrefs and Dropbox, what the surveys measure, Coolify, Dokploy and Kamal, the ops cost nobody invoices, and a break-even rule.
13 minOct 20, 2026Self-Hosting LLMs vs API Costs in 2026: The 62% Utilisation Break-Even
Self-hosting an open LLM vs paying per token in October 2026: H100 prices from $1.17 to $4.29 an hour, 8,973 tokens/s per GPU for gpt-oss-120b on vLLM, a $0.20 per million API price, the utilisation each price needs to break even, and the ops costs the spreadsheet leaves out.
13 minOct 20, 2026Multi-Tenant SaaS Architecture Explained: Silo, Pool, Bridge, Cells (2026)
Multi-tenant SaaS architecture explained: the AWS SaaS Lens silo, pool and bridge models, database-per-tenant vs shared schema with row-level security, noisy neighbours, cell-based architecture, and how to work out cost per tenant, with real examples.
12 minOct 20, 2026How to Set Up LiteLLM Proxy: Budgets, Fallbacks and Virtual Keys (2026)
Step-by-step LiteLLM proxy setup for v1.106.0: verify and pin the image, run it with Postgres on Docker Compose or Helm, write config.yaml, add fallbacks and retries, issue virtual keys with budgets and rate limits, log to Langfuse and OpenTelemetry, and troubleshoot the common errors.
13 minOct 20, 2026Kubernetes Requests and Limits Explained: 8% CPU Used, Resize GA in 1.35
Kubernetes requests and limits in 2026: how QoS classes and oom_score_adj decide who dies, why CPU limits cause CFS throttling, what cgroup v2 changed for OOM kills, in-place resize GA in 1.35, VPA InPlaceOrRecreate, and why clusters still use 8% of their CPU.
13 minOct 20, 2026How to Fix Kubernetes Pods Stuck in Pending: Causes and Commands (2026)
Pods stuck in Pending? Read the FailedScheduling event, then fix the cause: insufficient cpu or memory, untolerated taints, node affinity, unbound PVCs and volume zones, topology spread, quotas, scheduling gates, or Karpenter and Cluster Autoscaler not scaling.
12 minOct 20, 2026Kubernetes Cost Optimization in 2026: 8% CPU Use and What Fixes It
Kubernetes cost optimisation as of October 2026: Cast AI measures 8% CPU and 20% memory utilisation, Datadog puts 83% of container cost on idle, plus OpenCost vs IBM Kubecost 3.0, Karpenter 1.15 Balanced consolidation, in-place resize, spot and Graviton.
12 minOct 20, 2026Kubernetes Autoscaling Explained (2026): HPA, VPA, KEDA and Karpenter
Kubernetes autoscaling in 2026, explained: HPA v2 behavior and the new tolerance field, HPA scale to zero in 1.37, VPA 1.8 in-place resize, KEDA 2.21, Cluster Autoscaler vs Karpenter 1.15, GKE and AKS node auto-provisioning, and the mistakes that cost money.
13 minOct 20, 2026Continuous Profiling in 2026: OTel Profiles Alpha, eBPF Agents, Real Costs
Continuous profiling in 2026 explained: OpenTelemetry Profiles reached public alpha in March, the Elastic-donated eBPF profiler now ships as a Collector, Pyroscope 2.0 and Parca compared, the real overhead numbers, and how profiles turn into cloud savings.
13 minOct 20, 2026How to Reduce AWS Lambda Cold Starts: Init Duration to SnapStart (2026)
Step-by-step guide to measuring and reducing AWS Lambda cold starts in 2026: read Init Duration, price the August 2025 INIT billing change, pick runtime and arm64, bundle and lazy-load, tune memory, enable SnapStart and provisioned concurrency, and fix common errors.
14 minOct 20, 2026AI Gateways in 2026: Six Options Compared After LiteLLM's 40-Minute Backdoor
What an AI gateway (LLM gateway) does in 2026 and how LiteLLM, Portkey, Kong, Cloudflare, Envoy AI Gateway (now Agent Router) and OpenRouter compare on routing, budgets, caching and guardrails, plus the LiteLLM PyPI backdoor, three LiteLLM CVEs on CISA KEV, and build vs buy.
13 minOct 5, 2026FinOps for AI in 2026: 98% Manage Token Spend, 3 in 4 Can't Prove the Value, GPUs Run at 5%, and FOCUS Gets a Token Column in 1.5
The AI cost ledger as of October 2026: State of FinOps 2026 (1,192 practitioners, 98 percent manage AI spend, up from 31 percent in 2024), the Tokenomics Foundation finding that three in four enterprises cannot prove AI outcomes to the CFO, a per-million-token price table for Anthropic, OpenAI, Google and DeepSeek with cache and batch multipliers, a 20-step agent loop at $1.58 uncached and $0.42 cached, the FOCUS 1.2 to 1.5 roadmap and who exports which version, LiteLLM and Cloudflare budgets, Cast AI's 5 percent GPU utilisation, and $0.0037 of model cost inside a $0.99 resolution.
13 min