https://vantalect.com/blog/ Cloud Design Patterns on AWS and essays on measuring AI-assisted engineering, from Charles Sieg at Vantalect. en-us Sun, 18 Oct 2026 00:00:00 +0000 https://vantalect.com/blog/retries-backoff-jitter-idempotency/ https://vantalect.com/blog/retries-backoff-jitter-idempotency/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) A retry is extra load at the worst moment. Backoff, full jitter, a retry budget, and idempotency keys make it safe. https://vantalect.com/blog/circuit-breakers/ https://vantalect.com/blog/circuit-breakers/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) A slow dependency is more dangerous than a dead one. A circuit breaker with its own small bulkhead turns the first into the second on purpose. https://vantalect.com/blog/queue-based-load-leveling/ https://vantalect.com/blog/queue-based-load-leveling/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) A queue lets you size workers for the average load instead of the peak, if you cap the consumers and alarm on the age of the oldest message. https://vantalect.com/blog/load-shedding-and-backpressure/ https://vantalect.com/blog/load-shedding-and-backpressure/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) An overloaded service that tries to serve everyone serves no one. Reject excess work early, drop expired requests, and tell callers to slow down. https://vantalect.com/blog/cache-aside-request-coalescing/ https://vantalect.com/blog/cache-aside-request-coalescing/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) A cache your database cannot survive without is no longer an optimization. Coalesce misses, jitter expiry, and serve stale data while you refresh. https://vantalect.com/blog/event-fan-out-dead-letter-queues/ https://vantalect.com/blog/event-fan-out-dead-letter-queues/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) Fan-out lets each consumer fail alone, but only with its own queue, its own retry policy, and a dead-letter queue that pages someone. https://vantalect.com/blog/cqrs-read-models/ https://vantalect.com/blog/cqrs-read-models/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) When reads and writes want different shapes, scales, or engines, keep one model for commands and build read models from its changes. https://vantalect.com/blog/static-stability/ https://vantalect.com/blog/static-stability/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) A statically stable workload survives an Availability Zone loss without launching, scaling, or reconfiguring anything. Here is the arithmetic. https://vantalect.com/blog/cell-based-architecture/ https://vantalect.com/blog/cell-based-architecture/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) Redundancy protects you from hardware, not from your own bad deploy. Split the workload into independent cells so each failure reaches only one. https://vantalect.com/blog/observability-red-use-slos/ https://vantalect.com/blog/observability-red-use-slos/ Sun, 18 Oct 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) Green dashboards do not mean happy customers. Measure rate, errors, and duration, connect them with traces, and page on SLO burn, not on CPU. https://vantalect.com/blog/defending-the-numbers/ https://charlessieg.com/articles/leverage-factor-defending-the-numbers.html Mon, 02 Mar 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) The first leverage article drew hostile mail. This one answers it: what the estimates mean, why the METR study does not apply, the supervisory factor. https://vantalect.com/blog/agentic-coding-and-decision-fatigue/ https://charlessieg.com/articles/agentic-coding-decision-fatigue.html Sat, 28 Feb 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) Supervising AI agents means hundreds of accept-or-reject decisions a day, and decision quality falls by afternoon. What the science says, and what compensates. https://vantalect.com/blog/the-leverage-factor/ https://charlessieg.com/articles/leverage-factor-measuring-ai-engineering-output.html Tue, 24 Feb 2026 00:00:00 +0000 info@vantalect.com (Charles Sieg) Leverage is a small input controlling a large output. For engineering with AI agents, it is senior-engineer hours divided by agent minutes.