Creating a Cost-Effective and Highly Available Chatbot with OpenAI, AWS, and Discord

This blog post discusses how to create a cost-effective and highly available chatbot with OpenAI, AWS, and Discord. It explains the advantages of using these three tools together, and the challenges involved in creating a chatbot. Solutions proposed in this article include using EC2 Spot Instances for cost-effective hosting and using distributed locking with leases to prevent duplicate work and handle faults, providing a cost-effective and reliable way to create a chatbot.

Tutorial - Personal Todo List on AWS

Infrastructure as code, compute, storage, monitoring, deployments.


API contract verification, fuzzing.


Asset-centric threat modeling, attack modeling, the bare essentials.

Scaling and Performance

Micro-benchmarks, load testing, scaling limits and bottlenecks.


Chaos testing, gamedays.


Fault tolerance, replication, cells


Canaries, traffic replay, changes and deployments, feature flags and experiments.

Performance and Cost Efficiency

Latency vs. throughput, types of performance, opex vs. capex, capacity and scaling planning.

Functional requirements

End-to-end testing, unit testing, fakes, property-based testing, model-based testing, formal methods.

Events, Messaging, and Workflows

Queues and work, events and notifications, workflows - how hard can they be?

Design - Storage

Myth of the single solution, access patterns, latency, availability, consistency and idempotency, caching.

Design - Compute

Cattle not pets, latencies: cold vs. warm vs. hot, scalability vs. elasticity.

Change and Operations

Change is a continuum, deployment types, monitoring production, faking customers, testing in production.

Case Studies

Introduction Hi! Want more content like this? I don't publish often, but when I do you'll be the first to hear about it. Get updates by email Powered By ConvertKit

Case Studies

Two services with different focuses.

Appendix - Principles and Values

Documentation, tutorials, small experiments, asking questions the right way.

Appendix - How to Research a Managed Service

Tutorials, docs, experiments, scalability, limits and edge cases, deployments and changes. Example using SQS.