Steve Sanders

@stevendsanders.bsky.social

Startup CTO - helping nonprofits connect with donors Side projects: https://episode.ninja https://repotrends.com Building serverless event-driven architectures on AWS

One reason I like using AWS over other cloud vendors is that they heavily utilize the same core services that I use (Lambda, DynamoDB, etc) internally at both AWS and Amazon. The downside of this is that if they break one of those core services it results in a cascade that kills many others 😬

As the solo engineer at a tiny startup, I need to use every tool available to work as efficiently as possible. That includes AI. Sometimes it produces garbage and sometimes it's perfect on the first try. Today I vibe coded a 2FA TOTP generator app in about 5 minutes with a single short prompt 🤯

2FA TOTP Generator app screenshot

128 MB of memory is a terrible default even for simple Lambda functions, especially those that back anything user-facing. Serverless Framework had a good idea to default to 1024 MB while AWS CDK and AWS SAM still default to 128 MB. Who can spot when I gave this poor function more memory?

Lambda function average duration

Amazon: Our customer service chat bot uses a powerful LLM that can predict if you might want to return a recent order and suggest a nearby drop-off location! Customer: Can I just get a price adjustment (the item went on sale after I ordered) instead of returning it? Amazon: ...absolutely not.

Interesting pricing model for a SaaS that I saw today: Monthly, Yearly or Lifetime pricing where the Lifetime price is equal to 3x the Yearly price Getting the Lifetime payment up front seems nice (especially if you see high churn) while sacrificing long-term recurring revenue. What do you think?

SaaS Monthly/Yearly/Lifetime pricing