Sophia Willows' Blog

After you get an agent working, you’ll want to get it to run cheaply and quickly.

I’ve administered a lot of system design interviews over my career, and I’ve never had anyone use shuffle-sharding.

Exponential backoff with jitter is an effective retry strategy that helps avoid thundering herds, but most implementations only start applying jitter after the initial request fails.

There’s a subtlety in building AI agents that took me an embarrassingly long time to appreciate: your tool responses are prompts!

When comparing REST and GraphQL, it’s not just about payload size, flexibility, or tooling.

A few months ago Stripe announced their upcoming Order Intents API at Stripe Sessions.

Networks are unreliable.

I have some thoughts on the Google Cloud’s latest global outage!

Webhooks are surprisingly hard to process correctly.

Designing great resolver functions in GraphQL is quite unintuitive.

I’ve written about TypeScript enums in the past, and back then I encouraged the use of enums with string values instead of numeric ones due to the improved type safety and debuggability you get.

I typically describe my career as being the result of a lot of luck.

I’ve been extremely fortunate over the course of my career, and I’ve been able to learn from some truly incredible people working in software.

A common mistake in system design interviews is to underestimate a low but steady write rate.

Autofilling form fields has become an increasingly common and practical application of generative AI.

One of the powerful design patterns that can be employed within a GraphQL schema is the concept of collection lookups.