miguelpaoc

GPT-5 API's New Features for Developers, with Simple AI SDK Examples

Posted by miguelpaoc on Friday August 15, 2025 @11:32AM
from the self-post dept.

GPT-5 introduced three new API features: verbosity, reasoningEffort, and a unified model lineup. That gives developers more control over responses and costs.I wrote a concise guide with runnable AI SDK examples for each, so you can try them in under a minute.Blog link: https://medium.com/@miguel.paracuellos.inf/gpt-5-for-developers-3-new-api-features-you-should-be-using-today-d3886b5730eeI’d...

( Read More... | no comments | on HN )

Re:Introducing gpt-oss on Wednesday August 06, 2025 @03:25PM (#44816506)

I wrote a detailed breakdown of the models and why this release is a big deal — especially for devs who want to run or fine-tune LLMs locally: https://medium.com/@miguel.paracuellos.inf/openai-just-went-...

Curious how others are thinking of using these — would love to hear your setups.

Re: on Tuesday August 05, 2025 @10:25AM (#44798364)

At work, we’ve been struggling with some really slow Postgres queries in our SaaS platform - especially around reporting endpoints that hit orders, products, and user data. I spent a few weeks digging deep into performance (EXPLAIN ANALYZE + buffers became my best friend), and ended up applying 5 specific tricks that gave us up to a 10× speed-up.

I put together a write-up with the ones that helped most: lateral joins with json_agg, partial indexes, filter aggregates, etc. It’s focused on practical changes that have actually had an impact.

Curious how others have tackled Postgres performance at scale — I’d love to hear what’s worked for you, or anything you’d have done differently.