Quick: Expose Every Endpoint!
When LLMs make code generation faster, how should API devs find the balance between speed and quality? Also in this newsletter: httpretty, Spectral gets forked, and the API Report Card
I find myself in a weird bind recently.
AI seems to be causing everyone to throw good engineering out the window. Design, planning, architecture, and efficient software are gone in the pursuit of AI. I feel this pressure even more so with API Design, from shoving all your endpoints into MCP, to shoving multiple endpoints into one endpoint to expose via MCP, and finally exposing your DB directly to any agent.
Wherever you sit on the AI spectrum, good design, sound technical choices and focusing on good outcomes and product will get you much further than ramming AI into everything as fast as possible.
Take Google Satellite AI for example, what a colossally stupid idea.
-- Alexander
Ship Reliable Webhooks for Your Customers
If you're getting pinged at 4 in the morning for failed webhook deliveries, this is for you. Svix handles retries, idempotency, and security. Qualified startups get $12,000 in free credits.
Get Started for FreeI am Forking Spectral
Maintenance on Spectral has been fading since SmartBear acquired Spotlight, and Kin Lane has the numbers to prove it. The repository is down from 157 issues closed in 2021 to only 10 so far in 2026. OpenSource is rough right now, but when backed by a large org, we should expect more. After a supply chain attack, Kin has forked the repo and kept his long-standing invitation to donate it open.
The API Report Card
All enterprise software graded on whether you can integrate with it or not. A human scores each platform across six areas: Existence, Access, Coverage, Auth, Docs & DX, and Stability. Fun fact: according to the site, most fail.
An HTTP Request's Journey Through Node
Ishtmeet Singh does not disappoint with this fantastic blog post. An interactive journey through a HTTP request in Node triggered by someone clicking buy. An interactive clock so you can track the milliseconds through every step. A wonderful read over my morning coffee. Thank you for this, Ishtmeet.
Fetch Needs Error Codes
For those of you who don't know the frustration of getting a TypeError in JavaScript on a fetch request, I envy you. James Snell covers this gap in fetch and explains that with error signalling in HTTP/2 and HTTP/3, it is time for fetch to support error codes. There is an active proposal with TC39 in Stage 1 right now.
httpretty
A delightful package that prints the HTTP requests of your Go applications in your terminal. Simple and easy to set up, with nicely formatted outputs. Support has recently been added for multipart bodies, colour status, and more.
From the Community
Articles written and shared in our free Slack community.
AI Agents Don't Break Your APIs
AI Agents aren't breaking your API's they can't compensate for your poor documentation like humans can. Another killer blog post from Matthew Reinbold on how documentation is affecting agents and humans alike, and a simple campaign to update your API docs is not going to fix it.
Support APIs You Won't Hate
When you become an member, you'll get access to members-only content while directly supporting our work. Your support helps us to keep making resources for the API community.
Become a member todayThanks so much to our members: Kin, Juxt, Alex R, Nolan S, Brandon K, Frank, James D, Bill D, and Abdelhadi D. Your support means the world to us!
✌️ Until next time,
Alexander, Phil & Mike
