{"canonical":"https://www.readysetcloud.io/newsletter/233/","categories":null,"contentText":"🦸 Community Superhero Our community superhero this week is Adolfo Ochagavía, software engineer and consultant. Adolfo is a great blogger, talking about everything from building a compressor from scratch to JIT calculators. He has been contributing to open source projects and the Rust ecosystem for over a decade, and shares his lessons on his site. Thank you for everything you do, Adolfo!\n💯 Spotlight I’m kicking us off this week with a cool engineering blog from Uber Eats about how they rebuilt their search pipeline to decrease latency by 50%! The post is more story-driven than it is technical jargon, which helps explain the troubles they were having and how they mitigated each one. There’s a lot that goes into search of any kind, and I found it incredibly fascinating to see how they approach it at Uber. They even added a roadmap section to the end to tell us the methods they’re investing in to continue making improvements.\n🔥 My Favorite Content What do you do when an old project starts costing too much to run? You make a serverless version of it, of course! Which is exactly what Paul Santus did with SPIP, an old-school French CMS. Paul walks through his architecture choices, showing the 1 for 1 mappings of how it fit into an AWS serverless model. He covers the tricky pieces to his conversion and most importantly, the tradeoffs he had to make. Paul also shares the code in case you want to try it yourself. I like this article because of how well he does isolating the components of a legacy system and hot-swapping the infrastructure underneath with minimal UX changes.\n🤖 robot voice It’s easy to fall into the let’s make that an agent trap. With the frameworks and tooling available these days, it’s often easier to build an agent than it is to build an orchestrated pipeline that would solve the same problem, but deterministically. This was the main argument from James Anderson (who I’m not fully convinced is a person, he might just be a bot. The content is certainly written by an LLM.) in his post last week which cleverly calls most AI agents ‘if statements in a trench coat.’ He makes some good points about what to watch for when you should just be writing code or deferring to the judgment of an LLM. 404 · personality not found I published an article last week discussing using payment as authorization. I emphasize that we need to start thinking about API design differently with AI agents becoming increasingly important consumers of our products. Instead of going through lengthy registration and onboarding flows, I propose forgoing it in favor of pay-as-you-go operations on your API. The article discusses how to build with the x402 protocol using middleware I wrote for Lambda Powertools, but the takeaway I want you to have is the paradigm shift.\n💡 Tip of the Week I dance around it in this newsletter a lot, but there are some new trends in our industry that aren’t often called out but should be. Here’s Gergely Orosz pointing them out succinctly and masterfully. I agree with everything in this list.\nLast Words Here’s my monthly reminder to keep your voice in the content you create! It’s absolutely ok to use an LLM to help you create something, but using it to do everything will lose trust with your readers. When everything sounds the same, none of it is memorable - which is the whole point of content creation. You want to teach people a new skill or show them a new way of thinking, and where’s the value if it goes in one ear and out the other? So please stay creative and bring all of your human-ness to the table!\nWhat did I miss? What made you nod along (or 🙄)? Hit reply if you’re reading the email. Prefer socials? Ping me on Twitter, LinkedIn, or email.\nHappy coding!\nAllen\n","date":"2026-09-14T00:00:00Z","description":"We go heavy on engineering this week","image":"https://www.readysetcloud.io/images/newsletter.png","inLanguage":"en-US","lastmod":"2026-09-14T00:00:00Z","readingMinutes":4,"section":"newsletter","tags":null,"title":"Doubling Uber Eat's search speed","url":"https://www.readysetcloud.io/newsletter/233/","wordCount":660}