Frictionless, by Pagepro
Posts
How Uber Eats Delivers Billion Results & 1,145 PRs at Stripe

How Uber Eats Delivers Billion Results & 1,145 PRs at Stripe

Behind the scenes of Stripe, Uber Eats, and Salesforce

Chris Lojniewski
May 28, 2025

Hey there,

Last week, I turned 25. According to my Whoop, I'm 29, my ID says 39, but I feel much younger than that. While I wasn’t very active for most of my life, having kids changed my perspective. Now I'm motivated to stay active and ready for whatever sports or adventures they throw my way.

This week’s issue was inspired partially by mindset shifts and their influence on the industry. Read about how Stripe achieved 1,145 pull requests per day, ways to stop your employees from quitting, and training your junior devs for the AI-powered future.

We’ll also look at Uber Eats’ strategy for handling search queries, and celebrate the 30th birthday of JavaScript to finish it off.

Ready? Grab your coffee, relax, and enjoy Frictionless.

Chris Lojniewski | Pagepro LinkedIn

Wanna chat about the newsletter or have questions for me? Let’s connect!

www.linkedin.com/in/chris-lojniewski

In the Queue

Reduce Friction

A minimalistic purple and black graphic displaying bold text stats: "1,145 pull requests per day" and "99.99986% API reliability." The numbers highlight high developer productivity and near-perfect uptime, likely representing performance metrics for a robust engineering or API platform.

Source: Saile IT

Stripe: 1,145 Pull Requests Per Day

1,145 pull requests a day and under a minute of downtime all year? It sounds like a dream, but for Stripe, it’s business as usual. Their secret might lie in how the company shaped their engineering culture to support that kind of velocity.

Will Larson: Stuff I Learned at Carta.

Being a CTO is tough, and no one gets it better than someone who's been in the role. Will Larson shares what he learned during his time as CTO at Carta: how he approached messy architecture, rolled out better engineering representation, and dealt with the difficult parts of adopting LLMs.

This image is a comparison chart illustrating three distinct leadership styles: Absentee Management, Partnership, and Micromanagement. Each row describes how these styles behave in different managerial situations or characteristics. Here's a breakdown: Absentee Management is characterized by a lack of engagement, curiosity, and awareness. These leaders avoid involvement, fail to listen, don't set goals, and are often uninformed. Partnership represents an ideal leadership style where the leader is hands-on, collaborative, curious, informed, and actively problem-solves with the team. They share knowledge and context and create an environment of mutual trust. Micromanagement involves excessive control and over-involvement in minute details. These leaders do not listen well, give instructions without context, and pretend to know everything. They often make arbitrary decisions and hoard information. The chart uses clear, contrasting language to show how Partnership leadership strikes a balance between the extremes of being disengaged and being overbearing.

Source: Kim Scott on Medium

Why People Really Quit — And How Great Managers Make Them Want to Stay

Micromanagement can push people out, but so can being too hands-off. Enter “thought partnership”, a balanced way for leaders to stay involved without hovering. It could be exactly what we need to stop the brain drain.

Deepen Your Expertise

A split bar chart comparison showing document retrieval latency. The left graph (blue bars) represents unsorted index latency across various document IDs, with high variance and a retrieval time of 145ms. The right graph (green bars) shows sorted index latency with much more consistent and lower values, and a reduced retrieval time of 60ms. The header reads "Retrieval Latency Using Clustered Index By Location and Store."

Source: ByteByteGo

How Uber Eats Handles Billions of Daily Search Queries

When demand for the service exploded, Uber Eats' engineers had to act fast. To handle billions of daily search queries and 5x traffic spikes, they restructured their search into a multi-stage pipeline, balancing speed, accuracy, and growth.

5 Top CMS Platforms in 2025: Pagepro Picks

Picking the right CMS for your site can be a struggle. To make it easier, I’ve updated Pagepro’s list of the top 5 content management platforms: WordPress, Webflow, Drupal, Joomla, and Ubraco, covering what each one’s best at, and when to use it. Have a look!

Need Help Choosing the Right CMS?

We build custom CMS solutions that grow with your business.

pagepro.co/services/custom-cms-development

A code snippet written in Python is displayed. It defines two string variables: SYSTEM_PROMPT with the value "You are a cat.", and PROMPT with the value "In one sentence, what do you think about skateboarding?". A comment above SYSTEM_PROMPT reads # System prompt, and above PROMPT is the comment # Prompt. Below, a comment # Print Claude's response precedes a line that calls print(get_completion(PROMPT, SYSTEM_PROMPT)), which suggests sending the prompt to a language model named Claude for a response. The code humorously sets the AI's persona to a cat and asks for its opinion on skateboarding.

Source: Anthropic

Anthropic's Prompt Engineering Interactive Tutorial

If you’re looking to brush up on your prompt engineering skills, Anthropic has prepared a free guide on how to create optimal prompts for Claude. It will help you understand Claude's strengths and weaknesses and master building prompts from scratch.

AI Corner

A stylized illustration of speech bubbles designed to resemble submarines, each containing human profiles facing each other. The largest speech bubble has a propeller, suggesting movement or exploration, while others have elements like a periscope or headphones. The image uses a flat, minimalist style with blue and salmon-pink tones, symbolizing conversation and collaboration in the context of AI or tech.

Source: LeadDev

Tech CEOs Reckon With the Impact of AI on Junior Developers

As AI tools like Copilot become the norm, many tech leaders notice junior developers shipping code they don’t fully understand. They lose the ability to question the AI, and potentially erode their skills in the long term. If we don’t rethink how we train new engineers, we risk raising a generation that’s AI-dependent by default.

Salesforce Buys Data Management Major Informatica for $8bn: Here’s Why

Salesforce has been AI-first for a while. The $8B purchase of Informatica is another step in that direction. With its expertise in data integration and governance, Informatica will plug into Salesforce’s Agentforce platform for a smarter and safer automation for large-scale organizations.

New Capabilities for Building Agents on The Anthropic API

Exciting update from Anthropic! Their latest API release adds code execution, file handling, prompt caching, and easier system integration. We can now build agents that can reason and act in workflows.

Just Cool

Source: Deno

A Brief History of JavaScript

JavaScript might be old enough to be a senior dev by now, but it’s still powering most of the web. Take a look back at how the world’s most-used programming language has evolved over the past 30 years, and why it’s still going strong.

Let’s Stay in Touch! 📨

Do you have any comments about this newsletter issue or questions you want to ask? Drop me a message or book a meeting.

What do you think of today's email?

Your feedback helps me improve Frictionless.