- Frictionless, by Pagepro
- Posts
- Senior Devs Ship 2.5x More AI Code than Juniors
Senior Devs Ship 2.5x More AI Code than Juniors
Claude Is After User Data & Learning from Uber’s Mistakes

Hello!
Some weeks in tech remind me of triathlon training.
You don’t win by chasing shiny gear; you win by knowing when to push, when to hold back, and when to adapt.
Just look at the insights from this week: Amazon’s CTO says there are no “magic” technologies. The criminal case of Uber’s ex-CSO shows that one bad decision can cost you more than any bug. And the Salesloft data breach proved that a single weak SaaS link can put hundreds of companies at risk.
On the adaptation side, senior developers ship 2.5 times more AI code than their juniors, and Microsoft debuts its own in-house AI model, taking a step back from their partnership with OpenAI. Just in time, as the costs of cutting-edge AI keep growing, with many startups struggling to stay afloat.
And if all this sounds like a lot right now, take a moment to admire the code from the Apollo 11 mission, available on GitHub.
Grab your coffee and enjoy Frictionless!
In the Queue
Reduce Friction
"There Are No Magic Technologies" Amazon CTO Shares His Insights
When Werner Vogels speaks, you get decades of hard-won lessons from running Amazon’s backbone. Learn why FOMO is your worst advisor, planning for failure should be a priority, and building is always better than buying.
C-Suite Lessons From Joe Sullivan and the Uber Data Breach
57 million stolen records in the data breach was a blow to Uber’s reputation, landing their CSO, Joe Sullivan, in court in a first-of-its-kind criminal case. Sullivan’s disastrous decisions are a great lesson for all C-Suite leaders: security breaches don’t always happen because of technical failures.
Seeing Like a Software Company
Your Jira board isn’t the full story. The messy Slack pings, quick calls, and favors keep things moving more than any “official process.” Sean Goedecke explains why ignoring this ‘illegible’ work puts companies at risk.”
The Four Styles of Confidence on a Team
Confidence is contagious. The wrong kind drags everyone down. Overconfident, underconfident, midlevel, or proportionally balanced: see out where your team sits and what it means for performance.
Deepen Your Expertise
The Impact of the Salesloft Drift Breach on Cloudflare and Our Customers
Salesloft’s integration with Salesforce opened the doors to the company’s data, but they weren’t the only ones affected. Cloudflare’s account of the Salesloft Drift breach shows how a single weak link in the SaaS chain exposed hundreds of organizations and their strategy for addressing the issue.
Secret to Scaling a 100K+ Page Medical Platform
OmniaMed Communications, a trusted platform for the UK’s medical professionals, had a problem. A legacy system couldn’t support scale, speed, or sub-brands for their newest project, GPnotebook. The solution was quite simple: a Next.js and Sanity-powered migration.
First Attempt Will Be 95% Garbage: A Staff Engineer's 6-Week Journey with Claude Code
The best engineers I know don’t expect perfection on the first try. They expect to learn. Vincent Quigley from Sanity’s team details his first messy attempts at integrating and how pushing ahead helped him hand over 80% of initial implementations to AI.
Next.js Middleware – What Is It and When to Use It
Every app has hidden rules running in the background: who gets access, which pages redirect, what’s logged before a request even lands. In Next.js, middleware is where those rules live. My good friend Jakub explains when it’s worth adding that extra layer, and when you’re better off skipping it.
AI Corner
Cutting-Edge AI Was Supposed to Get Cheaper. It’s More Expensive Than Ever
Although the AI costs were supposed to go down, most companies watch them spiral instead. The “smarter” systems of today burn tokens just to think, and while big companies can weather the storm, startups struggle to keep their margins. The “dumber” but cheaper models might be a solution to this problem.
Microsoft Signals Shift from OpenAI with Launch of First In-House AI Models for Copilot
For years, Microsoft’s Copilot leaned on OpenAI under the hood. Now the company is testing models built entirely in-house, and the first results are promising. In the face of their cooling partnership with OpenAI, Microsoft might be looking for more technological independence.
Vibe Shift? Senior Developers Ship Nearly 2.5x More AI Code than Junior Counterparts
Fastly’s latest survey shows senior engineers ship almost 2.5x more AI-generated code than juniors. It might seem like they trust the AI more, but seniors are better at spotting mistakes in the code, letting them use AI more confidently. Will it mean faster development?
Anthropic Users Face a New Choice – Opt Out or Share Your Chats for AI Training
By September 28, all Claude users have to make a decision: share their chats for training or not. This new decision is a big change from Anthropic’s previous policy, where they did not use user data to improve their AI. Many users remain obvious to the upcoming change, and the questionable opt-out design drew criticism of the company’s practices.
Just Cool
Google Will Not Be Forced to Sell Off Chrome or Android, Judge Rules in Landmark Antitrust Ruling
A federal judge ruled that Google won’t have to spin off Chrome or Android. The company’s victory has been somewhat dampened by another ruling, which states that it can’t keep leaning on exclusive distribution deals to hold its search monopoly in place.
Apollo 11 Mission Code
Ever wanted to browse the code that literally took humans to the Moon? This repo holds the original Apollo 11 guidance computer source code for both the Command Module (Comanche055) and the Lunar Module (Luminary099).
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 |