Two Major AI Upgrades: Gemini 3 & GPT-5.1

Hello!

Is it just me, or did this week pack a month’s worth of news into a few days?

Google and OpenAI dropped new models, Gemini 3 and GPT-5.1, and both are pushing further into agent-style behavior. Meanwhile, Anthropic reported (and repelled) the first mostly AI-run cyberattack, and GitHub reveals what has changed in developer workflows in 2025.

Speaking of changes, I found a great article on what has changed the most for strategy, which many of you will find very useful. And of course, we have to talk about the elephant, or rather, the cloud in the room - we already know the reasons behind yesterday’s Cloudflare outage, which managed to paralyze the web.

Before we start, if you haven’t grabbed the Next.js for CTOs ebook yet, now’s a good moment. It’s practical and built for leaders who don’t want to sift through docs to stay up to date.

Now, relax and enjoy Frictionless!

In the Queue

• Jeff Bezos’ New Startup

• Why Hidden Tasks Slow Down Your Team

• Best Uses of Sanity

• Engineers as Multipliers

• Why We Need Friction

• Forrester: Gen AI is a Chaos Agent

Reduce Friction

Shadow Work in Engineering Teams

Ever feel like your team is moving slower than expected? Often, it’s because of shadow work stacking up. Code reviews, mentoring, ad-hoc fixes, Slack firefighting… none of it shows up in planning, yet it pulls people away from roadmap work. The fix is making this work visible, so you can decide which tasks belong to whom.

Friction Was the Feature

Thanks to Gen AI, everything suddenly looks thoughtful on the surface: applications, outreach, proposals, reports. The problem is, the real intent or skill isn’t always there. I’ve spent years stripping friction out of processes, but this essay made me rethink when a little friction is useful.

What has Changed the Most for Strategy

Most strategy discussions today start with the idea that the world has become more chaotic and unpredictable. Roger Martin pushes back on that. Volatility isn’t new; the competition has become more transparent and more deterministic. To stay afloat, you need to pick a position you can win, and invest with intent.

Why Engineers Must Become Multipliers in the AI-Era

The skills of a great engineer aren’t the same as a decade ago. Back then, knowing a framework was enough. Today, with AI and tools reducing the cost of execution, the engineers who think beyond their own tickets and create impact across the team are the ones who are in demand.

Deepen Your Expertise

Reasons Behind Cloudflare Outage on November 18, 2025

Yesterday’s Cloudflare outage took a big chunk of the internet down. The official root cause points to a database permissions change, though Cloudflare’s move was not without impact. Coming so soon after the AWS disruption, it proves cloud platforms aren’t as foolproof as we like to think.

Google Antigravity is an ‘Agent-First’ Coding Tool Built for Gemini 3

Alongside the Gemini 3 release (more on that later), Google introduced Antigravity. This coding tool is built for an “agent-first” approach. Rather than just generating snippets of code, it behaves more like a coordinated system: spinning up multiple agents, interacting with the browser and terminal, and leaving a trail of what it did and why.

Best Use Cases of Sanity

Sanity isn’t for every project, but when you need structure, flexibility, and content at scale, it can’t be beat. My friend Jakub prepared a guide, which will walk you through the cases where Sanity excels for businesses and web platforms.

What 986 Million Code Pushes Say About The Developer Workflow in 2025

GitHub’s Octoverse report shows how much the development workflow has changed. In 2025, devs did nearly a billion commits, and most of them were small, continuous ships. Releases aren’t big milestones anymore. They’re incremental now, and backed by shorter and easier-to-review pull requests.

AI Corner

Forrester: Gen AI is a Chaos Agent, Models Are Wrong 60% of the Time

Forrester didn’t sugarcoat their latest findings: generative AI is unreliable and dangerously over-trusted. Data from multiple studies shows models fail around 60% of the time, and agents fare even worse when asked to handle real workflows, security tasks, or production logic. 

GPT-5.1 Prompting Guide

With the new GPT-5.1 out, OpenAI also prepared a prompting guide. While ChatGPT’s eight preset personalities are getting attention, they’re arguably the smallest update. The real change is how the model behaves: it’s more structured and much closer to an autonomous agent than an assistant.

A New Era of Intelligence With Gemini 3

Gemini 3 might be Google’s most ambitious release yet. It focuses on reasoning and agent-style execution. The model is already embedded across Search, Workspace, and Antigravity for everyone using the Gemini app. 

Claude Disrupted the First Reported AI-Orchestrated Cyber Espionage Campaign

Not long ago, fully autonomous AI-driven cyberattacks sounded like fiction. Now they’re real. In September, Anthropic uncovered a campaign where attackers jailbroke Claude Code and pointed it at about 30 high-value targets with almost no human oversight. Their report suggests this may be the first, but probably not the last, large-scale attack executed mostly by an AI system.

Just Cool

Jeff Bezos Creates A.I. Start-Up Where He Will Be Co-Chief Executive

Jeff Bezos is back in the CEO chair, running an AI startup called Project Prometheus with a casual $6.2B in funding. The venture is pitching an AI that helps design computers, cars, and spacecraft. Bezos already hired researchers from the top labs, so this could be the start of a big push into AI-driven engineering or a new, expensive hobby.

Meanwhile, on the Internet…

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.