Logo
Join the CTO Newsletter for free!
 
Amazing CTO Logo

Amazing CTO | More happiness and success
🚀 125.4

by Stephan Schmidt

Happy 🌞 Sunday,

Welcome to my opinionated newsletter. This week’s insights:

  • 🧲 INSIGHTS: Why Retention Is So Hard for New Tech Products
  • 🧩 Why Do Software Developers Love Complexity?
  • 🧙‍♂️ Le Chatelier’s Principle: Why Your Organization Pushes Back
  • 🤖 AI Coding - it’s not working
  • 🩺 Quarterly System Health Check: Beyond the Dashboard
  • 🚀 Qwen Ecosystem: The new kid that takes over the block
  • 🧠 Why Language Models Hallucinate
  • 🧚‍♂️ Ex-Google Exec: The AI Jobs Myth
  • 🛡️ BackupGuardian: Validate Your Database Backups Before It’s Too Late

Good reading, have a nice Sunday ❤️ and a great week,

Stephan
CTO-Coach and CTO-veteran

Need support as an engineering manager? Thought about coaching? Let's talk—I helped many CTOs and engineering leaders with growth and making the right decisions under pressure, I can help you too.
🎁

If you only read one thing

Why Retention Is So Hard for New Tech Products (5 minute read)

Must read. “You can’t fix bad retention. No, adding more notifications will not fix your retention curve. You can’t A/B test your way to good retention” I know many of you only consider technology your domain, but you need to co-own product. Take a stand for products and features that work, don’t let people build things with engineers - you’re responsible for - that don’t work. The post has many more great insights, “Crazy viral growth with shitty retention fails. We’ve run this experiment many many times already, across multiple platforms and categories” If there is one thing I want to shout at founders every time, You can’t fix missing PMF with marketing! Work hard on PMF, and if you think you have it, I’d bet you don’t, work harder - hint,hint: retention.

https://x.com/andrewchen/status/1965419750525431873


🚀

Stories I’ve enjoyed this week

Why do software developers love complexity? (4 minute read)

Why do developers love complexity? There are many reasons, this is one of them. “Complexity also signals effort, expertise, and exclusivity. If you struggle to understand it, your brain rewards you with awe” Great article to understand what drives developers to love complexity - and creates problems to you, the manager. “Complexity shouts, ‘Look at me!’, while simplicity whispers ‘Did you notice?’.” IMHO the main driver is, developers are bored by the features they need to implement. Those writing a distributed real time syncing database don’t add a myriad of frameworks on top. Those bored by adding the third blue button to the application after rewritinge the same two forms - again! add as many interesting frameworks as they can to not be bored out.

https://kyrylo.org/software/2025/08/21/why-do-software-developers-love-complexity.html


Magical systems thinking - Works in Progress Magazine (18 minute read)

I see many CTOs struggle with transformations and reorganizations. Or with the simple thing of introducing a process into a system. Then I came across this article: “Le Chatelier’s Principle, the idea that the system always kicks back” and “Le Chatelier’s Principle provided an answer: systems should not be thought of as benign entities that will faithfully carry out their creators’ intentions. Rather, over time, they come to oppose their own proper functioning.” I’ve experienced this again and again. Systems are not clockwork. Not your complex deployments and not your product development organization. But there is much more in the article to help you understand.

https://worksinprogress.co/issue/magical-systems-thinking/


AI Coding (4 minute read)

Why AI Coding is not working? According to the author, it’s because “The best model of a programming AI is a compiler.” I’ve been saying this for a long time now, one-shotting MVPs and prototypes is the best way for agentic coding (Augmented coding is different). But where I see this as a good thing (excellent thing!), the author thinks this is bad. The fallacy is right there “While noobs and managers are excited that the input language to this compiler is English, English is a poor language choice for many reasons.” (Not sure if I’m the noob or the manager :-) If English is bad, how can a ticket describe what to build? Perhaps it’s not English, but you.

https://geohot.github.io//blog/jekyll/update/2025/09/12/ai-coding.html


My Quarterly System Health Check-in: Beyond The Dashboard (10 minute read)

Many, many, many, many good questions to ask in a quarterly review of your organization, e.g.

  • Will we deploy during peak hours, or during Friday nights?
  • Can we refactor anything as long tests pass? What are we afraid of refactoring?
  • Do we ignore alerts because they’re noisy?
  • Do people know (by memory) the approximate p99s of critical operations?
  • Do incentives reward short term velocity over long term maneuverability? Many CTOs don’t do this. A quarterly review like this has agency to move your organisation forward - instead of being reactive to other people’s wants. And this is one of your core problems - you’re reactive not following your plan. READ THEM ALL, they are that good.

https://blog.nilenso.com/blog/2025/09/05/my-quarterly-system-health-check-in-beyond-the-dashboard/


Qwen Ecosystem Expands Rapidly, Accelerating AI Adoption Across Industries (27 minute read)

Several people I know use Qwen coder for local AI. I think the battle is not fought out of cloud vs. local AI (currently cloud seems much better), but Qwen is a serious contender for local AI for developers (Looking forward to Apple releasing a M5 studio for local AI, the 395+ seems to be limited by memory bandwidth, which looks to be much more important for AI than for gaming). “Qwen3 Launched Models Optimized for Apple’s machine learning framework NVIDIA, AMD, Arm and MediaTek integrate Qwen3, unlocking enhanced AI performance across platforms”

https://www.alizila.com/qwen-ecosystem-expands-rapidly-accelerating-ai-adoption-across-industries/


Why language models hallucinate (12 minute read)

Important read - I see too many people getting hallucinations wrong. “Hallucinations are plausible but false statements generated by language models. [..] If you do not know the answer but take a wild guess, you might get lucky and be right. Leaving it blank guarantees a zero.” and “Nonetheless, accuracy-only scoreboards dominate leaderboards and model cards, motivating developers to build models that guess rather than hold back.” Hallucinations are the most probable filling of gaps in reality. Not right, but highly plausible.

https://openai.com/index/why-language-models-hallucinate/


Ex-Google exec: The idea that AI will create new jobs is ‘100% crap’—even CEOs are at risk of displacement (4 minute read)

I always get lots of flak when I say this, but no one yet could convince me (try and reply!) otherwise where the jobs come from when the majority of knowledge workers are laid off. Or perhaps nothing happens, because people silently just have more meetings to waste the time gained by AI.

https://www.cnbc.com/2025/08/05/ex-google-exec-the-idea-that-ai-will-create-new-jobs-is-100percent-crap.html


pasika26/backupguardian: Validate database backup files before migration to prevent costly failures (3 minute read)

Number one mistake CTOs make when implementing a backup strategy: Not testing backups. I have seen several CTOs who found out their backup was corrupted when they wanted to restore it. Minor: They also don’t know how long it takes to restore it, so if something happens, what are you going to tell the CEO how long restoring the database will take? Test restoring your database periodically (start now, no now, right now, on a Sunday? Yes!). And check that your backups are ok. A client of mine recently had to restore a database from backup on a Sunday, but not by their free will.

https://github.com/pasika26/backupguardian


Upcoming Events For Engineering Managers

  • September 23-24, CTO Craft Con: Berlin
  • September 25, Build & Lead – Why Everyone Hates Training
  • September 26, T3 Berlin 2025
  • November 3–4 LeadDev, Berlin
  • November 5-6, code.talks, Hamburg <- ❤️ I’m there
  • November 28, TechLeadConf, London

What is happening with Stephan?

I struck luck this week, with a post on LinkedIn that generated ~80,000 impressions. The driver seems to be reposts, it’s great if you can trigger them. Because people don’t talk enough about numbers, also: those impressions drove around 150 new followers. Sad news: We also made the decision to shut down Inkmi :-( On the bright side, there is a huge demand for my AI workshop, CTOs have trouble convincing and motivating their developers to use AI (because, what is in there for developers? - therefor, the workshop).

AI Workshop
Join the CTO newsletter!
Impressum