Amazing CTO | More happiness and success
🚀 76.3

by Stephan Schmidt

Happy 🌞 Sunday,

this week is different. I have a video for you by the excellent Adelina Chalmers who has great insights into the role of CTOs.

Everyone needs more money. What if you could hire more people for the migration? The big refactoring? AI? It is easier for Marketing to get that money than for the CTO.

Here comes help

“How to get your board to say “yes” to engineering spend!” (or the CEO!) I wrote about the topic in the past too, it is so important.

(Key phrase: “See how CTO got $5M for a technical redesign.”)

https://geekadelina.gumroad.com/l/yes

With the coupon code AMAZINGCTO you’ll get the video for £5 instead of £50 until the 21st of April.

Some form of disclaimer: As you know, I remove all tracking as far as possible (I also pay an expensive mail provider to not have tracking), and I don’t take any money, sponsorships or ads. So I also don’t get anything for the video. I include it, because Adelina as always is a great help to CTOs.

And you need more money.

Welcome to my opinions. This week’s insights

• 👷‍♀️ 80% of a winning strategy: Just avoid blundering
• 📐️ 04/24 Software Architecture Trends Report
• 🎸 The CTO Blues Song

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

Stephan
CTO-Coach and CTO-veteran

If you only read one thing

Avoid blundering: 80% of a winning strategy (24 minute read)

Love the article, not only is it a joy to read, it also writes about chess — which recently made the term blunder mainstream. Are there CTO blunders? Unnecessary technical complexity is a CTO blunder. The article mentions one blunder, something I tell every founder, because they don’t do it, “Didn’t talk to customers (and listen)” Why is talking so difficult? Once I’ve listened to two developers discussing for half an hour what the product manager meant with a point in the ticket. The product manager was sitting next door!

https://longform.asmartbear.com/avoid-blundering/

Image of the week

InfoQ Software Architecture and Design Trends Report - April 2024 (19 minute read)

Early majority for “Modular Monolith” a.k.a. Modulith. Currently writing https://www.inkmi.com as a modular Modulith with NATS.io to communicate between internal modules - love it so far and should be easy to move to microservices within some days if needed for scaling people. It seems the Modulith is going mainstream. The pendulum swings.

https://www.infoq.com/articles/architecture-trends-2024/

Song of the week

The CTO Blues

♫♫ Listen to the CTO Blues and sing along! ♫♫

Got a million problems, they're all on my plate
Running this company, it's an endless debate
Code's crashing, servers are down
And I'm drowning in bugs, can't turn it around

Meetings all day, never enough time
Trying to keep up, but I'm losing my mind
Tech debt piling up, deadlines getting missed
Wish I could escape, just one day of bliss

(lyrics and melody created by an AI)

https://suno.com/song/b88f120f-3254-4054-b195-ace952d132ac/

Stories I’ve enjoyed this week

The remoteness of remote work (22 minute read)

Very long read about remote work from a CTO perspective. My first read about long-term reflections of remote work: “It worked out great in the first year, started losing its sheen in the second year, and became detrimental to creativity and collaboration by the third year. “ Written from the perspective of a CTO, a manager. I wondered what the people think. The desire for more independence and the desire for personal interactions, we still have no clue what will happen: “What happens when an unstoppable force meets an immovable object?” ( https://en.wikipedia.org/wiki/Irresistible_force_paradox )

https://zerodha.tech/blog/remoteness-of-remote-work/

auto-code-rover (8 minute read)

One of the milestones of AI in software engineering is the automatic fix of bugs. State of the art: “AutoCodeRover is a fully automated approach for resolving GitHub issues (bug fixing and feature addition) […] On SWE-bench lite, which consists of 300 real-world GitHub issues, AutoCodeRover resolves ~22% of issues” Nearly a quarter of GitHub issues automatically fixed by LLMs.

https://github.com/nus-apr/auto-code-rover

Three major LLM releases in 24 hours (7 minute read)

Google Gemini Pro 1.5, OpenAI GPT-4 Turbo, Mixtral 8x22B. Three large language models in less than 24 hours. Things are speeding up. LLMs will become a commodity, what is the next frontier on top of LLMs? Will a LLM be like Linux?

https://simonwillison.net/2024/Apr/10/weeknotes-llm-releases/

Securing your Codebase (11 minute read)

Too many CTOs neglect security. And yes, don’t overinvest in security when you haven’t got customers is the right thing to do, but when you grow revenue and customers and employees, the impact of a security breach grows as fast as your company. If you grow exponentially, so does the risk impact. Great article with some basics. Creating awareness with developers is the first step. Show them this article.

Amazon cuts hundreds of jobs in cloud computing unit 7 : Over-hired or preparing for the great-AI layoff? This is what I wonder each time I read about layoffs these days.

https://tattle.co.in/blog/2024-04-03-securing-feluda-pt1

Fear makes you a worse programmer (5 minute read)

Some good points from Julia. I myself could not deploy without a lot of tests to feel safe. I could not radically change code without unit tests. Running unit tests and seeing them green gives me a good feeling. Often I see code bases which have not been refactored along the way, and developers now want “a refactoring” (misunderstanding what Refactoring is). These code bases also have low test coverage, or no tests at all. Write tests to continuously refactor small things without fear, so you’ll not have to “refactor everything”.

https://jvns.ca/blog/2014/12/21/fear-makes-you-a-worse-programmer/

How to send progress updates (3 minute read)

Progress updates are great. Self-promotion is important, otherwise your boss thinks you’re not working (unless you have worked on and invested in trust). Are they enough? No, of course not. But they set a baseline. Sending them is great too, don’t waste precious face time or 1on1 time on progress updates. Also: Not enough people do external and internal feature marketing for alignment and make people use features. Why develop features people don’t use in the first place? Progress updates help here. And this is something to try: “For example, if you send project updates every Tuesday, they will seem transactional, and no one will read them. If instead you send updates every 2–3 weeks, your audience will look forward to them because they’ll assume you have something new to say.”

https://www.spakhm.com/updates-howto

Parental control app exposes live GPS locations of kids on internet (0 minute read)

Not only is it tragic that a parental control app releases kids chats on the internet, but the reason was unnecessary complexity of its architecture, “that the app’s developers failed to configure authentication for their Kafka Broker Cluster” I guess you really need your Kafka and you’ve configured it correctly.

https://cybernews.com/security/parental-control-app-kidsecurity-data-leak/

What Is ChatGPT Doing … and Why Does It Work? (116 minute read)

Excellent—but very long and deep, but understandable and well written—article on how ChatGPT works.

https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-why-does-it-work/

The 37signals Guide to Internal Communication (17 minute read)

Many good points and an interesting read. Some nuggets like “Poor communication create more work.” and there are more. Communication is what we do most as managers; most companies have no communication guide though. During onboarding, I explain to all new hires how to communicate. “Don’t reuse old emails with old topics just because you don’t want to create a new email—at least change the topic!”

https://37signals.com/how-we-communicate

Engineers Pinpoint Cause of Voyager 1 Issue, Are Working on Solution (5 minute read)

When I debugged my first production system that wasn’t in the office, but in a data center, I feared crashing the machine or SSH and would need to go to the data center. Now assume the computer you debug is 24.3 billion kilometers away.

https://blogs.nasa.gov/voyager/2024/04/04/engineers-pinpoint-cause-of-voyager-1-issue-are-working-on-solution/

Oh, it’s a lot longer story than that. I worked as SGI from just around its peak… : Hacker News (16 minute read)

Why did SGI end, the inventor of GPUs and the spawn of the graphics card industry (SGIs DNA ended up in today’s AMD)? ” Because some engineers decided to implement a cool theory and were allowed to do it (no clipping, recursive rasterization, hilbert space memory organization).” So “Nothing ever made a deadline, they kept slipping by months.” And then SGI died. Which reminds me, in the 90s, a friend, and I tried to buy a SGI machine on every tech show we attended, without success of course. Today every cheap GPU is better at 1/10000th of the price of an SGI machine.

https://news.ycombinator.com/item?id=30920824