If you only read one thing
Interesting take on how AI can help with code, and an interesting take on self-healing code: “As of today, code-change authors at Google address a substantial amount of reviewer comments by applying an ML-suggested edit. We expect that to reduce time spent on code reviews by hundreds of thousands of hours annually at Google scale”
Picture of the week
Tweet of the week
I don’t agree with Paul Graham here. But I think the opinion is important enough for you to hear
“I’ve talked to multiple founders recently who have changed their minds about remote work and are trying to get people back to the office. I doubt things will go all the way back to the way they were before Covid, but it looks like they will go most of the way back.”
Stories I’ve enjoyed this week
Super interesting! As an industry, we have not enough evidence and facts, but shoot from the hip. They’ve scraped data from LinkedIn to see how many engineers, product managers, and data scientists a company has. Then they group it into 3 groups: Engineering First, Data First, and Product First. “With 37% of employees in engineering roles GitHub comes out top of the list. Companies with the relatively smallest engineering teams have around 12% of total employees in engineering roles.”
Many many insights into the CTO role. “Think beyond the IT function. Define your technology agenda to advance a greater sense of purpose for your entire organization and ecosystem.”
“If every time you write a blog post it takes you six months, and you’re sitting around your apartment on a Sunday afternoon thinking of stuff to do, you’re probably not going to think of starting a blog post, because it’ll feel too expensive.”
For this reason, things should happen fast in your company. Instead of a project taking weeks, focus hard and make it happen in days. Instead of hiring someone in 4 weeks hire someone in 4 days. Make the decision on hiring on the day of the interview, send out the contract, done. From interacting with companies and seeing the speed of their replies, I know how they work internally. From my interaction with CTOs, on how fast they are I can already see what problems they have. “I’ll think about the coaching and come back to you next week”* “Nooooo, tomorrow!” I want to shout. “if there’s something you want to do a lot of and get good at—like [..] fix bugs—you should try to do it faster.” I’ve seen bugs lying around for months.
Gorilla is a new language model specialized in using APIs. The paper gives several examples where GPT-4 and Claude fail, but Gorilla creates the right code. I myself had lots of problems with GPT-4 generating code and calling APIs that don’t exist, or no longer exist. Hope this flows back into GPT and open models like Falcon.
These are metrics a founder should know before talking to investors. And while they are interesting for a CTO (you should know these too, you should know the business side of things), I’m more interested in the concept:
I sometimes ask CTOs how many open bugs they have? They don’t know. Or how long is their lead time? They don’t know. Or how long code reviews wait? They don’t know.
If you don’t know your numbers, you’re not in control. We can debate if you should be in control, but the CEO for sure thinks you should be.
Many good ideas. We still haven’t mastered the language of GPT and use not enough ideas to get the best out of it. For example, “Produce cheat sheets: Write a cheat sheet for Markdown formatting.”
Engineering Management is hard. So I take every trick I can find to make it easier. Like this one “What are you most concerned about?: This can provide an engineer the room they need to bring up an issue that’s been bothering them. You may need to ask several questions of the form “tell me more about that” to get to the root of the issue.”
“For Keanu Reeves, it was obvious that the most productive path was to focus all of his attention on a single goal: becoming really good at Stahelski’s innovative brand of gun fu” Contrast this with productivity in startups: *“busyness and multi-faceted activity.” John Wick productivity is the true core of agility: Iterations (no one I’ve met does iterations, they all haste from feature to feature instead of iterating to perfection)
😂 “But what about the dumb and energetic? “Those,” Napoleon replied, “I shoot.””
When I look at the code of companies, I see all the blissful new ideas. “Sooner or later, the blissful productivity gives way to something that feels much more like… work. More like a grind.” But they are never complete, not all the code is migrated, that new code idea is only in one module. From the usage of hipp libraries I can see when some code is written. Finish what you’ve started. “Finishing requires courage”
The method (that also works for me to get things done):
What is a fractional CTO? I have been one several times, without knowing! Huge fan of the concept for early phase startups.
In 1994, Amazon was looking for its first developer. See where they are now.
On how much Go is shaped by the needs and experiences of Google. And I might add, if you’re not Google, Go might not be for you (Go is my goto language of choice)
Start with the why. “The first step I would recommend is to understand why you want to become a manager.” and something many new CTOs do not understand *“Firstly, embrace the change in mindset. Your success is now determined more by the success of your team than by your individual contributions.” They want to do too much themselves instead of empowering their team.
|Know someone who could benefit from this?|
Please forward the newsletter ❤️
Received this newsletter from someone? 📩 Just Join!