🌐 Global Day of Coderetreat Toronto is back! 🌐 Join us at Thoughtworks Canada’s shiny new office in downtown Toronto for a day of deliberate practice in Software Development. 💪 Coderetreat is a day to exercise your software development muscles. 🚢 Away from the everyday

We’re all set for #coderetreat today. Are you? #gdcr2024 PS: don’t forget your laptop!

#LeadDevNewYork is a wrap! To all the organizers, speakers, and attendees — thank you for an unforgettable experience! 🙌

An under-estimated skill! #LeadDev2024

Jason Wodicka’s tools for giving help #LeadDevNewYork

“Iceberg requests” … im totally stealing this @yaphi1 ! #LeadDevNewYork

Product roadmaps really being, “Something with X” for decades. Ain’t that the truth. #LeadDev

Hypotheses, not features! #LeadDev

#LeadDev inclusive / welcoming vibes ❤️

I never ask for permission to refactor. But I often ask for help. Don’t confuse asking for help with asking for permission.

Silos happen when the person... - Good at the thing - Responsible for the thing - Doing the thing ... is the same person. Collaboration means that these can be different people. It requires that people can be trusted to do things right and ask for help when they need it.

Therefore, the core of what we do as software developers will not change. We won't be replaced. How we do software development, however, will change radically. And LLM-powered AI tools will be pivotal in this change. More on this to come! 9/9

Consequently, we’ll still need humans who can exist outside the system while simultaneously being deeply involved inside the system. Only in this way can software developers effectively reflect on and adjust both the problem being solved and how it’s being solved. 8/9

But I don’t see it going that way. At its core, software development is problem solving via iterative guesswork. Getting big guesses wrong is costly. To be good at smaller guesses, we need to be close to the work. We get closer when our feedback loops get shorter. 7/9

A programming language optimized for LLMs to write and maintain would be very terse, to maximize the lines of code that can fit inside of an LLM's context window. It could look like minified JS or something else entirely. Source code would become indecipherable to humans. 6/9

But one could argue the following: We don't need humans to understand source code. We could develop a new programming language optimized for LLMs to write and maintain. Let's play that out hypothetically. 5/9

And we've had these programming languages for over 60 years. Java's 29 years old and has had 22 major versions. With every update to its syntax, Java's authors have tried to make Java and the codebases written in it more understandable to humans (with rare exceptions). 4/9