Nobody wants to touch the pricing code. Then two developers pick one rule, write down a few examples, and leave the next person a safer place to start. That’s how I like to add features to legacy code. https://t.co/31M11NGl3B
“What’s the most important feature?” The first time I ask, the answer is often “everything.” Come back two weeks later with something that works, and the answer can change. That’s a small dev team advantage worth building around. https://t.co/4qQC5zFFC8
Before you let AI clean up that strange conditional, ask why it's there. It might preserve a business promise nobody remembers.
This week's video: recovering the knowledge in old code before changing it.
https://t.co/yASAinqxpe
AI writes code faster. Why does a small change still feel dangerous?
A failing test might expose broken behavior—or a dependency on an internal detail. The question is: what promise is it protecting?
This week's video: https://t.co/wPcdcQmwUY
AI implements a refund rule: cancel seven days before the workshop. The tests pass.
Seven days before what? Midnight? The start time? Whose time zone?
Are You Outsourcing Code or Understanding?
https://t.co/sribPbzTWA
AI writes code faster. Who reviews it—and who designs what comes next?
One experienced developer. Two queues. Less time to help others learn.
EP02: why understanding can become the bottleneck, and why building together matters.
https://t.co/JtE504lhi5
“How can I add more features so people will use this?” already assumes why they aren't using it.
In EP01 of Agents and Agency, I explore how AI can help us question that assumption—and test what we learn.
Watch: https://t.co/KjP4mNRMe4
If AI can help us create almost anything, what should we actually choose to create?
The future does not belong to better prompt engineers.
It belongs to better thinkers.
Final episode in this Thinking with AI arc:
https://t.co/39lIBoabfb
When answers become cheap, what becomes valuable?
For me, working with AI has made one thing clearer:
The quality of the question changes the quality of the collaboration.
New Thinking with AI episode:
https://t.co/xxPuje4NtM
Why does encapsulation matter again in the AI era?
Because as systems get larger, we cannot trace everything.
We need responsibility.
We need boundaries.
We need contracts.
New Thinking with AI episode:
https://t.co/mvRxe1FfXN
If AI can write code faster than ever, does software craft still matter?
I think this is the wrong moment to abandon craft.
Faster code is not the same thing as clear, validated, valuable software.
New Thinking with AI episode:
https://t.co/L0FsV2y67O
What should you never delegate to AI?
When the work matters, there is a line between delegating a task and giving away the understanding you need to maintain, extend, and trust the system.
Where do you draw that line?
https://t.co/6YGVGTFRR2
Why does encapsulation matter again in the AI era?
Because as systems get larger, we cannot trace everything.
We need responsibility.
We need boundaries.
We need contracts.
New Thinking with AI episode:
https://t.co/HJOd4Y2fQn
If AI can write the code, what’s left for developers?
I think the value moves from implementation to understanding:
context
decomposition
validation
systems thinking
changeable design
New Thinking with AI episode:
https://t.co/CNPowY1Fzb
Most people think better prompting is about getting AI to understand you.
I think the more interesting part is discovering what you didn’t understand yet yourself.
New Thinking with AI episode:
AI is a mirror for your understanding.
https://t.co/8ZvqUAAacR
Everyone asks:
"If AI writes the code, what happens to software developers?"
I think they're asking the wrong question.
The most valuable skills were never programming languages
They're thinking skills
New video
https://t.co/HfUQNmTReV
#AI#SoftwareEngineering#Programming