Tom Redman retweeted
How do we deliver greatness in the age of AI? This is how:
1. Introduce a Product Editor role (h/t @lil_dill)
This is a person whose job it is to ruthlessly reject mediocrity. To avoid being overwhelmed, they must enforce their standards across the product organization.
Leverage the fact that people want their code to make it to production by ruthlessly rejecting subpar work. Features often look great when viewed in isolation, but fall apart under scrutiny when viewed holistically. The Editor holds the holistic view and constantly communicates it.
This person must have great taste, tier 1 communication skills, and extremely high agency. They are the closest thing to a real customer in your organization.
How do we deliver greatness in the age of AI? This is how:
1. Introduce a Product Editor role (h/t @lil_dill)
This is a person whose job it is to ruthlessly reject mediocrity. To avoid being overwhelmed, they must enforce their standards across the product organization.
Leverage the fact that people want their code to make it to production by ruthlessly rejecting subpar work. Features often look great when viewed in isolation, but fall apart under scrutiny when viewed holistically. The Editor holds the holistic view and constantly communicates it.
This person must have great taste, tier 1 communication skills, and extremely high agency. They are the closest thing to a real customer in your organization.
5. Done is not good and good is not great
Our tools changed, but why did our standards? It’s extremely tempting to consider code-done as job-done.
Code-done is not job-done and it certainly is not job-done-great. They are different things.
To deliver a great product, you need to go two levels above what your agent does by default.
Your agent will get you to done but they will not get you to good, let alone great.
You must use your human thinking brain to elevate done to good and even then, you’re not done. Do one more pass with your meat folds to elevate it further, to something you’re really proud of.
I’ve found that I can do done-to-good in the same day, but good-to-great requires that I sleep on it.
6. Use AI to create great products
Wait, what? Isn’t this whole thing about how AI doesn’t make great? No. AI doesn’t make great by default.
Back in the olden times, it was common to spend a lot of time and energy working on a feature that you knew was 90%, but acheiving the remaining 10% would take a disproportionate amount of more time and energy.
So, we leave it there and call it good enough. Which, when software was expensive, was the right thing to do! You couldn’t spend your time on diminishing returns in the pursuit of greatness. Maybe you could take some learnings away for your next feature, but this one is ready for prod.
But now, software is cheap! AI is more than happy to burn to kens on diminishing returns, and if the result is greatness, it’s tokens well-spent.
The price of greatness has dropped.
Use your agents to leverage your creativity – try new things, experiment with different interactions, and crush the smallest of details.
let your team's agents talk to and learn from each other:
soopdoop.com/
Currently works with @superset_sh
It's funny to think about these new experiences while building, like opening a PR link that "I" authored, and thinking "I hope I already merged this", then seeing that sweet, sweet purple chip.
"Nice, no PRs open"
if you’ve ever built a house, you understand the leverage that large machinery has over manual labor.
llms are the heavy machinery of software development. professionally, its negligent not to use heavy machinery.
What's the @resend of phone calls? @zenorocha anybody you'd recommend in the space?