As a tech lead or eng manager, you so frequently get request from above or from other teams to drop what you are doing and work on this thing they need, *now*.
During my 4 years at Uber after asking these questions, 9 out of 10 times it turned out it wasn't really urgent:
1. "What is the impact of this work you're asking for?" If the impact is unclear: sorry, but we can't do the work. Why would we?
Just this question made the requester realize half the time they just think it's urgent, but don't know what the work will actually result in.
2. "Do you have a spec that is agreed with stakeholders?" A writeup answering the "why" and the "what" that is signed off by relevant business folks.
I've seen so much engineering work thrown out as later the business goes "that's not what we wanted, why didn't you tell us?"
3. "We're not committing to any work before we have done a rough estimation."
With #1 and #2 done, many stakeholders will come and say "drop what you're doing, this is a 3-day work we need ASAP."
Hold your horses. You don't make estimates: the team doing the work does...
4. Make the cost of dropping what you're doing very clear.
This cost is always forgotten by the person coming with the request. But it's a relevant one: wrapping up work, onboarding to the new work, then later onboarding to the old work. Plus a hit on morale for a sudden change!
Uber has some very hectic times when there were reasons we needed to do some new work ASAP. Like a regulation change that means the company would be banned from operating in a region if not building something.
Even in such a place, most "urgent" things turned out to be noise.
The way I always approached these requests was to educate the person coming with them, and have them realize their work is actually not as urgent or as important or as impactful of what the team is already doing.
Doing so meant building empathy both ways, and less hard feelings.
A huge upside of this approach: when committing, you *can* commit with a very high certainty that you will not be interrupted with your work.
The alternative: take on this "super urgent" work, then someone else comes along saying " I need you to drop what you are doing *now*..."
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Short past tenure is less of a blocker for up to sr eng positions IME.
Not having had a *single* ~3+ year gig can become a blocker for staff+ levels, though. Because "how can we trust you can build for the long run if you've never stayed long around to find out?"
I just realized how I’ve heard virtually nothing about the Vision Pro. I saw dozens of people on my feed pre-order it, unbox it, share first impressions.
Now it’s just… nothing. Nothing good, nothing bad.
It reminds me of the Microsoft HoloLens launch and subsequent… quiet.
When it becomes available I’ll probably get one with a 14-day return. If I don’t get value out of it… just return.
Safe to assume it’s happening at bulk. $3,500 is too expensive to be a dust holder for days
I don't often agree with Scott Galloway, but his analysis on why the Vision Pro will fail - written a year ago, well before its release! - I'm pretty sure will go down as an "I told you so."
Marissa Mayer raising $20M for her startup, and getting ~1,000 installs with her well-known persona shows how it is REALLY hard to build a “hit app” in 2024. Like 100x as hard as 10 years ago. What worked in 2012 (“just build an app!” clearly doesn’t:
Via @platformer:
Full article:
The only person I’ve seen to be able to repeatedly pull off launching a hit B2C *app-only* launch in the 2020s is @nikitabier. And he has plans, strategies, and is obsessed with viral loops & understands today’s users (+FOMO)platformer.news/marissa-mayer-…
@nikitabier I actually applaud Mayer she went from corporate exec with enough money to retire for good: instead go back to building and this an honest shot! And hope she keeps at it.
Sunshine is like 95-98% of all startups in terms of outcome. Startups are just damn hard!! 0 to 1 is hard.
What if the single best use case for GenAI / LLM is in the coding domain? Code generation, autocomplete, debugging assist etc. Thanks to coding being one of the very few places where
1. Training data is plentiful
2. Hallucinations can be limited w a feedback loop
Show me another field that
1. Uses text in a “machine-like” form like coding does (very simple “grammar”)
2. Can have rapid loops to eliminate hallucination
I can think of one coming close: customer support offering a vetted list of text resources as a response. What else?
What if other fields that lack all of these see far far less success - and us devs assume they will? But they have different constraints…
And so here we are devs, trapped in a bubble where GenAI is very promising.
And yet we don’t see the limitations outside this bubble?
Turned out that the only interns that other company could hire were ones that were either local (no housing!) or with parents who paid “the difference.”
Uber paid a new grad salary + housing. It was €3,700/month + €1,000/month housing.
That dev is still at Uber, now a sr eng.
This hire was an absolute standout one. Smart, hard-working, humble, always positive.
I sometimes think how this other tech company “pushed” this hire away from them because they refused to pay even a living wage for their interns.