Getting shit done
The product makes the call on should we do this at all, Engineering and UX essentially has the veto power on how we do it, and then at the end of the day, the PM has to put their body on the line for is this ready to ship. If they say yes and the thing’s garbage, then the PM has to wear it. That’s generally how that works.
Shopify has an internal system called GSD, which is a bit like Jira or a task management app, except it’s not really a task management app. It’s a project stakeholder reviewing tool. It stands for “Get Shit Done.”
We have a system called OK1 and OK2, and for any particular team there’s a front line of reviewers. On OK1s, it’s usually the directors from product, UX, engineering, and sometimes data who all have to sign off. And then it comes up to the senior leadership team (SLT), which we call OK2, which includes, again, product, UX, engineering, and data, who all have to sign off on it as well.
Doing the right thing for merchants
Tobi’s also really good at pushing out the values of being merchant-obsessed. This is also one of the things that I learned when I got to Shopify that was one of those nice luxuries. Shopify has been and still is such a successful company that Tobi was able to both establish and live by a set of very merchant-focused values for a very long time.
There’s such an insane culture here of doing the right thing for merchants, even at the expense of revenue or other users like buyers or developers. You can have that culture if you have a successful business where you can live that and the business still succeeds. That’s a cultural thing that runs through everything.
Hiring people who understand both what good and bad look like
I have a really simple interview process. Not to give too much of the game away, but I only ask four questions. The first question is “Tell me about a product that you were responsible for that went well, and tell me the story of that thing.” The second question is “Tell me about a product that you’re responsible for that didn’t go well, and tell me the story of that thing.” And then I do the same two questions for a product manager who worked for you.
I’m looking to understand how a person thinks about what good or bad looks like. How did they react when things went wrong? How insightful is their analysis of what they would do differently, especially in the product that went well or the PM who went well; how much reflection is there on what could have been done better? Even in the things that went well, is this person tortured by that thing that they didn’t do well even in that product that was ridiculously successful?
Marc Andreessen’s AI notes/thoughts.
- Think of AI as a sophisticated autocomplete that probabilistically uses all the text ever written to accurately predict the next word.
- AI doesn’t ‘understand’ or ‘think’ about data in the same way humans do, and doesn’t distinguish between sarcasm and actual information. Instead, it follows a narrative that is jointly constructed with the user, which can be used to explore various fiction and non-fiction scenarios.
- AI reduces the cost of knowledge work by orders of magnitude. As a result, it will free up spending power, increase quality of life, and create new jobs
- The next stepping stone would be AI correlating information in new and interesting ways. For instance, to generate new chemical compounds, mathematical formulas, product ideas, fictional scenarios, etc.
- Two theories explain bias in AI systems: The first theory suggests that AI systems reflect biases in the training data. The second theory involves direct censorship and restraints on the model.
- Some companies are already trying to get regulatory capture around AI, erect barriers to new startups, and ban open-source AI by arguing that it’s too dangerous.
- VC-backed companies are having a hard time. The more money a company raised, the more pain they’re feeling.
- Incumbents and market leaders are quickly become adept at deploying cutting-edge AI using internal teams and open-source, off-the-shelf technology, cutting out what seemed to be good opportunities for VC-backed startups.
- Indie devs are building small, cash-flowing businesses by quickly shipping niche AI-powered products in niche markets.
- A small number of promising Moonshot companies with unproven technology hold the most potential for VC-sized returns.
A few questions that change your perspective
- What hard but necessary conversation am I putting off?
- If you fired yourself today and took over your life from scratch, what would you stop doing?
- If you had to eliminate all but the most essential activity, which one would remain?
- What would your 80-year-old self say about your decisions today?
- What happens if I go off the grid for a week unannounced?
- What would this look like if it were easy?
- Are the people I love getting my best energy
Others
- The Shopify story
- IMAX emulates Palm Pilot (the OG smartphone from the ’90s) software to power Oppenheimer’s 70 mm release
- Use Visual Look-Up on iPhone to identify plants, dog breeds, and more from images.
Substack edition
How Shopify builds products
Getting shit done
The product makes the call on should we do this at all, Engineering and UX essentially has the veto power on how we do it, and then at the end of the day, the PM has to put their body on the line for is this ready to ship. If they say yes and the thing’s garbage, then the PM has to wear it. That’s generally how that works.
Shopify has an internal system called GSD, which is a bit like Jira or a task management app, except it’s not really a task management app. It’s a project stakeholder reviewing tool. It stands for “Get Shit Done.”
We have a system called OK1 and OK2, and for any particular team there’s a front line of reviewers. On OK1s, it’s usually the directors from product, UX, engineering, and sometimes data who all have to sign off. And then it comes up to the senior leadership team (SLT), which we call OK2, which includes, again, product, UX, engineering, and data, who all have to sign off on it as well.
Doing the right thing for merchants
Tobi’s also really good at pushing out the values of being merchant-obsessed. This is also one of the things that I learned when I got to Shopify that was one of those nice luxuries. Shopify has been and still is such a successful company that Tobi was able to both establish and live by a set of very merchant-focused values for a very long time.
There’s such an insane culture here of doing the right thing for merchants, even at the expense of revenue or other users like buyers or developers. You can have that culture if you have a successful business where you can live that and the business still succeeds. That’s a cultural thing that runs through everything.
Hiring people who understand both what good and bad look like
I have a really simple interview process. Not to give too much of the game away, but I only ask four questions. The first question is “Tell me about a product that you were responsible for that went well, and tell me the story of that thing.” The second question is “Tell me about a product that you’re responsible for that didn’t go well, and tell me the story of that thing.” And then I do the same two questions for a product manager who worked for you.
I’m looking to understand how a person thinks about what good or bad looks like. How did they react when things went wrong? How insightful is their analysis of what they would do differently, especially in the product that went well or the PM who went well; how much reflection is there on what could have been done better? Even in the things that went well, is this person tortured by that thing that they didn’t do well even in that product that was ridiculously successful?
Marc Andreessen’s AI notes/thoughts.
Think of AI as a sophisticated autocomplete that probabilistically uses all the text ever written to accurately predict the next word.
AI doesn’t ‘understand’ or ‘think’ about data in the same way humans do, and doesn’t distinguish between sarcasm and actual information. Instead, it follows a narrative that is jointly constructed with the user, which can be used to explore various fiction and non-fiction scenarios.
AI reduces the cost of knowledge work by orders of magnitude. As a result, it will free up spending power, increase quality of life, and create new jobs
The next stepping stone would be AI correlating information in new and interesting ways. For instance, to generate new chemical compounds, mathematical formulas, product ideas, fictional scenarios, etc.
Two theories explain bias in AI systems: The first theory suggests that AI systems reflect biases in the training data. The second theory involves direct censorship and restraints on the model.
Some companies are already trying to get regulatory capture around AI, erect barriers to new startups, and ban open-source AI by arguing that it’s too dangerous.
Who is winning in AI now?
VC-backed companies are having a hard time. The more money a company raised, the more pain they're feeling.
Incumbents and market leaders are quickly become adept at deploying cutting-edge AI using internal teams and open-source, off-the-shelf technology, cutting out what seemed to be good opportunities for VC-backed startups.
Indie devs are building small, cash-flowing businesses by quickly shipping niche AI-powered products in niche markets.
A small number of promising Moonshot companies with unproven technology hold the most potential for VC-sized returns.
A few questions that change your perspective
What hard but necessary conversation am I putting off?
If you fired yourself today and took over your life from scratch, what would you stop doing?
If you had to eliminate all but the most essential activity, which one would remain?
What would your 80-year-old self say about your decisions today?
What happens if I go off the grid for a week unannounced?
What would this look like if it were easy?
Are the people I love getting my best energy