Igor Faletski retweeted
Things I stole from @Shopify while I was at Stan:
1. Almost optimize for churn.
Shopify's mission is to increase entrepreneurship, so it makes starting as easy as possible knowing most businesses will fail. A few breakout merchants can make the economics of the entire cohort work.
2. Give different work different clocks.
Growth can work in short feedback loops. Core product may be building infrastructure that needs to last for decades. Holding both to the same measurement window rewards the wrong work.
3. Optimize totals, not rates.
The easiest way to improve a conversion rate is to make the previous step harder. Fewer people enter, a higher percentage convert, and the dashboard turns green, even though fewer people reach the outcome you care about.
4. Measure value over the period in which it emerges.
Shopify looks at the total GMV created by a merchant cohort over 1, 3, and 5 years. A few breakout merchants can make the economics of an entire cohort work. Short-term retention alone misses that.
5. Revisit your “winning” experiments.
An early lift may only pull demand forward. Shopify preserves holdout groups and checks them months later to see whether an experiment created durable value or merely borrowed it from the future.
6. Personalize onboarding with what you learn.
Collecting more information during signup can be worth the friction when you use it to give people better guidance. The goal isn't the shortest signup flow. It's getting more people to the outcome they came for.
7. Put onboarding inside the product.
Shopify has a “no wizard” principle. Instead of teaching people in a separate setup flow, it brings the useful parts of a wizard, like simplification and personalization, into the actual product experience.
The common thread: don't optimize the appearance of progress. Design the product, incentives, and measurement system around the outcome you actually want to create.
Igor Faletski retweeted
Replying to @kaijdavey
Great question. My rough framing is that AI has given engineering a 100x boost, product 10x, and design 5x. That imbalance creates real strain: engineers can merge lots of individual changes while the overall UX gets less coherent, and designers burn out trying to keep up.
Design is a deep craft. It takes thought and a view of the whole experience. We haven’t given designers tooling that matches how quickly we can now produce code.
That’s why we’re building Inspect for designers. The ambition is to give them the 200x boost. Then engineers can be the ones trying to keep up :)
Igor Faletski retweeted
just got off the phone with a new engineer hire at a portco with a very wise insight
“AI has been replacing my job since I got a CS degree, yet I’m busier than ever - we can all just be more ambitious”
this is the same insight that the inimitable @JensenHuang has for us all
Igor Faletski retweeted
People rest when things are settled. Nothing about this is settled. Every week the map redraws itself and you either keep up or wake up irrelevant. That's exhausting and also the most alive I've felt in this industry. Both things are true at once.
Igor Faletski retweeted
the potential of the current moment makes me feel a little physically ill
maybe it’s lack of sleep
Igor Faletski retweeted
It makes sense to optimize model routing at the harness layer instead of the gateway layer if you want to hillclimb on accuracy/cost for any e2e task.
Every task is solved by a combination of a model mixture and agent harness. Every task requires a different mixture of models (+harness logic) to be at the pareto frontier of accuracy and cost.
* If you only optimize the model mixture at the gateway layer, you lose the broader context encoded in the harness and only optimize at the LLM completion layer.
* If you optimize the model mixture at the harness layer, you can make the "optimal" model choices a priori and while in the agent loop
For any given task, the model and harness are probably co-optimized together, so that the model mixture can only exist with this specific harness shape and vice versa
Igor Faletski retweeted
Everyone who thinks AI slop will ruin code efficiency/performance is going to be so surprised when everything is absurdly well-optimized John Carmack style machine code.
Igor Faletski retweeted
Critics: I’ll believe it when I see it
Builders: I believe it so I’ll make it
Igor Faletski retweeted
Startups win because it's easier to accelerate on the right foundations. Incumbents lose when they chase shiny objects
Lightspeed's @buckymoore says the real opportunity in the AI app layer is in large industries far enough afield from where the model providers are today — and where the context engineering to get customer data into the model is extremely nuanced and messy.
"I think this is kind of the elephant in the room right now — whether post-training open-source models combined with the unique user feedback you get from being an application provider is defensible enough."
"That is going to be an inevitable challenge for any of these industries that hit a maturation point of AI adoption, like legal and software engineering have."
"But on the other hand, there are some industries where they're very large, they're far enough afield from where the model providers are today — and probably will continue to be — and the context engineering to actually get the customer data into the model is just so messy. It requires going across different business functions, it requires a lot of hands-on forward-deployed engineering."
"Those are the kind of companies that we get really excited about. Because I think being really good at that is not only defensible, but it also allows you to generate a feedback loop with your customers, where you hear a lot of their secrets. And those secrets allow you to feed that back into how you make your product better at the expense of anyone else playing in the space. Because if you're serving the customer, they're only serving you those secrets."
"I think Palantir is a good example of this in the pre-AI era, and I think we're going to see many companies ascend in that same way."


























