The most powerful thing a product can do is give its user a premise.1 A premise is the foundational belief that shapes a user’s behavior. A premise can normalize actions that people otherwise might not take, held back by some existing norm

AirBnb. The premise: It’s ok to stay in strangers’ homes.

the idea of staying in strangers’ homes for short stays was doubted even by the founders. Crashing in someone’s spare room wasn’t unheard of, but it might be seen as weird, taboo, or even dangerous.

Bumble. The premise: It’s ok for women to ask men out.

The best way to follow through on a premise is to make it the core feature of the app. Bumble did, requiring that women make the first move on the app. A woman would be presented with a list of her matches and would have to make the first "move" before men could reply. This of course became a powerful differentiating feature and marketing hook.

Substack. The premise: It’s ok to charge for your writing.

Substack's premise aimed to normalize the hardest part of internet writing: getting paid. They aimed to show that independent authors could succeed at making a living (and subscription models aligned with this ethos). In doing so, Substack also made the less-hard parts of internet writing even easier. You could start a newsletter and keep it free until you felt confident about going paid. This not only normalized the end goal but also lowered the barrier to getting started.

A premise is valuable not only for “products,” but also for experiences.As I recently shouted, people still underestimate the power of giving a social event a premise. Hackathons, housewarmings, happy hours and the like are hangouts with a narrative. They have a good premise — a specific context that makes it more comfortable to do something that can be hard: socialize. (Side note: some of the best tv series and films are built on great premises.)

Premises work best on end consumers, prosumers, small business freelancers, and the like. Many two-sided marketplaces serving two of these stakeholder groups tend to have a good premise. For example, Kickstarter's premise for the creator might be: It’s ok to ask for money before you've built a product.

what I cheered as an analyst was Altman’s clear articulation of the company’s priorities: lower price first, speed later. You can certainly debate whether that is the right set of priorities (I think it is, because the biggest need now is for increased experimentation, not optimization), but what I appreciated was the clarity.

The fact that Microsoft is benefiting from OpenAI is obvious; what this makes clear is that OpenAI uniquely benefits from Microsoft as well, in a way they would not from another cloud provider: because Microsoft is also a product company investing in the infrastructure to run OpenAI’s models for said products, it can afford to optimize and invest ahead of usage in a way that OpenAI alone, even with the support of another cloud provider, could not. In this case that is paying off in developers needing to pay less, or, ideally, have more latitude to discover use cases that result in them paying far more because usage is exploding.

You can, in effect, program a GPT, with language, just by talking to it. It’s easy to customize the behavior so that it fits what you want. This makes building them very accessible, and it gives agency to everyone.

Stephen Wolfram explained: For decades there’s been a dichotomy in thinking about AI between “statistical approaches” of the kind ChatGPT uses, and “symbolic approaches” that are in effect the starting point for Wolfram|Alpha. But now—thanks to the success of ChatGPT—as well as all the work we’ve done in making Wolfram|Alpha understand natural language—there’s finally the opportunity to combine these to make something much stronger than either could ever achieve on their own.

This new model somewhat alleviates the problem: now, instead of having to select the correct plug-in (and thus restart your chat), you simply go directly to the GPT in question. In other words, if I want to create a poster, I don’t enable the Canva plugin in ChatGPT, I go to Canva GPT in the sidebar. Notice that this doesn’t actually solve the problem of needing to have selected the right tool; what it does do is make the choice more apparent to the user at a more appropriate stage in the process, and that’s no small thing.

ChatGPT will seamlessly switch between text generation, image generation, and web browsing, without the user needing to change context. What is necessary for the plug-in/GPT idea to ultimately take root is for the same capabilities to be extended broadly: if my conversation involved math, ChatGPT should know to use Wolfram|Alpha on its own, without me adding the plug-in or going to a specialized GPT.

the obvious technical challenges of properly exposing capabilities and training the model to know when to invoke those capabilities are a textbook example of Professor Clayton Christensen’s theory of integration and modularity, wherein integration works better when a product isn’t good enough; it is only when a product exceeds expectation that there is room for standardization and modularity.

To summarize the argument, consumers care about things in ways that are inconsistent with whatever price you might attach to their utility, they prioritize ease-of-use, and they care about the quality of the user experience and are thus especially bothered by the seams inherent in a modular solution. This means that integrated solutions win because nothing is ever “good enough”

the fact of the matter is that a lot of people use ChatGPT for information despite the fact it has a well-documented flaw when it comes to the truth; that flaw is acceptable, because to the customer ease-of-use is worth the loss of accuracy. Or look at plug-ins: the concept as originally implemented has already been abandoned, because the complexity in the user interface was more detrimental than whatever utility might have been possible. It seems likely this pattern will continue: of course customers will say that they want accuracy and 3rd-party tools; their actions will continue to demonstrate that convenience and ease-of-use matter most.

Prompt engineers not only need to get the model to respond to a given question but also structure the output in a parsable way (such as JSON), in case it needs to be rendered in some UI components or be chained into the input of a future LLM query. They scaffold the raw input that is fed into an LLM so the end user doesn’t need to spend time thinking about prompting at all.

From the user’s side, it’s hard to decide what to ask while providing the right amount of context.From the developer’s side, two problems arise. It’s hard to monitor natural language queries and understand how users are interacting with your product. It’s also hard to guarantee that an LLM can successfully complete an arbitrary query. This is especially true for agentic workflows, which are incredibly brittle in practice.

When we speak to other people, there is a shared context that we communicate under. We’re not just exchanging words, but a larger information stream that also includes intonation while speaking, hand gestures, memories of each other, and more. LLMs unfortunately cannot understand most of this context and therefore, can only do as much as is described by the prompt

most people use LLMs for ~4 basic natural language tasks, rarely taking advantage of the conversational back-and-forth built into chat systems:Summarization: Summarizing a large amount of information or text into a concise yet comprehensive summary. This is useful for quickly digesting information from long articles, documents or conversations. An AI system needs to understand the key ideas, concepts and themes to produce a good summary.ELI5 (Explain Like I'm 5): Explaining a complex concept in a simple, easy-to-understand manner without any jargon. The goal is to make an explanation clear and simple enough for a broad, non-expert audience.Perspectives: Providing multiple perspectives or opinions on a topic. This could include personal perspectives from various stakeholders, experts with different viewpoints, or just a range of ways a topic can be interpreted based on different experiences and backgrounds. In other words, “what would ___ do?”Contextual Responses: Responding to a user or situation in an appropriate, contextualized manner (via email, message, etc.). Contextual responses should feel organic and on-topic, as if provided by another person participating in the same conversation.

Prompting nearly always gets in the way because it requires the user to think. End users ultimately do not wish to confront an empty text box in accomplishing their goals. Buttons and other interactive design elements make life easier.The interface makes all the difference in crafting an AI system that augments and amplifies human capabilities rather than adding additional cognitive load.Similar to standup comedy, delightful LLM-powered experiences require a subversion of expectation.

Users will expect the usual drudge of drafting an email or searching for a nearby restaurant, but instead will be surprised by the amount of work that has already been done for them from the moment that their intent is made clear. For example, it would a great experience to discover pre-written email drafts or carefully crafted restaurant and meal recommendations that match your personal taste.If you still need to use a text input box, at a minimum, also provide some buttons to auto-fill the prompt box. The buttons can pass LLM-generated questions to the prompt box.

Duolingo had already implemented several gamification mechanics successfully, such as the progression system on the home screen, streaks, and an achievements system. And second, top digital games at the time had much higher retention rates than our product, which I took as evidence that we hadn’t yet reached the ceiling for gamification’s impact.

The moves counter allowed users only a finite number of moves to complete a level, which added a sense of scarcity and urgency to the gameplay. We decided to incorporate the counter mechanic into our product. We gave our users a finite number of chances to answer questions correctly before they had to start the lesson over.

Depressingly, the result of all that effort was completely neutral. No change to our retention. No increase in DAU. We hardly got any user feedback at all.

When you are playing Gardenscapes, each move feels like a strategic decision, because you have to outmaneuver dynamic obstacles to find a path to victory. But strategic decision-making isn’t required to complete a Duolingo lesson—you mostly either know the answer to a question or you don’t. Because there wasn’t any strategy to it, the Duolingo moves counter was simply a boring, tacked-on nuisance. It was the wrong gamification mechanic to adopt into Duolingo. I realized that I had been so focused on the similarities between Gardenscapes and Duolingo that I had failed to account for the importance of the underlying differences.

Referrals work for Uber because riders are paying for rides on a never-ending pay-as-you-go system. A free ride is a constant incentive. For Duolingo, we were trying to incentivize users by offering a free month of Super Duolingo. However, our best and most active users already had Super Duolingo, and we couldn’t give them a free month when they were already in a plan. This meant that our strategy, which needed to rely on our best users, actually excluded them.

Now when looking to adopt a feature, I ask myself:Why is this feature working in that product?Why might this feature succeed or fail in our context, i.e. will it translate well?What adaptations are necessary to make this feature succeed in our context?

Our failure with the Gardenscapes-style moves counter hadn’t actually disproved any of the original reasons why we believed gamification still had upside for Duolingo—we had only learned that the moves counter was a clumsy attempt at it. This time, we would be more methodical and intelligent about features we added or borrowed.

We deliberately made our leaderboard as casual and frictionless as possible; users were automatically opted in and could progress to the top of the first league by merely engaging consistently in their regular language study. By keeping the game mechanic exciting, but making it simpler than in FarmVille 2, we felt like we had struck the right balance of adopting and adapting.