Alon retweeted
Sharing an interesting recent conversation on AI's impact on the economy.
AI has been compared to various historical precedents: electricity, industrial revolution, etc., I think the strongest analogy is that of AI as a new computing paradigm (Software 2.0) because both are fundamentally about the automation of digital information processing.
If you were to forecast the impact of computing on the job market in ~1980s, the most predictive feature of a task/job you'd look at is to what extent the algorithm of it is fixed, i.e. are you just mechanically transforming information according to rote, easy to specify rules (e.g. typing, bookkeeping, human calculators, etc.)? Back then, this was the class of programs that the computing capability of that era allowed us to write (by hand, manually).
With AI now, we are able to write new programs that we could never hope to write by hand before. We do it by specifying objectives (e.g. classification accuracy, reward functions), and we search the program space via gradient descent to find neural networks that work well against that objective. This is my Software 2.0 blog post from a while ago. In this new programming paradigm then, the new most predictive feature to look at is verifiability. If a task/job is verifiable, then it is optimizable directly or via reinforcement learning, and a neural net can be trained to work extremely well. It's about to what extent an AI can "practice" something. The environment has to be resettable (you can start a new attempt), efficient (a lot attempts can be made), and rewardable (there is some automated process to reward any specific attempt that was made).
The more a task/job is verifiable, the more amenable it is to automation in the new programming paradigm. If it is not verifiable, it has to fall out from neural net magic of generalization fingers crossed, or via weaker means like imitation. This is what's driving the "jagged" frontier of progress in LLMs. Tasks that are verifiable progress rapidly, including possibly beyond the ability of top experts (e.g. math, code, amount of time spent watching videos, anything that looks like puzzles with correct answers), while many others lag by comparison (creative, strategic, tasks that combine real-world knowledge, state, context and common sense).
Software 1.0 easily automates what you can specify.
Software 2.0 easily automates what you can verify.
Alon retweeted
IDEA ABSORPTION
Idea generation is one of my biggest strengths as a problem solver but the inability to watch when / where i was doing it, was a big weakness early in my career as an executive / leader. I didn’t even realize I was inundating my org with ideas, until a peer Eng leader came to me and said “Gokul you’ve got to stop generating idea confetti. The org needs focused execution, not more ideas.” I’ve never forgotten that line from nearly two decades ago. I made two changes: (a) ideating in very small, mostly 1:1,settings with senior leaders, to flesh out a concept (I do this with AI today!), vs in groups, unless the goal of the group meeting was to ideate, and (b) trying to be Socratic in phrasing: “should we consider X” and being clear that X is a suggestion, not a directive, and it’s the team’s job to decide if X makes sense vs all the other things on their plate. Learnt this the hard way after being presented work output by someone who had worked on a project for a week based on misinterpreting a throwaway comment from me at a meeting.
It was cathartic to see that Bezos had the same issue early on in his career. From a recent interview:
“Jeff Wilke came to me one day he worked for Amazon for a quarter of a century but this is when he probably knew me only for a year and he said “Jeff you have enough ideas to destroy Amazon.” And this was such a shocking idea for me
Jeff said you have enough ideas per minute per day per week to destroy Amazon. I was like what do you what do you mean? He’s like you have to release the work at the right rate that the organization can accept it and he was a manufacturing expert and so you know his view of the world was every time I released an idea I was creating a backlog a queue work in process and because it was just stacking up it was adding no value and in fact it was creating distraction and so he said ‘Look you have to figure out when to release these new ideas at a rate that the organization can accept them and this was I mean this sounds so obvious but it was not obvious at the time to me and this was a profound insight for me and so I started prioritizing the ideas better keeping lists of them keeping them to myself until the organization was ready for the ideas and then I also started figuring out how can I build an organization that can be ready for more ideas that’s about having the right senior team and the right leadership and getting those people the executive bandwidth so they could do more ideas per unit time and so and and that is what we built we built a company that’s very good at inventing and doing more than one thing at a time and you do want to build as the company gets bigger you do want to be able to do more than one thing at a time but that idea of releasing the work was very profound for me and it made it made us operationally more effective while still being inventive and do you think you’re a better inventor.”
Alon retweeted
It's true, claude is better than writing in OpenAI's o1 model.
However, o1 can do something claude cannot: process 3-4 hours of transcripts and create detailed worksheets, summaries, and FAQs that cite what the speaker has said word for word, and recall every question answer pair.
Alon retweeted
I’m hiring Product Designers for @airtable AI!
We’re creating tools to make AI actually work for work, helping teams easily connect data, generate insights, and automate workflows.
Alon retweeted
I'm hiring a seasoned product designer to join our @airtable AI team! ✨
We’re creating AI experiences that will completely transform how teams build custom apps and workflows with Airtable.
Alon retweeted
Join me tomorrow for the Ultimate Product-led Onboarding Playbook live event with Reforge!
I'll be sharing the playbook from my time working with 100s of companies while at Dopt and leading Growth Design at Dropbox.
Almost 3,000 people have registered already. Sign up link ↓
Alon retweeted
Super excited to host the Ultimate Product-led Onboarding Playbook live event with Reforge!
Building exceptional product-led onboarding is super hard. What does good onboarding even look like? Which patterns are effective, and why? How do you measure success?
The Ultimate Product-led Onboarding Playbook is designed from first principles (understand your users!) to answer those questions and help you nail onboarding for your unique users and product.
It's based on my time working with 100s of companies in the Dopt community and my time as Head of Growth Design at Dropbox.
During the event, I'll:
• Dive into a 10-step playbook for driving activation through improved product onboarding.
• Host a live Q&A to address all your burning questions about product onboarding.
1,300 people have registered already! RSVP and submit questions via the link in the comments. I'd love to hear what's interesting to you!
Alon retweeted
Leading Growth Design at @Dropbox taught me the power of new user onboarding journey maps. Now, at @doptcom, we're using the same framework. Here's how it works:
All too often journey maps are packed with research but miss important details. A quality journey map focuses on the path to value and eliminates unnecessary friction. Friction to help users discover, understand, and realize value is good. Unnecessary friction that distracts from realizing the core value is bad.
Here's where to start:
1. First, get grounded. Define your ideal customer profile and the value they’re going to get from your product.
2. Define the high-level user journey moments: Setup, Aha, and Habit.
3. The setup moment is the necessary steps your users need to take so they can experience the core value of your product
4. The aha moment is when they first experience the core value of your product
5. The habit moment is when they build a habit around the core value
Now the trickiest part: the solution user journey. There are tons of onboarding patterns in the world, which will work best?
1. Map out each step of the journey! Don't expect a user will just know how to use your product or why it's valuable. Create an onboarding experience that guides the user to value that's relevant to them.
2. Each step should move the users closer to the key moments identified above. Remove those that don't.
3. Again, get grounded. Understand the type of user you serve and their journey. Horizontal SaaS? Show them how you solve their specific use case. Developer tool? Provide great docs and resources. Collaboration app? Solve the cold start problem.
4. Activating a user is never enough. For B2B products, you must activate their collaborators, team, and the broader company. Include team activation points.
5. Break the journey down by persona if you have a multi-persona product. Show how they collaborate.
6. Include moments of friction, delight, and any open questions to align your team.
These steps should result in a journey that drives wins for the user and wins for the business: the true measure of a great user journey.
In my new user journey map artifact for @reforge , I break down how I applied this new user journey mapping framework to our onboarding at Dopt (including video commentary!). Link below!
Alon retweeted
5 months ago I broke down Figma's new AI onboarding. They're back with a new iteration! The most interesting part: 𝗔𝗜 𝗽𝗿𝗼𝗺𝗽𝘁𝘀 𝗮𝗿𝗲 𝘁𝗵𝗲 𝗻𝗲𝘄 𝘁𝗲𝗺𝗽𝗹𝗮𝘁𝗲𝘀. An updated breakdown 👇
In the last iteration, an AI input form was presented side-by-side with pre-built usecase templates.
This makes sense, they both solve the cold start problem of helping users go from from 0 → .5 (they don't get users to 1 yet).
I wondered if, in the future, AI entry points and templates would merge.
With this new iteration, Figma has essentially made templates from pre-written, use case prompts!
Templates are great at presenting use cases (push). Prompting is great at capturing intent (pull).
This experience is starting to have the best of both worlds, where the entry point into prompting is structured and easy to understand, but I can still edit the prompt to fit my intent.
I like this and I think it points to a strong vision for AI-generated getting-started experiences (rather than just throwing me into an input).
Of course, templates still play a big role for Figma and Figma has a huge community with tons of helpful resources.
But now AI has a much larger in-product entry point than templates.
I could imagine other tools taking a similar approach!
Alon retweeted
Chatbots don't work for B2B SaaS. They're removed from the user and app context, prompting is too imprecise, and they don't enable you to take action.
In our interview on the Building with AI podcast, we chatted about how we found opportunities beyond the chatbot to deliver great in-app assistance.
My favorite part is that the interview included me and Karthik, our lead engineer on the project. We were able to bring the full product, design, and engineering perspectives to the conversation — so important for building AI features!
We dig deep into:
🔹 How the product and design thinking helped us create "promptless" in-app AI assistance, moving away from traditional chatbot experiences
🔹 Leveraging in-product contexts, such as UI screenshots, to deliver highly relevant and context-aware support
🔹 Technical details about the opportunities and challenges of building multi-modals
🔹 How product, design, and engineering worked together to deliver AI Assist and how that's different than building non-AI features
Links to the full interview below!
Alon retweeted
Forget chatbots 👏
In our latest episode, @haroonc sits down with @philvb (Co-founder, CPO) and @karthikramen (Staff Engineer) at @doptcom to discuss guiding principles for AI-powered UX, the significant technical challenges in developing AI Assist, key learnings from building AI products, and much more!
Alon retweeted
Latest customer using Dopt! Here's ZenML's onboarding hub built with Dopt.
A few reasons I like this:
🔹 This pattern makes sense for ZenML: onboarding hubs are most successful in products that have longer setup experiences where users need more assistance, like dev tools.
🔹 The checklists give structure to the journey toward aha moment and the support, invite, and resource entry points help users get unstuck.
🔹 There are two separate checklists in the hub: a "Starter Setup" and "Production Setup". The truth is onboarding is more like a set of aha moments where the user realizes deeper levels of value. I like how these checklists work together to get a user all the way to production. We've considered something like this at Dopt -- it's cool to see our customers go live with it!
🔹 The hub may look like a lot, but it's all tucked into an entry point in the navigation. The user can use the rest of the app and easily return here when needed.
I wrote a full breakdown of the pattern on our blog, link in the comments!
Alon retweeted
So fun to demo @doptcom's AI Assist at the @ProductHunt AI Open Demo Night! Each presenter had 2 minutes to give a ⚡️ demo and then got live feedback from the community. Really cool to meet other builders in the AI space and chat about AI tech, pricing, and UX!
Alon retweeted
When building AI features you never have full confidence “it will work”. There’s an inherent risk:
You can't know if an AI result is good without building the feature to see the result.
But you don't want to build the feature only to discover the AI response isn't good.
So how do you determine if it's worth building?
Yesterday we launched AI Assist to the world. Over the past few months, we went from low-confidence early AI ideas to launching AI assist, our new tools to build AI-powered in-product assistance into your product.
Building AI Assist was unlike any other feature I've built in my career. It felt much riskier than non-AI features.
By navigating the challenges of building AI assist, we learned 3 product lessons for mitigating risk when building AI features:
Lesson 1️⃣ : It's too easy to dream. Get grounded with technical research.
Make time for designers and developers to do research to learn what’s possible. It may end up being a sunk cost (you’ll have to be comfortable with that risk), but research is the best input to creating AI solutions that are grounded in reality.
Lesson 2️⃣: Quantify risk to select opportunities
Unlike the traditional value vs effort prioritization of non-AI product work, AI requires a sense of the probability of success across a set of bets (kind of like a Growth team!). Quantifying the risk helps navigate where to invest time.
Lesson 3️⃣: Building is the only way to know what to build
When building non-AI products, you have high certainty a feature will work after discovery and definition. Not so with AI features. With AI features, you must iteratively build to increase the probability of success and continue or decide to cut your losses and move on.
I wrote a case study about building AI Assist and dug deeper into these lessons on our blog, link in the comments!
Alon retweeted
Today we're launching AI Assist on Product Hunt. Build AI-powered assistance seamlessly into your product ✨
This was a huge, whole team effort. From inception to launch in < 3 months. I couldn't be more excited to get it into folks' hands.
Alon retweeted
Last week, @montereyai launched their onboarding hub powered by Dopt. It was so well designed I signed up for their product and explored it myself. Here are a few standout features:
𝗦𝗲𝗮𝗺𝗹𝗲𝘀𝘀 𝗱𝗮𝘁𝗮 𝘀𝗼𝘂𝗿𝗰𝗲 𝗰𝗿𝗲𝗮𝘁𝗶𝗼𝗻
The coolest part: the checklist is powered by Dopt and because we provide flexible dev tools, the create data source checklist item is their data source creation flow. You can complete the task of creating a data source - a core workflow of the product - right in the context of the getting started experience. I love how Dopt enables this like this and I think it's the most impressive part of their onboarding 👏
𝗚𝗲𝘁𝘁𝗶𝗻𝗴 𝘀𝘁𝗮𝗿𝘁𝗲𝗱 𝗯𝘂𝗶𝗹𝘁 𝗶𝗻𝘁𝗼 𝘁𝗵𝗲𝗶𝗿 𝗻𝗮𝘃𝗶𝗴𝗮𝘁𝗶𝗼𝗻
The Getting Started page is a tab in the navigation, which I think is a good pattern for complex products with a longer time to value. This nav pattern allows users to easily return to their getting started journey as the user completes other key tasks on the way to their aha moment.
𝗗𝗼𝗰𝘀 𝗮𝗻𝗱 𝘀𝘂𝗽𝗽𝗼𝗿𝘁 𝗽𝗮𝘁𝗵𝘀
The hub provides links to documentation and a "book time with us" option, offering users multiple paths for learning and assistance throughout their journey.
You can explore the hub by signing up for Monterey's product or view an example hub we created - link below!
Alon retweeted
👀 Another sneak peek of AI assist! We're in early access with customers, including ourselves. Now we can point at anything in Dopt and get AI-generated help and relevant sources from our docs.
It does a good job of understanding some tricky core Dopt concepts (including our user-targeting expression language) and returning a helpful natural language explanation. We were impressed with this one!
It's not perfect yet. The response says the element is a button, it's actually a block. And there's some more design polish to be done. But these are areas we can improve on before we launch in a few weeks.
Bonus detail: we wired this up to ⌘E for explain 🙂. So far it's felt surprisingly natural, but we'll see if it sticks (it's completely customizable with our SDKs).
Alon retweeted
Most VC advice out there overcomplicate design partnerships. As we build our new AI Assist feature, we’re running 0 → 1 customer development for a second time. Simple works for us, here’s what we’ve learned:
1️⃣ 𝗧𝗵𝗲 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝗼𝘄𝗻𝘀 𝘁𝗵𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺
Since we’ve been around for a bit, we already know our ideal customer profile and their JTBD: building quality in-product onboarding and education.
2️⃣ 𝗪𝗲 𝗼𝘄𝗻 𝘁𝗵𝗲 𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻
We believed the was a compelling new AI solution to our customer’s problem that wouldn’t fit into existing mental models. We had to own designing a valuable solution, no customer would tell it to us.
We had many ideas and through iterations decided to build tools that enable users to point at anything in your product to get AI-generated contextual help.
3️⃣ 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗲 𝗾𝘂𝗮𝗹𝗶𝗳𝗶𝗲𝗱 𝗹𝗲𝗮𝗱𝘀
We sought out existing and new customers who fit our ideal customer profile via cold and warm outbound and marketing to generate inbound.
Basically, we want to reach our ICPs who are building onboarding and education to see if they’re willing to spend time with us to give feedback on our solution.
4️⃣ 𝗔𝘀𝗸 𝗳𝗼𝗿 𝘁𝗶𝗺𝗲
We asked for a 30 or 45 min AI solution feedback meeting with 3 goals:
• Product discovery with mockups and prototypes to get feedback
• Test positioning and messaging
• Determine if there’s interest in being a design partner. At this point, the best gauge for perceived value is their willingness to spend time with us.
5️⃣ 𝗕𝗿𝗶𝗻𝗴 𝗼𝗻 𝗱𝗲𝘀𝗶𝗴𝗻 𝗽𝗮𝗿𝘁𝗻𝗲𝗿𝘀
If there’s interest in our solution, then we ask if they’re willing to become a design partner.
They’ll get early access in exchange for spending regular time with us as we build. We get feedback from them on the product and input to help us hone pricing & packaging, positioning, and messaging.
If they’re not interested, then we determine why and iterate. Is the solution not valuable enough? Or are we not talking about it correctly? Or more usually, both 🙂. If there’s overwhelmingly no interest, then stop the project.
6️⃣ 𝗜𝘁𝗲𝗿𝗮𝘁𝗲
We loop through these steps as fast as possible. We take the feedback from the calls and iterate on each part: improve messaging to generate better leads and improve product mockups to be more valuable to bring on more design partners.
We were on our 6th Figma page with 10s of iterations per page before we started showing prototypes.
𝗧𝗵𝗮𝘁’𝘀 𝘄𝗵𝗮𝘁 𝘄𝗼𝗿𝗸𝘀 𝗳𝗼𝗿 𝘂𝘀!
Now we’re pretty far along in building AI assist and we have a healthy set of design partners in early access.
If you’re interested in what we’re building, comment below / DM and we’ll let you know when we can give you access.
The image is a slide from our AI solution feedback meetings ✨
















