The startup, which just won $7.5 million in financing, uses AI virtual users to replace human traffic: uploading event logs→ generating behavioral characters→ simulating interactions→ outputting prediction reports, with zero code, zero risk, and zero privacy touch throughout the process. In the most heavily regulated financial and medical scenarios, Blok has lined up a waitlist for 60 customers – the next scene of product development may no longer require human trial and error.
While product teams are still anxious about waiting 4-6 weeks for A/B test results, a startup called Blok is completely disrupting this traditional model with AI virtual users. They have compressed product testing from “weeks” to “hours”, from “reactive” to “predictive”, which is not only an improvement in efficiency, but also a fundamental change in product development philosophy.
The product, called Blok, announced the completion of a $7.5 million funding round to build an AI-powered product testing platform. But as I delved into the company’s technology architecture and business model, I realized that this was far from a simple tool upgrade story. Blok is solving a fundamental problem plaguing the entire tech industry: how to empower product teams to make product decisions faster and more accurately in an era of skyrocketing user interface complexity.
What struck me even more was that Blok’s founders, Tom Charman and Olivia Higgs, were not tech novices, but serial entrepreneurs with deep experience in data science, behavioral modeling, and product development. They have modeled human behavior for national security agencies, published consumer applications serving millions of users, and have cutting-edge research experience in behavioral science, healthcare, and finance. This deep cross-disciplinary background allows them to think about the ancient and modern issue of product testing from a whole new dimension.
The deep dilemma of traditional product testing
As I dived deeper into the problem Blok was trying to solve, I realized that traditional product testing methods were facing unprecedented challenges. This challenge comes not only from the technical level, but also from fundamental changes in user expectations and product complexity.
How can product managers do a good job in B-end digitalization?
All walks of life have taken advantage of the ride-hail of digital transformation and achieved the rapid development of the industry. Since B-end products are products that provide services for enterprises, how should enterprises ride the digital ride?
View details >
The first is the problem of slow feedback loops. In many organizations, the experimentation cycle is extremely slow, and teams need to wait 4-6 weeks to prove an idea wrong. The process of waiting for statistical significance results is frantic, and you have to be careful to avoid “peeking traps” (making false positive judgments based on preliminary positive signals before the experiment is completed). Even before the test goes live, there is a lot of preparation work such as initial design and experimental setup, feature implementation, traffic coordination, and team alignment.
The second is the issue of cultural barriers. A big reason why product experimentation can’t be scaled in-house is a cultural issue, not a technical one. The real challenge is to develop experimental thinking within and across teams. Terms like p-value, confidence intervals, and sample size calculations often introduce unnecessary cognitive burden on teams that only want to make better product decisions and are reluctant to spend time performing statistical rigor. The result is that the team abandons experimentation altogether and goes back to the old path of relying on intuition or the opinions of the highest-paid or most charismatic people in the boardroom. Or one or a few data experts become bottlenecks—usually a data scientist or product manager with an analytics background—who need to handle every request, run every analysis, and take on the burden of validating every product change.
The third is the risk of real user testing. 80% of A/B tests fail, at best no impact, but at worst they annoy customers and lead to costly errors. This is especially true in regulated or trust-sensitive sectors such as consumer health and finance. Testing with real users can pose risks of miscommunication, trust erosion, and even regulatory violations.
Finally, there is the problem of team resource conflicts. Another common bottleneck in experimentation is flow distribution. Different teams compete for the same user base to run their A/B tests. The results are prioritized for the most “mission-critical” tests, and many valuable ideas can never be verified. In some cases, it is easy to overlook “spaghetti-style” experiments, causing one test to leak into another, producing biased results. There are also frustrated engineers who have to roll back all of them once the failed experimental features are running.
I find that the root of these problems lies in the fundamental limitations of existing testing methods: they are inherently reactive, not predictive. Teams can only understand user reactions after features have been built and deployed, and this passive feedback mechanism is becoming increasingly unsuitable in a rapidly changing market environment.
Blok’s Technological Revolution: A Paradigm Shift from Reaction to Prediction
At the heart of Blok’s innovation lies in the shift from reactive to predictive product testing, a transformation that is not just an upgrade in technology but a fundamental change in mindset. They have built a complete AI-powered ecosystem of user behavior simulations.
At the technical architectural level, Blok’s workflow reflects a deep understanding of user behavior modeling. Customers start by uploading event log data from popular analytics platforms such as Amplitude, Mixpanel, or Segment. This historical data is crucial for modeling Blok’s behavior. Based on the ingested data, Blok’s AI performs complex behavioral modeling, creating diverse user personas. These roles are designed to represent the vast majority of the app’s user base, capturing different usage patterns and preferences.
The development team then submits their Figma design along with detailed experimental parameters, including defining the hypotheses they want to test and the specific user goals they want to achieve with the new feature. Blok’s user role agent starts to act, running the simulation multiple times. These AI agents interact with the proposed designs like real users, exploring different paths and encountering potential challenges.
At the end of the simulation, Blok provides comprehensive insights. This includes detailing the overall report of experimental findings, highlighting areas that performed well, and identifying areas for improvement. Teams receive role-specific reports and customized recommendations. Paying homage to the current technological landscape, Blok also includes a chatbot interface that allows users to query experimental results and gain deeper understanding through natural language interactions.
This structured approach ensures that teams can make informed decisions based on data-driven predictions rather than guesswork, fundamentally enhancing the efficiency and effectiveness of application testing. What’s more, Blok moves the experimentation phase forward to the exploratory phase of product development, making it possible to validate design prototypes, functional hypotheses, and lightweight prototypes without writing a single line of code.
In terms of technical depth, Blok’s behavioral modeling is not just simple user segmentation but complex modeling that incorporates psychological statistics. They train agents using various foundation models and then actually start predicting how different customers will behave and respond to different growth experiments. The ultimate goal of this approach is to achieve true personalization: My version of Uber and your version of Uber may be two fundamentally different products, based not only on the behavioral profile we have, but also on how we use these different products.
The investment logic behind the $7.5 million financing
Blok’s funding story itself is a brilliant example of how investors can identify paradigm-shifting opportunities. The $7.5 million financing was completed in two rounds, and its investor lineup and investment logic are worthy of in-depth analysis.
The $5 million seed round was led by MaC Venture Capital and included employees from Discord, Google, Meta, Apple, Snapchat, and Pinterest. The Pre-seed round was led by Protagonist, with participation from Rackhouse, Ryan Hoover’s Weekend Fund, and Blank Ventures. This investor mix includes both professional VCs and industry experts from major tech companies, reflecting the widespread recognition of Blok’s technological path.
The investment case for Marlon Nichols, managing partner of MaC Venture Capital, is particularly noteworthy. He notes that Blok is often compared to Optimizely and Amplitude, but these tools are more reactive. Blok is surpassing them by providing a predictive testing layer. “We support Blok because we believe product development is at an inflection point. Teams are releasing faster than ever, but they still make critical decisions based on A/B testing and intuition. Blok’s simulation engine disrupts this model by allowing teams to predict user behavior before writing a single line of code. ”
This investment logic reflects a significant trend: investors are beginning to recognize that traditional product development methodologies require a fundamental update in the age of AI. Tools that provide predictive insights, reduce trial and error costs, and improve decision-making efficiency will become the new infrastructure.
Notably, Blok chose to attract such high-quality investor participation at a relatively early stage. This not only provides sufficient financial support for the company, but more importantly, establishes a strong industry network. Angel investors from major tech companies not only provide funding but also provide real user feedback and market validation opportunities.
Blok’s current business model is a SaaS subscription while also balancing compute costs. The company aims to achieve median multi-million dollar revenue this year and gradually open up the platform to a wider customer base. This growth trajectory reflects the growing demand for advanced predictive application testing solutions in a world where user expectations for seamless digital experiences are higher than ever.
Target market and early customer validation
Blok’s marketing strategy demonstrates a thoughtful positioning choice. Currently, Blok operates on a back-of-waiting list, working closely with initial customer bases, primarily in the financial and healthcare sectors. These industries are ideal testing grounds for Blok technology, as they operate under strict regulations and have zero tolerance for poor user experiences or flawed experimentation.
Choosing finance and healthcare as your initial market is a strategically smart decision. In these areas, the ability to thoroughly test and validate features before releasing them publicly is not just an advantage – it’s a necessity. Blok’s predictive AI simulations provide a critical safety net, ensuring that sensitive applications are robust and user-friendly from the start.
According to founder Tom Charman, large and small companies face different problems. Small companies don’t have enough user base to test their products and get real-time feedback, while large companies want to avoid cramming features into their apps, making them clunky. Blok tries to get to a point where companies don’t need to release features on an experimental basis and wait weeks or months to see results.
In a real-world use case, one particularly noteworthy example is Blok’s collaboration with a company to help them predict whether customers are likely to switch from free to paid. Through this process, they identified a significant problem: customers with the highest lifetime value (LTV) didn’t actually complete the onboarding process. This is a really big problem because product teams are often (at least in the founders’ opinion) more focused on local optimals than global optimals, short-term rather than long-term. By understanding the different types of people who come to the platform and then ensuring it’s built for the best customers, you can avoid growth issues six months later.
Currently, Blok has built a waitlist of about 60 customers, who have joined in the past few weeks. The waitlist is growing at a rate of 30-40% per month, which bodes well for them considering that no one knows who they are, which is entirely through word of mouth. The team plans to start accepting the waitlist by the end of this year and start growing the company.
The cross-disciplinary background of the founding team
The background of Blok’s founding team constitutes a significant competitive advantage for this company. Tom Charman and Olivia Higgs are serial entrepreneurs who have co-founded several companies in areas such as travel and learning. What’s more, their professional background provides a unique perspective on solving complex user behavior modeling problems.
Tom Charman’s background is particularly noteworthy. A double major in economics and political science, he started his business while in college, growing his first full-fledged startup all the way to the good shape of his graduation. His turn to data science stemmed from a deep interest in behavioral economics and game theory, particularly concepts like the prisoner’s dilemma, which led him to think about the nature of human behavior. This interdisciplinary background allows him to understand user behavior patterns from a broader perspective.
In the field of data science, Tom has accumulated 10-15 years of experience building companies, with both success stories and lessons from failures. In addition to his entrepreneurial experience, he has worked with various organizations and governments, even worked for the United Nations, and has given TEDx talks on artificial intelligence and quantum computing. This diverse experience has given him a deeper understanding of the social impact of technology applications.
Interestingly, in his spare time, Tom enjoys exploring abandoned buildings, doing so-called “urban explorations”, taking pictures of abandoned places where humans no longer exist. This focus on human behavior traces also reflects his deep motivation for user behavior modeling to some extent.
Olivia Higgs mentioned in the interview that the team interviewed more than 100 product engineers to understand the problems faced by the product team. This in-depth user research sets the stage for Blok’s unique approach to AI innovation. She emphasized that the demand for advanced testing solutions is growing as the complexity of modern interfaces increases. As users interact with technology through diverse channels like chat and voice, introducing new visual UI elements requires meticulous attention to avoid friction.
The team’s security background also provides Blok with additional technical depth. In Tom’s view, there is also a huge opportunity for user behavior modeling in security, but they chose to focus on the product area because they wanted to adopt an approach that never touched personal data. This philosophy of privacy protection stems from their European background and deep understanding of data protection regulations such as GDPR.
In-depth application of AI Agent in product testing
Blok’s use of AI Agents goes far beyond simple user simulations, and they have built a complete ecosystem of intelligent product testing. The depth and breadth of this application have shown me the immense potential of AI in product development.
At the user behavior modeling level, Blok uses AI agents based on behavioral science and product data to simulate how different user types explore products, spot friction points, and react to changes – all before the experiment goes live. Think of it as a sandbox environment to test product decisions on a virtual user base.
The core of this approach is to make behavior archiving very fine. They don’t just simply group users, but really understand the different types of people within the platform and then add psychographic statistics on top of that. When you can define these different behavioral profiles, you can train agents using various foundation models and then actually start predicting how different customers will respond to different growth experiments.
In terms of technical implementation, Blok faces an interesting challenge: how to conduct in-depth user behavior analysis without touching personal data. Their solution is to build synthetic datasets, which means no one’s data is exposed. The benefit of this approach is that product teams can start getting to know their customers without making them feel spied on.
To address data privacy concerns, the team is exploring a new frontier of cryptography called zero-knowledge. This technology, which originated from cryptocurrency and blockchain processes, is now being applied to data privacy protection. They are also considering using data enclaves so they don’t have to worry about personally identifiable information.
On a more technical level, the team is working on interesting hardware from Intel and other companies that could be valuable in achieving their privacy goals. This exploration of emerging technologies reflects the team’s innovative thinking in solving complex technical problems.
From a data demand perspective, Blok has established minimum data thresholds for different types of companies. For B2C companies, they need data from tens of thousands of customers, not hundreds of thousands or millions, but tens of thousands to be in a good position. For B2B companies, because they collect more data, they need data from thousands of customers. This relatively low data barrier allows more small and medium-sized companies to benefit from Blok’s technology.
The value of Blok isn’t just limited to product teams, it’s designed to be used across functions, covering different teams across the organization. This comprehensive range of use cases exemplifies the true value of AI-driven test platforms.
In marketing teams’ apps, Blok can perform conversion rate optimization (CRO) by running pre-launch test simulations on landing pages, sign-up flows, or ad listings. It also validates information and tests different content variations or calls-to-action with virtual users. This capability allows marketing teams to optimize key conversion nodes before they go out at scale.
For product teams, Blok’s application is more direct and central. When it comes to onboarding optimization, teams can evaluate which processes increase user activation before implementation. In terms of feature adoption, it is possible to predict which feature variants users are more likely to interact with. This predictive power allows product teams to launch new features with greater confidence.
Design teams also get great value from Blok. They can get early input on design concepts, saving time costs on usability studies or real user recruitment. This prototyping feedback capability is particularly valuable for design iterations, as it enables the identification of potential issues before the design becomes code.
I find the value of Blok’s cross-team application to be that it solves a fundamental problem in organizations: how to get different teams to make decisions based on the same user understanding. Traditionally, product, design, and marketing teams have often made decisions based on disparate assumptions and data sources, and this fragmented decision-making often leads to inconsistencies in the user experience.
Blok provides a unified user behavior simulation platform that allows all teams to test and validate their ideas based on the same user behavior model. This unity not only improves the quality of decision-making, but also reduces communication costs and coordination complexity between teams.
What’s more, Blok allows each team to test independently without competing for limited real user traffic. This ability to “test infinitely” fundamentally changes the culture of experimentation within organizations, allowing more ideas to be validated and greatly lowering the barrier to innovation.
Competition with traditional testing tools
In understanding Blok’s competitive landscape, I’ve found that they are facing competition not just from traditional A/B testing tools but from the entire product decision-making methodology. The depth and complexity of this competition extend far beyond surface-level functional contrasts.
Tools like Optimizely and Amplitude do do some great work when it comes to traditional competitors, but they’re more reactive. When Marlon Nichols says that Blok is outgoing them by providing a predictive testing layer, he points to a key difference: the difference in the temporal dimension. Traditional tools tell you what happened in the past, and Blok tells you what might happen in the future.
In the context of content, some companies are doing similar work, predicting how different customers will respond to different types of copy based on user profiles. But Blok chose not to enter the advertising space because, in the opinion of the founders, there are some privacy issues involved in the advertising space, which is not in line with their values.
Tom Charman mentioned in the interview that they saw some very early-stage companies similar to them, but it felt like they had only recently reached the point where technology was just keeping up with a really big problem. This is a really big problem that they have wanted to solve for a long time, but only just got to the point where it can actually be solved now. That’s why he thinks this could become a pretty hot area over the next 12 to 18 months.
From a technical moat perspective, Blok’s strength lies not only in the technology itself but also in their deep understanding of user behavior modeling. This understanding comes from the founding team’s cross-border experience in multiple fields: from behavioral modeling of national security to the large-scale application of consumer products, from academic research to business practice.
I think Blok’s biggest competitive advantage is their problem-solving methodology. Instead of making incremental improvements within the existing A/B testing framework, they redefined the problem itself: from “how to do better A/B testing” to “how to make better product decisions without the need for A/B testing.”
The value of this paradigm shift lies in the fact that it addresses not only the efficiency of existing methods but also risk, cost, and culture. When teams no longer need to do trial and error with real users, they can innovate more boldly, iterate faster, and make decisions with greater precision.
Business model and growth strategy analysis
Blok’s business model reflects a deep understanding of the SaaS space and a clear understanding of its value proposition. As a SaaS product, they use a subscription-based model but also balance computational cost considerations, reflecting the unique nature of AI-driven products.
In terms of integration difficulty, Blok chose the most simplified path. According to Tom Charman, the integration process is “super simple” and only requires a few lines of code and you can get to work. The entire process takes only a few days, and once set up, results start to be seen and real value delivered within a few days. This low-friction integration strategy is a key factor in the success of SaaS products.
In terms of customer acquisition strategy, Blok currently mainly relies on word-of-mouth communication. This organic growth approach is particularly valuable considering they are still in their relatively early stages. The 30-40% monthly growth rate of the waitlist, achieved purely through word-of-mouth without any marketing promotion, shows that the product really solves real pain points.
The company’s revenue target of reaching a median of several million dollars this year is neither too aggressive nor too conservative, reflecting a realistic assessment of market opportunities. Considering that their target customers are primarily B2B businesses, this revenue level means they need to acquire a decent number of corporate customers, which explains why they choose to expand their customer base carefully.
In terms of pricing strategy, while the specific price is not disclosed, it can be inferred that Blok needs to find a balance between providing value and controlling computational costs. AI-driven products often face the challenge of increasing computing costs with usage, requiring companies to refine their pricing strategies.
From the perspective of long-term business model development, Blok has huge room for expansion. Once they establish a strong position in product testing, they can apply the same technology to more areas such as security, marketing automation, customer service, and more. This platform-based development path provides the company with diversified growth opportunities.
Profound impact on the product development industry
The emergence of Blok represents a profound change that the entire product development industry is undergoing. The impact of this change will extend far beyond the technical level, reaching every aspect of organizational culture, decision-making processes, and industry standards.
First, Blok is redefining what it means to be an “experimental culture.” Traditional experimental culture requires teams to have statistical knowledge, wait patiently for results, and accept the risk of failure. But in Blok’s mode, experimentation becomes a lightweight, risk-free, instant feedback activity. This shift will allow more non-technical team members to participate in product decision-making, fundamentally changing the organization’s decision-making structure.
Second, this technology is redefining the timeline for product development. When Olivia Higgs says, “We compress weeks into hours,” she’s not just describing an increase in efficiency, she’s describing a whole new cadence of product development. Under this new rhythm, the speed of product iteration will be greatly accelerated, the market responsiveness will be significantly improved, and the competitive landscape of the entire industry may change.
Third, Blok’s technology gives small teams new weapons to compete with larger companies. Traditionally, only companies with a lot of user traffic could conduct effective A/B testing. But Blok’s virtual user simulation allows smaller companies to access the same quality of user insights, and this democratization will rebalance the industry’s competitiveness.
Fourth, this technology is driving product development from “bold assumptions and careful verification” to “accurate prediction and quick action”. When teams can accurately predict user behavior, they can be bolder in product innovation while taking on fewer risks. This change in risk-benefit ratio will encourage more product innovation.
Finally, Blok’s success could lead to a whole new division of labor in the industry. In the future, there may be new occupations such as “virtual user modelers” and “behavioral predictive analysts”, and the professional division of labor in the entire product development will be more refined.
From a broader perspective, this technological trend represented by Blok is transforming product development from an art to a science. While creativity and intuition are still important, data-driven decision-making will become more precise and reliable. This shift not only improves product quality but also reduces the probability of product failure, thereby improving the efficiency of resource use across the industry.
While Blok has made significant progress in technological innovation, the technical challenges they face also deserve an in-depth analysis. These challenges are not only relevant to Blok’s own development, but also reflect the technological frontiers of the entire AI-driven product testing landscape.
The core challenge is the accuracy of user behavior modeling. While AI can simulate a large number of user behavior patterns, real human behavior often contains many irrational and emotional factors that are difficult to capture entirely through data and algorithms. Tom Charman mentioned in the interview that they are working on how to integrate psychostatistics into behavioral modeling, which is an extremely complex interdisciplinary problem.
Blok’s user behavior simulations are based on historical data and existing user groups, which can reinforce existing user biases and overlook potential new user groups. If a product’s historical users are primarily from specific demographic groups, models trained on this data may not accurately predict the behavior of other groups. This limitation can lead product teams to unknowingly exclude certain user groups, exacerbating the digital divide.
The second important challenge is the control of calculation costs. AI-powered user simulation requires significant computational resources, especially when it involves simulating a large number of different user personas and complex interaction scenarios. How to control costs while maintaining simulation quality is a key issue for Blok to continuously optimize. This also explains why they place a special emphasis on balancing computational costs in their business model.
Data privacy and security are another significant challenge. While Blok has chosen a synthetic data-based approach to avoid direct processing of personal information, ensuring that behavioral patterns extracted from raw data do not compromise user privacy remains a concern that requires ongoing attention. The zero-knowledge proof and data enclave technologies they are exploring are promising, but they still face many challenges in engineering implementation.
The generalization ability of the model is also a key technical challenge. There are significant differences in user behavior patterns in different industries, different cultural backgrounds, and different product types. How to build a model that can accurately reflect the characteristics of specific user groups and have sufficient generalization capabilities is the core problem that the Blok technical team needs to solve.
In terms of future development direction, Blok has several technological evolution paths worth paying attention to. The first is the real-time learning capability of the model. Current models are primarily trained on historical data, but ideal systems should be able to adjust and optimize simulations in real-time based on the latest user behavior patterns.
The second is the simulation ability of multimodal interaction. With the popularity of voice, gestures, eye movements, and other interaction methods, Blok needs to incorporate these new interaction patterns into user behavior simulations. This requires not only technological breakthroughs, but also a deep understanding of human-computer interaction theory.
The third is cross-platform and cross-device user behavior modeling. Modern users often use the same product on multiple devices and platforms, and how to build a unified cross-platform user behavior model is a technical challenge with important business value.
Finally, Blok needs to consider how to integrate emerging AI technologies (such as large language models, multimodal models) into their user behavior simulation system. These new technologies may offer new possibilities for user behavior prediction, but they also require Blok’s continuous investment in R&D resources to explore and validate.
Prediction of the evolution of the competitive landscape of the industry
Based on my in-depth analysis of Blok and my observations of the product testing industry as a whole, I have some forward-looking thoughts on the future evolution of the competitive landscape. This evolution will not only affect the market position of existing players but may also give rise to entirely new market segments and business models.
First, I predict that traditional A/B testing tool providers will face significant transformation pressure. Companies like Optimizely and Amplitude may be marginalized if they don’t quickly integrate predictive testing capabilities. But the advantage of these companies is that they have a large number of existing customers and data resources, and if they can successfully transform, they will still be highly competitive.
Second, I expect more big tech companies to start deploying in this space. Companies like Google, Microsoft, Adobe, and others are motivated to develop similar capabilities to enhance their existing suite of product development tools. This competition may drive technological advancements across the industry, but it will also create more competitive pressure for startups like Blok.
Third, I think there will be more competitors that focus on specific verticals. For example, user behavior simulation tools specifically for specific industries such as e-commerce, fintech, and healthcare. This trend towards specialization may make the market more segmented, but it also presents opportunities for generic platforms like Blok to compete differently.
Fourth, I predict that the open source community will also play an important role in this area. As the technology matures, open-source user behavior simulation frameworks may emerge, which will lower the barrier to entry but also increase competition. Blok needs to find a balance between the open-source trend and its commercial value.
From an investment perspective, I think this space will attract more venture capital attention. As first-mover players like Blok prove the viability of business models, more capital will enter this market, driving technological innovation and market expansion. But it also means that competition will become more intense.
In terms of technology trends, I expect future competition to focus on several key dimensions: simulation accuracy, computational efficiency, ease of integration, privacy protection, and industry-specific optimization. Companies that lead in these dimensions will gain a competitive advantage.
Most importantly, I think the growth of this industry will drive a fundamental change in product development methodologies. Product teams in the future may use virtual user testing as part of their standard workflow, just as they use version control and continuous integration today. This change in methodology will create significant market opportunities, but it will also require continuous innovation from all players to adapt to change
