Key Moments
Max Hodak: What Really Kills Deep Tech Startups?
Y CombinatorWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
Deep tech startups fail not due to technology, but due to poor operational infrastructure like purchasing and hiring. Building these systems is crucial for speed and survival.
Key Insights
Deep tech companies often fail due to organizational and logistical issues, not technological shortcomings, with purchasing and hiring being critical infrastructure areas.
Science's retinal implant, which helps blind patients see, has undergone three clinical trials, with one patient reading a 300-page novel using the device.
A company's speed is determined by its infrastructure, which includes efficient purchasing systems that avoid bottlenecks and allow highly paid employees to focus on innovation, not procurement.
Science uses a company-wide voting system involving 7-8 employees to review initial job applications, averaging judgment and preventing bottlenecks.
The 'iGEN reviews' system, inspired by Google's PageRank, uses a weighted voting mechanism to provide continuous, unbiased feedback on employee performance every 4-6 weeks.
Founders cannot delegate their judgment and must make differentiated decisions, as becoming average in startups leads to failure; action produces information and helps overcome being stuck.
Deep tech startups are often derailed by infrastructure, not technology
Max Hodak, CEO of Science, emphasizes that the primary reason deep tech startups fail is not a lack of technological innovation, but deficiencies in their operational infrastructure. He likens this to artists discussing cheap turpentine instead of art, drawing a parallel to professionals focusing on logistics over strategy. While pure software companies might operate with minimal physical infrastructure, deep tech companies, which often deal with the physical world (e.g., rockets, drugs, brain-computer interfaces), face continuous needs for purchasing equipment, raw materials, and complex components. The speaker highlights that the core challenge lies not just in acquiring these items but in establishing systems that allow for efficient and timely procurement, which directly impacts the company's ability to move quickly and effectively.
Streamlining procurement is vital for maintaining speed
A common pitfall for startups is the inefficient purchasing process. Initially, founders might rely on credit cards, but as the team grows, this becomes unmanageable. Approving every purchase can lead to significant delays, especially when dealing with expensive or specialized equipment. For instance, a $3,000 power supply might seem excessive, prompting a founder to consider waiting for an auction to save money. However, Hodak argues that the cost of delaying a highly paid employee's work for two weeks can far outweigh the savings on the component. Implementing a formal procurement system, while necessary for budget control, can also introduce its own delays due to vendor account setup, paperwork, and quoting processes. Science's solution involves building internal software that integrates purchasing with other operational workflows, ensuring speed and control without sacrificing efficiency. This focus on infrastructure allows for rapid iteration, which Hodak posits is the ultimate determinant of success in deep tech.
Accurate cost attribution and resource management are essential
Beyond just buying things, a critical infrastructure component is the ability to attribute costs accurately, especially when dealing with bulk materials used across various experiments. When the cost of individual experiments or manufactured items is unknown, they are effectively perceived as free, leading to wasteful practices. Understanding the true cost of producing something, whether it's a cell line or a wafer, requires detailed tracking of resources, including rent, depreciation, and future volume projections. Science developed extensive internal software, dubbed 'Helix,' to manage these processes. By capturing every step from procurement to manufacturing in a database, they can correlate data and determine precise costs, such as $40,000 per wafer iteration. This granular understanding of expenses is crucial for financial planning, determining runway, and making informed pricing decisions, preventing the common startup issue where budgets are quickly depleted due to a lack of financial visibility.
Hiring processes must be efficient and scalable
Hiring is another universal challenge that separates successful startups from failures. Hodak emphasizes the importance of recruiting from existing networks and the 'scene' from which the startup emerged, as these individuals already understand the company's language and mission. However, relying solely on this network is insufficient for scaling. For public hires, a defined, efficient process is paramount. Science employs a four-step system: online application capture, company-wide initial voting, a phone screen focused on judgment, horsepower, and agency, and a take-home assignment or technical call. The initial voting stage involves randomly selecting 7-8 employees to assess candidates, averaging judgment across the company and preventing single points of failure. This structured approach, supported by internal software, ensures that hiring doesn't become a bottleneck, allowing for rapid review and assessment. Approximately 17% of initial applications proceed to a phone screen, and half of those move to a homework assignment.
Continuous performance feedback is key to organizational health
Traditional annual performance reviews are often seen as disruptive, infrequent, and slow to surface existing issues. Hodak advocates for a more continuous feedback mechanism. Science uses a system called 'iGEN reviews,' which prompts employees every 4-6 weeks to answer a single question: 'Knowing how this person turned out, would you vote again today for their hire?' This process, inspired by Google's PageRank algorithm and utilizing eigenvector centrality, constructs a weighted graph of feedback across the company. An employee's vote is weighted more heavily if they are highly rated by others. This system provides ongoing insight into performance, identifies potential issues early, and helps average judgment across the organization, moving away from a single annual review cycle. Techniques like Monte Carlo dropout are used to detect potential collusion or voting rings, ensuring the integrity of the feedback.
Infrastructure dictates iteration speed, which drives success
Hodak reiterates that the rate of iteration is what separates success from failure in deep tech. Companies that can learn and adapt weekly will inevitably outperform those that learn monthly. This speed is directly enabled by robust infrastructure, encompassing purchasing, recruiting, spending, and performance review processes. These 'boring' logistical elements are as crucial as the core technical knowledge. Companies often fail not because their technology is flawed, but because they lack the systems to manage an organization of hundreds of people and extensive physical infrastructure. Establishing solid foundational systems early on makes subsequent growth and execution significantly easier, preventing the loss of control over spending and the need for drastic, coarser measures later on.
Founders must own their judgment and differentiate
A fundamental challenge for startup founders is the inability to delegate their core judgment. While advice and external input are valuable, the ultimate decisions must align with the founder's own differentiated thinking. In academic settings, averaging performance can lead to acceptable outcomes; in startups, becoming average leads to failure. Founders must strive for judgment that is not just good, but distinctively so. This often requires making decisions that may not be popular or immediately obvious, especially during high-stakes moments years into the company's development. The ability to act based on one's own conviction, even when alone in that perspective, is critical for achieving significant outcomes. This process of making decisions and observing their outcomes generates crucial information, helping founders refine their judgment over time.
Biotech and neural interfaces offer profound impact but require careful navigation
While the field of BCIs and neural interfaces is interdisciplinary and complex, Hodak highlights their potential to significantly improve lives, particularly in healthcare and longevity. He notes that while AI may focus on super-intelligent machines, BCIs aim for conscious machines, with potential convergence in the future. However, in the near term, BCI's impact is seen as a 'longevity story' and a form of advanced healthcare. Unlike pharmaceuticals, which often provide incremental improvements, neural engineering can achieve dramatic effect sizes, such as restoring movement or vision almost instantly. This engineering-driven approach, treating the brain as a computer, bypasses many difficult biological problems. Hodak also touches upon the challenging engineering aspects of implants, including power, thermal constraints, and packaging, stressing the need for material science advancements. The biotech field remains capital-intensive and demanding, requiring founders to commit a decade or more. However, the power to heal and reshape the world through such interventions offers a unique and profound impact, distinct from economic or military power.
Mentioned in This Episode
●Software & Apps
●Companies
●Organizations
●People Referenced
Common Questions
Deep tech startups often face significant challenges in infrastructure, particularly with procurement, inventory management, and accurate cost attribution. Unlike pure software companies, those dealing with physical products must manage the complexities of acquiring diverse materials and equipment, which can lead to delays and cost overruns if not handled efficiently.
Topics
Mentioned in this video
Mentioned as an example of a company that wouldn't be hassled over a $3,000 purchase.
Mentioned as a place where top computer scientists are now located.
Mentioned as a company where top rocket scientists work and as an example of internal software development ('Warp Speed').
Mentioned as a company where top rocket scientists work.
Mentioned as a place where top computer scientists are now located and for its original PageRank algorithm.
Famously has a lot of internal software that is key to its operations.
Famously invested heavily in internal tools, leading to significant efficiency gains.
Internally uses a large piece of software called 'Warp Speed' for manufacturing and R&D processes.
More from Y Combinator
View all 621 summaries
57 minHow To Design In The Agent Era
43 minGarry Tan: "Personal AGI Is How You Stay Under Your Own Power"
37 minBuilding the First Data Centers in Space
50 minWaymo Co-CEO Dmitri Dolgov: The Demo Is Only 1% Of The Work
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free