CHAPTERS
- 0:22 – 0:31
Diode’s mission: AI that automates PCB (circuit board) design
Diana Hu opens by introducing Diode’s founders and their rapid traction post–YC, including a newly closed Series A. The founders frame Diode as an AI-enabled PCB design shop that automates board design end-to-end.
- •Diode uses AI to automate printed circuit board design
- •Positioned as an “AI-enabled design shop” rather than a point tool
- •Fast progress after YC Summer ’24 and $11.4M Series A led by a16z
- 0:31 – 0:51
Who’s buying: Fortune 100 + frontier hardware startups
The team describes a surprisingly strong customer roster for such an early company. They serve both large enterprises and ambitious hardware startups that need PCB design capacity quickly.
- •Customers include Fortune 100 companies
- •Also serving hardware-forward startups (e.g., robotics and autonomous systems)
- •Broad demand from teams that need boards designed without building a full in-house EE org
- 0:51 – 2:46
Origin story: software engineer meets hardware pain firsthand
Lenny recounts moving from software into hardware during COVID, mentored by Davide, and discovering the pace and tooling gap in electronics. They later worked together again and saw repeated “silly mistakes” in boards despite strong teams.
- •Early exposure: Lenny joined a hardware team and learned under Davide
- •Hardware iteration cycles are slow compared to software feedback loops
- •Tooling limitations lead to avoidable errors and long turnaround times
- 2:46 – 4:01
Initial YC thesis: “compiler-like” verification for PCB mistakes
The founders originally believed the key need was verification—catching errors the way compilers catch bugs in code. They built technical infrastructure to generate boards, inject mistakes, and train systems to detect issues.
- •Original product idea: automated PCB mistake detection/verification
- •Built a generative pipeline to create boards and inject errors for training
- •Motivated by LLMs’ ability to parse documentation and spot inconsistencies
- 4:01 – 5:02
Customer reality check: 100+ conversations and a painful pivot
After extensive user discovery, they learned customers didn’t value the “find my mistakes” pitch—whether because of pride or process, it wasn’t the perceived pain. The real pain was getting the design produced in the first place.
- •Users responded: “We don’t make mistakes,” so verification wasn’t compelling
- •Key lesson: pain point wasn’t auditing existing designs
- •Pivot insight: the bottleneck is generating the PCB design itself
- 5:02 – 5:29
Finding the right problem through YC network: the Jetson Orin moment
A chance conversation with another YC founder revealed repeated demand for custom Jetson Orin development boards in robotics. That concrete, repeatable need gave Diode a crisp wedge and immediate direction.
- •Signal came from a batchmate hearing repeated customer requests
- •Specific wedge: custom Jetson Orin dev boards for robotics teams
- •YC environment accelerated discovery of a high-intensity problem
- 5:29 – 6:40
First deal and instant validation: customers want the full solution
Unlike the lukewarm reception to verification, the new pitch landed immediately—an instant “yes.” They realized buyers didn’t want a component tool; they wanted someone to own the outcome because PCB talent is scarce and outsourcing is common.
- •First conversation under new framing produced an immediate yes
- •Market driver: shortage of PCB designers and limited internal bandwidth
- •Value proposition: deliver finished boards, not just productivity tooling
- 6:40 – 11:28
Services-as-a-product: solving the 80/20 with AI + owned verification
Diana challenges how a services-like offering becomes venture-scale; Davide explains the model: AI makes them dramatically more productive, while the team still owns correctness. They combine automation with a verification pipeline and explicit acceptance criteria, resembling software delivery but faster.
- •AI enables “software-like margins” on what looks like services
- •Key concept: 80/20 automation works when experts own the last-mile quality
- •Diode owns verification and delivery against client acceptance criteria
- 11:28 – 11:56
Reframing PCB design as a software problem (and unlocking LLM capability)
Lenny explains that modern models already “know” much of electrical engineering, but traditional PCB tools are visual and don’t match how LLMs work. By translating PCB design into code, Diode lets models express their latent knowledge, then outputs visuals engineers are comfortable reviewing.
- •LLMs have broad EE knowledge but need a code-friendly interface
- •Traditional EDA tooling is graphical and largely unchanged for decades
- •Use code as an intermediate representation; export conventional visuals for humans
- 11:56 – 12:31
Velocity in practice: building internal tools to ship 100+ boards
The founders discuss how they split responsibilities—Davide leading EE and Lenny leading software—treating the EE team as the first customer of internal tooling. That tooling makes it feasible for a tiny team to manage many clients and produce an unusually high volume of designs.
- •Two-person early team relied on custom software to multiply output
- •Davide’s EE workflow accelerated by tools built by Lenny’s software team
- •Open-source tooling (e.g., KiCad) helped, but Diode built a faster layer on top
- 12:31 – 13:33
Technical architecture: language design + Rust/Wasm for conservative industries
To sell into aerospace/medical and other security-sensitive sectors, Diode made deliberate technical choices. They created a schematic language that works for both humans and LLMs, and built an isolated core compiler in Rust with Wasm bindings to enable local, browser-based compilation and visualization.
- •Designed a schematic language optimized for both LLM generation and human review
- •Rust-based compiler implements core logic with strong isolation properties
- •Wasm bindings enable in-browser visualization and real-time local compilation
- •Supports strict environments (including air-gapped deployments) and configurable modules
- 13:33 – 14:21
Future outlook: bringing AI’s “10x” gains from software into the physical world
Davide describes the broader ambition: apply AI acceleration to real-world, physical systems, where reliability and correctness matter deeply. They see hardware as the next major frontier for top engineers and a core driver of long-term impact.
- •Vision: apply AI-driven speedups to designing physical objects
- •Hardware seen as the next “Everest” after software automation gains
- •Emphasis on real-world reliability and systems that work outside the lab
- 14:21 – 15:28
Hiring and team building: researchy problem-solvers who also ship
Lenny closes with what they look for in recruits: curiosity, comfort with open-ended exploration, and the discipline to deliver practical outcomes. They also emphasize product-minded engineers who can build customer-facing interfaces for reviewing and handing off designs.
- •Seeking engineers excited by ambiguous, multi-solution technical problems
- •Balance experimentation with shipping real boards and products
- •Strong focus on customer experience: review tools, interfaces, and collaboration workflows
