Jul 26, 2024 · 50m · sourcery

Text-to-CAD: AI Revolutionizing Hardware Design with Jordan Noone of Zoo · Sourcery with Molly O'Shea

Jordan Noone · 40m spoken Molly O'Shea · 5m spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

In this episode of Sourcery, host Molly O'Shea interviews Jordan Noone, Executive Chairman of Zoo and former co-founder of Relativity Space, to explore how Zoo's GPU-optimized computational geometry engine and AI-driven Text-to-CAD tools are modernizing hardware engineering.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Molly holds 11.9% of the talking time here. How this is scored →

Molly as informed peer 3.1 Guest teaching 5.5 Guest disagreement 0.9 Molly pushing back 0.4
05100:0015:0030:0045:000:00–2:25 · Molly as informed peer 3/10 Jordan Noone on the Text-to-CAD Launch Molly introduces Jordan Noone and gently resets the conversation after a heavy technical pitch, asking him to step back and contextualize the market size. Jordan explains the trillion-dollar hardware development ecosystem still stuck using 1980s mouse-driven paradigms.2:25–5:57 · Molly as informed peer 4/10 Traditional CAD vs. Modern Computation Molly prompts Jordan to explain the traditional aerospace CAD workflow. Jordan details how geometric design has remained essentially unchanged since Autodesk's AutoCAD in the 1980s, explaining the combinatorial explosion of manual interface communication.5:57–9:18 · Molly as informed peer 3/10 Addressing the Hardware-Software Gap Molly asks which specific niche of hardware development Zoo targets. Jordan educates on the asymmetry between modern automated software CI/CD pipelines and manual, bottlenecked hardware engineering workflows.9:18–13:32 · Molly as informed peer 3/10 Zoo's Product Suite Overview Jordan gives a deep technical explanation of Zoo's GPU-accelerated computational geometry engine built in CUDA from scratch, describing computational geometry as a lost art abandoned since the 1980s.13:32–15:50 · Molly as informed peer 1/10 Sponsor Announcement: Archer Aviation The segment begins with a sponsor read by Molly for Archer Aviation, followed by Jordan describing Zoo's v1.0 unification of Text-to-CAD and direct modeling tools to solve prompt-engineering limits.15:50–21:40 · Molly as informed peer 4/10 Developer Platform and Open-Source Dogfooding Molly poses two competing questions regarding onboarding non-technical mechanical engineers versus providing open-source APIs. Jordan explains their dual strategy of dogfooding open tools while protecting and monetizing the backend geometry engine API.21:40–30:21 · Molly as informed peer 3/10 The Origins and Leadership of Zoo Jordan shares his backstory from USC Rocket Propulsion Lab to SpaceX, Relativity Space, and Embedded Ventures, explaining how CEO Jess Frazelle's frustrations at Oxide Computer led to the inception of Zoo.30:21–35:02 · Molly as informed peer 3/10 Technical Team Breakdown and ML Training Molly inquires about the team's composition. Jordan breaks down their specialized engineering groups across computational geometry, ML, infrastructure, and virtual dataset generation via parallel cloud computation.35:02–44:52 · Molly as informed peer 4/10 Hardware Lifecycle and CNC Toolpathing Molly asks why a software company requires an in-house factory and data center. Jordan explains their expansion downstream into CNC toolpath generation and design-for-manufacturability, necessitating in-house physical hardware verification.44:52–47:40 · Molly as informed peer 3/10 Returning to Relativity's Original Headquarters Molly notes Zoo's location in Inglewood and asks about their venture funding history. Jordan explains operating lean on $6M, highlighting their lead investor Venrex and angel backing from former GitHub CEOs Nat Friedman and Tom Preston-Werner.47:40–49:43 · Molly as informed peer 3/10 Future Product Roadmap and Milestones Molly asks for the roadmap over the next 6-12 months. Jordan recaps the upcoming launch allowing manual editing of Text-to-CAD outputs and beta testing of end-to-end manufacturing workflows before Molly closes the show.0:00–2:25 · Guest teaching 5/10 Jordan Noone on the Text-to-CAD Launch Molly introduces Jordan Noone and gently resets the conversation after a heavy technical pitch, asking him to step back and contextualize the market size. Jordan explains the trillion-dollar hardware development ecosystem still stuck using 1980s mouse-driven paradigms.2:25–5:57 · Guest teaching 6/10 Traditional CAD vs. Modern Computation Molly prompts Jordan to explain the traditional aerospace CAD workflow. Jordan details how geometric design has remained essentially unchanged since Autodesk's AutoCAD in the 1980s, explaining the combinatorial explosion of manual interface communication.5:57–9:18 · Guest teaching 6/10 Addressing the Hardware-Software Gap Molly asks which specific niche of hardware development Zoo targets. Jordan educates on the asymmetry between modern automated software CI/CD pipelines and manual, bottlenecked hardware engineering workflows.9:18–13:32 · Guest teaching 7/10 Zoo's Product Suite Overview Jordan gives a deep technical explanation of Zoo's GPU-accelerated computational geometry engine built in CUDA from scratch, describing computational geometry as a lost art abandoned since the 1980s.13:32–15:50 · Guest teaching 4/10 Sponsor Announcement: Archer Aviation The segment begins with a sponsor read by Molly for Archer Aviation, followed by Jordan describing Zoo's v1.0 unification of Text-to-CAD and direct modeling tools to solve prompt-engineering limits.15:50–21:40 · Guest teaching 5/10 Developer Platform and Open-Source Dogfooding Molly poses two competing questions regarding onboarding non-technical mechanical engineers versus providing open-source APIs. Jordan explains their dual strategy of dogfooding open tools while protecting and monetizing the backend geometry engine API.21:40–30:21 · Guest teaching 6/10 The Origins and Leadership of Zoo Jordan shares his backstory from USC Rocket Propulsion Lab to SpaceX, Relativity Space, and Embedded Ventures, explaining how CEO Jess Frazelle's frustrations at Oxide Computer led to the inception of Zoo.30:21–35:02 · Guest teaching 6/10 Technical Team Breakdown and ML Training Molly inquires about the team's composition. Jordan breaks down their specialized engineering groups across computational geometry, ML, infrastructure, and virtual dataset generation via parallel cloud computation.35:02–44:52 · Guest teaching 7/10 Hardware Lifecycle and CNC Toolpathing Molly asks why a software company requires an in-house factory and data center. Jordan explains their expansion downstream into CNC toolpath generation and design-for-manufacturability, necessitating in-house physical hardware verification.44:52–47:40 · Guest teaching 4/10 Returning to Relativity's Original Headquarters Molly notes Zoo's location in Inglewood and asks about their venture funding history. Jordan explains operating lean on $6M, highlighting their lead investor Venrex and angel backing from former GitHub CEOs Nat Friedman and Tom Preston-Werner.47:40–49:43 · Guest teaching 4/10 Future Product Roadmap and Milestones Molly asks for the roadmap over the next 6-12 months. Jordan recaps the upcoming launch allowing manual editing of Text-to-CAD outputs and beta testing of end-to-end manufacturing workflows before Molly closes the show.0:00–2:25 · Guest disagreement 1/10 Jordan Noone on the Text-to-CAD Launch Molly introduces Jordan Noone and gently resets the conversation after a heavy technical pitch, asking him to step back and contextualize the market size. Jordan explains the trillion-dollar hardware development ecosystem still stuck using 1980s mouse-driven paradigms.2:25–5:57 · Guest disagreement 2/10 Traditional CAD vs. Modern Computation Molly prompts Jordan to explain the traditional aerospace CAD workflow. Jordan details how geometric design has remained essentially unchanged since Autodesk's AutoCAD in the 1980s, explaining the combinatorial explosion of manual interface communication.5:57–9:18 · Guest disagreement 1/10 Addressing the Hardware-Software Gap Molly asks which specific niche of hardware development Zoo targets. Jordan educates on the asymmetry between modern automated software CI/CD pipelines and manual, bottlenecked hardware engineering workflows.9:18–13:32 · Guest disagreement 1/10 Zoo's Product Suite Overview Jordan gives a deep technical explanation of Zoo's GPU-accelerated computational geometry engine built in CUDA from scratch, describing computational geometry as a lost art abandoned since the 1980s.13:32–15:50 · Guest disagreement 0/10 Sponsor Announcement: Archer Aviation The segment begins with a sponsor read by Molly for Archer Aviation, followed by Jordan describing Zoo's v1.0 unification of Text-to-CAD and direct modeling tools to solve prompt-engineering limits.15:50–21:40 · Guest disagreement 1/10 Developer Platform and Open-Source Dogfooding Molly poses two competing questions regarding onboarding non-technical mechanical engineers versus providing open-source APIs. Jordan explains their dual strategy of dogfooding open tools while protecting and monetizing the backend geometry engine API.21:40–30:21 · Guest disagreement 2/10 The Origins and Leadership of Zoo Jordan shares his backstory from USC Rocket Propulsion Lab to SpaceX, Relativity Space, and Embedded Ventures, explaining how CEO Jess Frazelle's frustrations at Oxide Computer led to the inception of Zoo.30:21–35:02 · Guest disagreement 1/10 Technical Team Breakdown and ML Training Molly inquires about the team's composition. Jordan breaks down their specialized engineering groups across computational geometry, ML, infrastructure, and virtual dataset generation via parallel cloud computation.35:02–44:52 · Guest disagreement 1/10 Hardware Lifecycle and CNC Toolpathing Molly asks why a software company requires an in-house factory and data center. Jordan explains their expansion downstream into CNC toolpath generation and design-for-manufacturability, necessitating in-house physical hardware verification.44:52–47:40 · Guest disagreement 0/10 Returning to Relativity's Original Headquarters Molly notes Zoo's location in Inglewood and asks about their venture funding history. Jordan explains operating lean on $6M, highlighting their lead investor Venrex and angel backing from former GitHub CEOs Nat Friedman and Tom Preston-Werner.47:40–49:43 · Guest disagreement 0/10 Future Product Roadmap and Milestones Molly asks for the roadmap over the next 6-12 months. Jordan recaps the upcoming launch allowing manual editing of Text-to-CAD outputs and beta testing of end-to-end manufacturing workflows before Molly closes the show.0:00–2:25 · Molly pushing back 2/10 Jordan Noone on the Text-to-CAD Launch Molly introduces Jordan Noone and gently resets the conversation after a heavy technical pitch, asking him to step back and contextualize the market size. Jordan explains the trillion-dollar hardware development ecosystem still stuck using 1980s mouse-driven paradigms.2:25–5:57 · Molly pushing back 1/10 Traditional CAD vs. Modern Computation Molly prompts Jordan to explain the traditional aerospace CAD workflow. Jordan details how geometric design has remained essentially unchanged since Autodesk's AutoCAD in the 1980s, explaining the combinatorial explosion of manual interface communication.5:57–9:18 · Molly pushing back 0/10 Addressing the Hardware-Software Gap Molly asks which specific niche of hardware development Zoo targets. Jordan educates on the asymmetry between modern automated software CI/CD pipelines and manual, bottlenecked hardware engineering workflows.9:18–13:32 · Molly pushing back 0/10 Zoo's Product Suite Overview Jordan gives a deep technical explanation of Zoo's GPU-accelerated computational geometry engine built in CUDA from scratch, describing computational geometry as a lost art abandoned since the 1980s.13:32–15:50 · Molly pushing back 0/10 Sponsor Announcement: Archer Aviation The segment begins with a sponsor read by Molly for Archer Aviation, followed by Jordan describing Zoo's v1.0 unification of Text-to-CAD and direct modeling tools to solve prompt-engineering limits.15:50–21:40 · Molly pushing back 1/10 Developer Platform and Open-Source Dogfooding Molly poses two competing questions regarding onboarding non-technical mechanical engineers versus providing open-source APIs. Jordan explains their dual strategy of dogfooding open tools while protecting and monetizing the backend geometry engine API.21:40–30:21 · Molly pushing back 0/10 The Origins and Leadership of Zoo Jordan shares his backstory from USC Rocket Propulsion Lab to SpaceX, Relativity Space, and Embedded Ventures, explaining how CEO Jess Frazelle's frustrations at Oxide Computer led to the inception of Zoo.30:21–35:02 · Molly pushing back 0/10 Technical Team Breakdown and ML Training Molly inquires about the team's composition. Jordan breaks down their specialized engineering groups across computational geometry, ML, infrastructure, and virtual dataset generation via parallel cloud computation.35:02–44:52 · Molly pushing back 0/10 Hardware Lifecycle and CNC Toolpathing Molly asks why a software company requires an in-house factory and data center. Jordan explains their expansion downstream into CNC toolpath generation and design-for-manufacturability, necessitating in-house physical hardware verification.44:52–47:40 · Molly pushing back 0/10 Returning to Relativity's Original Headquarters Molly notes Zoo's location in Inglewood and asks about their venture funding history. Jordan explains operating lean on $6M, highlighting their lead investor Venrex and angel backing from former GitHub CEOs Nat Friedman and Tom Preston-Werner.47:40–49:43 · Molly pushing back 0/10 Future Product Roadmap and Milestones Molly asks for the roadmap over the next 6-12 months. Jordan recaps the upcoming launch allowing manual editing of Text-to-CAD outputs and beta testing of end-to-end manufacturing workflows before Molly closes the show.

speaking balance: gold is Molly, purple is the guest (3 minute bins)

0:00 · Molly 44.4% · guest 55.6%0:00 · Molly 44.4% · guest 55.6%3:00 · Molly 1.1% · guest 98.9%3:00 · Molly 1.1% · guest 98.9%6:00 · Molly 13% · guest 87%6:00 · Molly 13% · guest 87%9:00 · Molly 14.1% · guest 85.9%9:00 · Molly 14.1% · guest 85.9%12:00 · Molly 23.4% · guest 76.6%12:00 · Molly 23.4% · guest 76.6%15:00 · Molly 24.9% · guest 75.1%15:00 · Molly 24.9% · guest 75.1%18:00 · Molly 0% · guest 100%18:00 · Molly 0% · guest 100%21:00 · Molly 10.5% · guest 89.5%21:00 · Molly 10.5% · guest 89.5%24:00 · Molly 0% · guest 100%24:00 · Molly 0% · guest 100%27:00 · Molly 0% · guest 100%27:00 · Molly 0% · guest 100%30:00 · Molly 12.3% · guest 87.7%30:00 · Molly 12.3% · guest 87.7%33:00 · Molly 29.6% · guest 70.4%33:00 · Molly 29.6% · guest 70.4%36:00 · Molly 0% · guest 100%36:00 · Molly 0% · guest 100%39:00 · Molly 0% · guest 100%39:00 · Molly 0% · guest 100%42:00 · Molly 3.9% · guest 96.1%42:00 · Molly 3.9% · guest 96.1%45:00 · Molly 17.8% · guest 82.2%45:00 · Molly 17.8% · guest 82.2%48:00 · Molly 9.5% · guest 90.5%48:00 · Molly 9.5% · guest 90.5%
Sharpest disagreement ▶ 27:20 Dismissal of shallow CAD tools and dashboard companies

Jordan forcefully dismisses legacy software incumbents and modern 'dashboard startups', pointing out that there has not been an engineering software IPO since 1986 and that browser ports fail to solve real workflows.

Hardest push from Molly ▶ 1:21 Reframing the dense technical elevator pitch

Molly interrupts Jordan's dense initial pitch to demand context, explicitly asking him to step back and clarify the real-world scale and target customer before going deeper into tech jargon.

Biggest teaching moment ▶ 10:05 Explaining the loss of computational geometry skills

Jordan educates the host on how modern software talent completely shifted toward ML, causing core 1980s algorithmic skills in computational geometry to become an almost extinct discipline.

Molly holds their own ▶ 2:25 Prompting aerospace development cycle comparisons

Molly demonstrates industry grasp by contextualizing Jordan's 60-year disruption claim and framing the discussion around specific aerospace engineering cycles and interface bottlenecks.

the scores for every segment, with the reasoning behind each
ChapterTopicMolly as informed peerGuest teachingGuest disagreementMolly pushing backWhy
Jordan Noone on the Text-to-CAD Launch 3512 Molly introduces Jordan Noone and gently resets the conversation after a heavy technical pitch, asking him to step back and contextualize the market size. Jordan explains the trillion-dollar hardware development ecosystem still stuck using 1980s mouse-driven paradigms.
Traditional CAD vs. Modern Computation 4621 Molly prompts Jordan to explain the traditional aerospace CAD workflow. Jordan details how geometric design has remained essentially unchanged since Autodesk's AutoCAD in the 1980s, explaining the combinatorial explosion of manual interface communication.
Addressing the Hardware-Software Gap 3610 Molly asks which specific niche of hardware development Zoo targets. Jordan educates on the asymmetry between modern automated software CI/CD pipelines and manual, bottlenecked hardware engineering workflows.
Zoo's Product Suite Overview 3710 Jordan gives a deep technical explanation of Zoo's GPU-accelerated computational geometry engine built in CUDA from scratch, describing computational geometry as a lost art abandoned since the 1980s.
Sponsor Announcement: Archer Aviation 1400 The segment begins with a sponsor read by Molly for Archer Aviation, followed by Jordan describing Zoo's v1.0 unification of Text-to-CAD and direct modeling tools to solve prompt-engineering limits.
Developer Platform and Open-Source Dogfooding 4511 Molly poses two competing questions regarding onboarding non-technical mechanical engineers versus providing open-source APIs. Jordan explains their dual strategy of dogfooding open tools while protecting and monetizing the backend geometry engine API.
The Origins and Leadership of Zoo 3620 Jordan shares his backstory from USC Rocket Propulsion Lab to SpaceX, Relativity Space, and Embedded Ventures, explaining how CEO Jess Frazelle's frustrations at Oxide Computer led to the inception of Zoo.
Technical Team Breakdown and ML Training 3610 Molly inquires about the team's composition. Jordan breaks down their specialized engineering groups across computational geometry, ML, infrastructure, and virtual dataset generation via parallel cloud computation.
Hardware Lifecycle and CNC Toolpathing 4710 Molly asks why a software company requires an in-house factory and data center. Jordan explains their expansion downstream into CNC toolpath generation and design-for-manufacturability, necessitating in-house physical hardware verification.
Returning to Relativity's Original Headquarters 3400 Molly notes Zoo's location in Inglewood and asks about their venture funding history. Jordan explains operating lean on $6M, highlighting their lead investor Venrex and angel backing from former GitHub CEOs Nat Friedman and Tom Preston-Werner.
Future Product Roadmap and Milestones 3400 Molly asks for the roadmap over the next 6-12 months. Jordan recaps the upcoming launch allowing manual editing of Text-to-CAD outputs and beta testing of end-to-end manufacturing workflows before Molly closes the show.

Statements from this episode (19)

Opinion
Noone: Hardware development labor and software represent a $1 trillion market
“And so for us, we view it as about a trillion dollar sized market. That's both the labor and the software license costs for actually running those hardware development programs.”
Jordan Noone Jul 26, 2024 ▶ 1:51
Opinion
Noone: Hardware design software has not been fundamentally updated since the 1980s
“Then it last was updated in the eighties. And it's still a industry fully driven by engineers with a mouse clicking on a screen.”
Jordan Noone Jul 26, 2024 ▶ 2:07
Insight
Noone: Manual CAD interfaces scale engineering communication complexity quadratically
“It's a very, again, labor-driven, very slow process that ends up compounding, you know, the number of times or number of parts you have interfacing squared. That's the number of interfaces you have. That's the number of engineers you have having kind of cross …”
Jordan Noone Jul 26, 2024 ▶ 4:18
Assertion Not checkable as stated
Jordan Noone: Zoo sees highest adoption from combined hardware-software teams
“The people who we tend to see the most adoption from are people that see software development and hardware development side by side.”
Jordan Noone Jul 26, 2024 ▶ 7:13
Assertion Partly supported
Noone: Zoo built the first GPU-optimized cloud CAD geometry engine using CUDA
“So our view was we would start with this geometry engine. That's the world's first GPU optimized cloud-based API accessible CAD engine. We built that from scratch. So that's CUDA, which is kind of GPU level machine language CUDA level implementation of those c…”
Jordan Noone Jul 26, 2024 ▶ 10:12
Assertion Contradicted
Noone: Hardware engineers historically lacked code-based CAD automation tools
“And that code allows scripting, automation, scaling, things like equations and database lookups and logic, things that sound obvious to anyone who's done software before. Of course, you'd think that software could write a loop that changes the geometry of a fe…”
Jordan Noone Jul 26, 2024 ▶ 11:54
Disclosure
Zoo to merge Text-to-CAD directly into CAD modeling app in V1.0
“Right now we're doing a push for what we call version one point O of both of those products. Those will actually merge together that rather than generating text to CAD and having a part, you can now do that in the modeling app and then make downstream changes.”
Jordan Noone Jul 26, 2024 ▶ 14:16
Insight
Noone: The inability to edit outputs is generative AI's biggest bottleneck
“You get 95% of the way there, and there's one feature in your image that you need different, and there's nothing you can do about it. Then, other than try to slightly change your prompt, slightly perfect your prompt, it's a huge bottleneck in the generative AI…”
Jordan Noone Jul 26, 2024 ▶ 14:39
Disclosure
Noone: Zoo monetizes the majority of its traffic via API
“And the reason we started open sourcing those tools is one that those tools still drive traffic through our API. That's where we monetize a majority of our incoming traffic and is that they still create traffic on top of us.”
Jordan Noone Jul 26, 2024 ▶ 19:21
Assertion Supported
Noone was the youngest person to receive FAA rocket launch clearance
“I was the first person in the world, the youngest person in the world to get FAA clearance to fly a rocket to space.”
Jordan Noone Jul 26, 2024 ▶ 22:14
Assertion Supported
Noone: Relativity Space has 1,200 employees and raised $1.2B
“Relativity is now a 1200 person company, raised about 1.2 billion in private capital it's headquartered down in Long Beach here in Los Angeles.”
Jordan Noone Jul 26, 2024 ▶ 24:30
Assertion Contradicted
Noone: There has not been an engineering software IPO since 1986
“There hasn't been an engineering software IPO since 1986. It's been 40 years of the same major incumbents.”
Jordan Noone Jul 26, 2024 ▶ 28:37
Assertion Not checkable as stated
Noone: Zoo users generated 13,000 CAD files on Text-to-CAD launch day
“The day we launched text-to-CAD in December of last year, we had 13,000 CAD files generated the first day of launching the platform.”
Jordan Noone Jul 26, 2024 ▶ 33:50
Disclosure
Noone: Zoo is building its first local GPU data center
“We're starting to build out our first local data center. And for our own GPUs, both for the geometric rendering side and the ML training side, we use those on both sides of the stack there.”
Jordan Noone Jul 26, 2024 ▶ 34:46
Insight
Noone: CNC toolpath generation is the costliest part of machine shop prototyping
“Generating a CNC toolpath is the expensive part of running a machine shop, especially something doing cutting edge hardware where you're prototyping, you're making one off runs. Most of that cost is going into engineering labor to generate the instructions for…”
Jordan Noone Jul 26, 2024 ▶ 39:34
Assertion Not checkable as stated
Noone: Programmatic Design for Manufacturability checks do not exist without human review
“There is no way to programmatically do a DFM check. You know, for a part, there's some software that kind of does it. You're still going to be sending it to your machinist. You're going to send it to the manufacturing engineer.”
Jordan Noone Jul 26, 2024 ▶ 42:04
Disclosure
Noone: Zoo will launch text-to-CNC toolpath generation within 18 months
“If it's text straight to the CNC machine, then that will be a capability that will come out in the next year, year and a half from us as a company is you can have your text prompt, send something to a CNC machine because you have that toolpath generated in pro…”
Jordan Noone Jul 26, 2024 ▶ 43:45
Disclosure
Zoo has raised about $6 million total in venture capital
“So historically we've raised only about six million in private venture capital.”
Jordan Noone Jul 26, 2024 ▶ 46:07
Disclosure
Former GitHub CEOs Tom Preston-Werner and Nat Friedman backed Zoo as angels
“Two of the angels in us are two of Jess's previous coworkers, which is Tom Preston Werner, the original GitHub founder, CEO, and then Nat Friedman, who took over the CEO for Tom after that.”
Jordan Noone Jul 26, 2024 ▶ 47:09
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 160 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.