Jul 19, 2026 · 41m · sourcery
AMD, Starcloud, Coatue..10 Hot Takes From The Biggest Names in AI · Sourcery with Molly O'Shea
Max Cook · 9m spoken Molly O'Shea · 8m spoken Ramin Hasani · 5m spoken Philip Johnston · 2m spoken Matan Grinberg · 2m spoken Rodrigo Liang · 1m spoken Robin Rombach · 1m spoken Mark Papermaster · 1m spoken Pim de Witte · 32s 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 recorded at the RAISE Summit in Paris, host Molly O'Shea interviews leading AI founders, investors, and executives to explore hot takes on open-source models, agentic workflows, efficient compute hardware, and the future of enterprise automation.
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 24.6% of the talking time here. How this is scored →
Molly as informed peer 2.0 Guest teaching 4.6 Guest disagreement 1.6 Molly pushing back 1.0
speaking balance: gold is Molly, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 13:50 Glorified data science take
Pim dismisses broader AI hype by bluntly calling most AI glorified data science downstream of simple datasets.
Hardest push from Molly ▶ 14:09 Pushing on sponsor plugMolly playfully steers the conversation into a sponsor plug for MongoDB when Pim refuses to disclose where data is stored.
Biggest teaching moment ▶ 34:56 Ramin on multidimensional model scalingRamin educates the audience on why single-axis pre-training scaling has hit economic and physical barriers, presenting Liquid AI's three-axis approach.
Molly holds their own ▶ 9:22 Molly cites Factory enterprise metricsMolly demonstrates domain awareness by citing enterprise open source adoption statistics shared by Matan from Factory.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Molly as informed peer | Guest teaching | Guest disagreement | Molly pushing back | Why |
|---|---|---|---|---|---|---|
| Max Cook on Interaction Models and Open Source | 3 | 5 | 2 | 1 | Max Cook delivers deep analyses of human-computer interaction models and open source versus closed-source economics. Molly primarily facilitates and prompts, occasionally citing data points she heard from other guests like Matan of Factory. | |
| Henri on the RAISE Summit Concept | 1 | 4 | 3 | 1 | Henri and Pim provide quick hot takes in a lighthearted, informal vibe. Pim dismisses the AI hype by arguing most AI is just glorified data science. | |
| Mark Papermaster on Agentic Workflows and Hardware | 2 | 4 | 1 | 1 | Mark Papermaster discusses agentic workflows and CPU/GPU hardware demands, followed by Robin Rombach outlining open weights philosophy. Molly sticks to introductory, open-ended questions. | |
| Sponsor Segment: Brex | 2 | 5 | 2 | 1 | Following an ad read, Matan Grinberg explains enterprise open source token growth metrics. Molly listens and prompts for year-end predictions. | |
| Philip Johnston on Space Data Centers and SpaceX | 2 | 4 | 1 | 1 | Philip Johnston outlines StarCloud's vision for space data centers and why SpaceX's launch dominance makes it historically undervalued. Molly asks friendly clarifying questions. | |
| Chris of Good Future Media on Short-Form Video Strategy | 2 | 5 | 1 | 1 | Chris breaks down the economics and psychological tactics behind short-form clipping and hooks. Molly asks for practical tips and pricing details. | |
| Sponsor Segment: MongoDB | 2 | 6 | 2 | 1 | Ramin Hasani delivers an elaborate technical monologue on shifting from one-dimensional capability scaling to a three-axis optimization model (intelligence, cost/efficiency, substrate). Molly asks brief prompts. | |
| Rodrigo Liang on SambaNova's Series F and SM-50 Chip | 2 | 4 | 1 | 1 | Rodrigo Liang reveals SambaNova's Series F funding and the launch of their SM-50 inference chip, explaining premium inference economics. Molly keeps the interview celebratory and concise. |
Statements from this episode (15)
Cook: Mobile apps will consolidate to 2-4 agentic AI interfaces
“We're going from a world in which you have 60 to 80 apps to an agentic AI world where you've got a lot fewer than that. Two, three, four apps where everything else plugs in as an API or an MCP.”
Cook: Keyboards and mice will be obsolete within five years
“We're gonna look back in five years and say, of course we're not using keyboards and mice anymore. Like, we've come up with a better way to interact with intelligence.”
Cook: Enterprise open-source LLM spend dropped from 19% to 11%
“Enterprise LLM spend was 19% open source last year, and it's 11% this year. So it's dropped.”
Cook: Open source rules production, while closed models dominate discovery
“Jesse's explanation, I think, is the correct one, which is that open source is good for production grade, like production scale workflows. Yet we're so early in this AI journey that there's not that many production grade workflows. And so if you are trying to …”
Pim de Witte: Most AI development is just glorified data science
“Most of AI is just glorified data science.”
Papermaster: Agentic workflows drive CPU-to-GPU ratios to one-to-one
“Now with these agentic workflows, you actually need both. In fact, the ratio of CPU to GPU is becoming like one to one.”
Papermaster: AMD uses thousands of AI sub-agents for chip design
“We have thousands of sub-agents just doing incredibly complex chip design, speeding how fast we can get the next generation to market.”
Rombach: Black Forest Labs builds multimodal models for physical AI
“We are building multimodal visual models for content creation and for now physical AI.”
Grinberg: 90% of AI tokens will use open models within 12 months
“My hottest take is that in the next 12 months, 90% of tokens will be going to open models.”
Grinberg: Enterprise open model token share surged past 10% by May
“At the beginning of the year, we saw enterprises were using less than one percent open models, like less than one percent of their tokens were going to open models. Then in around March, it kind of crossed that one percent threshold. And then in May, it crosse…”
Johnston: SpaceX will reach a $10 trillion valuation within years
“My hottest take right now is that the SpaceX IPO is the, gonna be viewed historically as the most undervalued IPO of all time. I think they will tear through 10 trillion within, you know, certainly within the next couple of years, and then have almost unlimite…”
Johnston: StarCloud secretly flew three Arm GPUs on its first spacecraft
“We flew an ARM chip on our first spacecraft. We flew three ARM GPUs on our first spacecraft. Nobody knows about that, but we did.”
Johnston: StarCloud 3 targets 10 megawatts of compute per Starship launch
“We are about to move into a much, much larger facility. We're building a kind of campus to set up a huge manufacturing line for what we're calling StarCloud three, which is, like, the vehicle which will fit on Starship, so we can potentially fit up to 10 megaw…”