Jul 19, 2026 · 56m · sourcery
SemiAnalysis, Altimeter, Nebius, Glean.. 12 Hot Takes From Biggest Names in AI · Sourcery with Molly O'Shea
Qasar Younis · 13m spoken Molly O'Shea · 9m spoken Dylan Patel · 5m spoken Arvind Jain · 2m spoken Apoorv Agrawal · 2m spoken Ariel Cohen · 2m 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
Recorded at the RAISE Summit in Paris, host Molly O'Shea interviews prominent AI founders, analysts, and tech executives to gather hot takes on enterprise model routing, hardware infrastructure scaling, data management, and the shifting economics of artificial intelligence.
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 17.7% of the talking time here. How this is scored →
Molly as informed peer 3.5 Guest teaching 3.8 Guest disagreement 2.1 Molly pushing back 1.2
speaking balance: gold is Molly, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 22:19 Dylan Patel condemns token budgeting
Dylan aggressively dismisses the practice of token budgeting as loser mentality, arguing that companies clamping down on token usage will fail to adapt.
Hardest push from Molly ▶ 23:12 Molly presses Dylan on firing an employeeMolly directly challenges Dylan on industry rumors about him firing an employee over using Haiku, forcing him to clarify the real workplace story.
Biggest teaching moment ▶ 17:48 Dylan details European power grid regulationsDylan explains why data center construction in France faces regulatory pushback from neighboring EU countries who rely on French nuclear energy to balance their grids.
Molly holds their own ▶ 1:33 Molly cites macro employment dataMolly validates and extends Qasar's commencement point by citing published economic data showing that AI has so far correlated with increased total jobs.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Molly as informed peer | Guest teaching | Guest disagreement | Molly pushing back | Why |
|---|---|---|---|---|---|---|
| Interview with Kasser from Applied Intuition | 4 | 4 | 1 | 1 | Molly engages in a supportive and conversational interview with Qasar, referencing recent job report statistics and Marc Andreessen's attendance at their conference. Qasar shares detailed analogies comparing modern AI infrastructure spend to the Long-Term Capital Management collapse and lessons from John DeLorean's General Motors memoir. | |
| Sponsor Segment: Brex Financial Platform | 5 | 5 | 6 | 3 | After an initial sponsor read, Dylan Patel brings sharp takes, dismissing token budgeting as a loser mentality and jokingly teasing Molly. Molly keeps up with technical terminology, prompting him on co-design, hardware pricing trickle-down, and rumors about firing employees over model selection. | |
| Interview with Mark from Nebius | 3 | 3 | 1 | 1 | Mark delivers corporate talking points on Nebius's multi-threaded data center portfolio strategy. Molly facilitates the overview of their infrastructure expansion without challenging the framing. | |
| Interview with Arvind Jain from Glean | 3 | 4 | 1 | 1 | Arvind explains Glean's context graph technology and token optimization benefits in the enterprise. He also predicts open-source models will take over almost all AI inferencing within two years. | |
| Interview with Laura from New York Stock Exchange | 2 | 1 | 0 | 0 | A purely collaborative and promotional segment between Molly and Laura from the New York Stock Exchange, focusing on on-the-ground vibes and media storytelling. | |
| Interview with Apoorv Agrawal from Altimeter | 4 | 5 | 2 | 2 | Apoorv lays out his framework of the four seasons in AI and advises enterprise leaders to plan for climate rather than weather. Molly follows up directly to clarify what constitutes the underlying climate. | |
| Interview with Nikhil from Turbo Puffer | 3 | 5 | 1 | 1 | Nikhil details the architectural mechanics of Turbo Puffer, breaking down how caching layers over object storage allow large-scale vector search at a fraction of standard costs. Molly guides the technical walkthrough. | |
| Sponsor Segment: MongoDB Database Platform | 3 | 4 | 2 | 1 | Following the MongoDB ad, Barak presents Wonderful AI's contrarian thesis that geographic expansion outpaces vertical specialization in AI deployment. Molly prompts him to elaborate on their playbook. | |
| Interview with Max from Legora | 4 | 4 | 4 | 1 | Max shares strong opinions criticizing European startup vacation culture and urging founders to adopt a global competitive work ethic. Molly engages on token economics and consumption pricing models in legal tech. | |
| Interview with Gil from Merge | 4 | 4 | 4 | 1 | Gil rejects the premise of token maxing, stating it produces zero tangible productivity gains, and stresses security concerns regarding enterprise LLM integration. Molly brings up Alex Karp's commentary on data sovereignty. | |
| Interview with Ariel Cohen from Navan | 3 | 3 | 1 | 1 | Ariel explains why hallucinations cannot be tolerated in enterprise travel systems and recaps Navan's revenue milestones. The tone remains fully collaborative. |
Statements from this episode (13)
Younis: Applied Intuition has barely touched its venture capital funding
“Yeah, we're still in that phase, which is great, which we're super happy about.”
Younis: Physical AI will outshine early LLMs over the next 25 years
“I think the physical world and AI going into that is, I think when we look back, I think that's going to be the bigger story. And then look back 25 years from now.”
Younis: AI Industry Is in an Opulent Bubble and a Pullback Is Coming
“The cost structure that we're currently seeing from whether it's from how much we spend on model training and development to employees to where we're at that phase of the bubble where it's like, you know, like the, all the opulence in the companies is what peo…”
Younis: Autonomous data collection vehicles are expanding to global cities
“When I go around Paris, you see data collection vehicles. And you're like, it's even coming here. Like, you know, it's just like not Sunnyvale. And so now, like every city I go to typically with a trained eye, you can see like, oh, that's a data collection. We…”
Patel: EU countries block French data centers to protect nuclear power
“Paris is you know, they have all this spare power. One of the funny things is like, they're like, oh, okay, let's build a bunch of data centers. But then Spain, Italy, Belgium the Netherlands and Germany All freaked out because it's like, oh wait, all this spa…”
Patel: Open-source AI is dying as Chinese labs shift to licensing
“The theme here is that all, like, there's multiple Chinese model labs who are telling all the inference guys, hey, yeah, yeah, our next model's not gonna be open source, we're gonna license it to you. So open is, is dying quickly, unfortunately.”
Patel: Nvidia B200 and B300 server pricing increased prior to production
“I think you've already seen server pricing go up like B 200, B 300 that you're buying now versus last year. The price is significantly higher. We've seen next generation hardware receive price increases before they've even started production. Like, here was th…”
Patel: Strict token budgeting is a 'loser mentality' that stalls productivity
“I love token maxing. I think anyone who doesn't token max is going to get left behind. I think all this token budgeting stuff is, is loser mentality. If you're token budgeting hardcore, then your people are not going to learn the new workflows and then they're…”
Jain: Open source models will dominate AI inferencing within two years
“Well, I think open source models are going to dominate AI inferencing. You know, you'll see in the next two years, we'll shift to where's almost no open source to where's going to be almost all open source.”
Agrawal: Enterprises must plan for multi-model routing over single-lab lock-in
“So if you're a CIO, if you're a CEO, if you're about to make a multi-million, billion-dollar decision of which lab to build your intelligence on, plan for the climate, not for the weather. What does that mean? That's multi-model routing. That's evals. That's t…”
Agrawal: Decagon routes roughly 90% of its production through open-source models
“As Jesse from Decagon wrote very eloquently a big portion of their volume goes through open source, about 90% now.”