Finscale in English

There are a handful of genuinely interesting ideas - orchestration lock-in as the real AI moat, architectural resilience as a preliminary filter before any financial analysis, and the clarification that agentic AI is still software - but a large portion of the 34 minutes is spent on biographical background, self-promotion of the newsletter, and vague macro claims about companies going 'to the moon or to zero'. Insight density is diluted by the conversational setup.

“agents are software. Like what is an agent? An agent is fundamentally a unit of software. So when we say software is at risk, what we're saying is software as a service, SaaS as a way of delivering SaaS is at risk.”

“the value goes to the companies that are able to orchestrate that intelligence and to use that intelligence to put it in a system and to actually produce tangible business outcomes”

The orchestration-as-moat thesis published a year ago (before it became consensus), the architectural resilience as a mandatory pre-filter, and the explicit pushback on the 'deterministic enterprise vs. probabilistic LLM' defensive framing are genuinely contrarian positions. The castle-helicopter metaphor is evocative and the argument that the AI labs will simply solve enterprise reliability is a meaningful challenge to popular defensive SaaS narratives. However, the broader macro framing ('paradigm shift,' 'moats,' 'winners and losers') is well-worn territory.

“with generated AI, the attacks are coming from the sky, the helicopter that no one had thought about, and they're coming from the foundations of the castle that are shaking from below, from above”

“I am very wary of a lot of defensive positions on um, software right now...LLMs are probabilistic. The enterprise needs deterministic software...I personally think that that is wrong”

Raphaëlle d'Ornano has genuine practitioner depth - 800+ diligence engagements, founder of an advisory firm built from scratch, now a hedge fund manager with a publicly tested investment thesis and verifiable named calls on specific stocks. However, she is a first-time fund manager without a long audited track record, and she straddles the practitioner/thought-leader line via her Substack, which limits the ceiling here.

“I completed over 800 diligence engagements across every single field of technology. Software, Internet, deep tech, et cetera.”

“I was maybe ahead on the SaaS apocalypse. So I was, uh, short on a lot of software names starting in November. And so obviously that played out very well”

There are concrete anchors - the 11% switching statistic, named companies (CoreWeave, Figma, Sigma, Fermi), an 85%/15% portfolio split, 800 engagements, the 280-page public manifesto - but critical claims are left unsubstantiated: the source for the 11% figure is not given, fund performance is described only as 'very superior to many other funds' with nothing disclosed, and the SaaS apocalypse argument rests on directional assertion rather than data.

“11% of enterprise builders switched AI providers last year, despite the switch, uh, being technically trivial”

“I have approximately uh, 85% of my portfolio which is in the US I have uh, many stocks in Japan and in Asia”

The host asks a few structurally good questions - why publish the thesis publicly, why hedge fund over VC, how did you raise without a track record - and these draw out real content. But the interview is compromised by extended biographical setup, active promotion of the guest's newsletter mid-interview, and a consistent failure to follow up on vague performance claims or challenge the macro thesis with any friction. The closing question on human impact is too philosophical to yield operational learning.

“Why did you decide to make it uh, public, uh, and to give access to everyone. It's very detailed, uh, very interesting. But I guess for your competitors it's also a source of inspiration.”

“you are ah, first fund manager. This is the first. You don't have credentials. Um, so how uh, were you able to raise”