Nova wins the Deel global pitch competition
Deel, the payroll and global workforce platform now doing over $1 billion ARR, ran a startup competition across seven regions this year: New York, Berlin, London, Tel Aviv, Dubai, and more. 35,000 startups applied globally, roughly 3,500 in the Dubai region alone. Nova was among the 40 finalists, placing in the top 1% of applicants.
Of those 40, only 10 won. Nova was one of them.
Ramón then traveled to Paris for the global finals, where five companies each took home a $1 million prize. The competition there was a different level entirely: one of the five winners was already tracking toward $100M ARR by end of year. The field was also mixed in terms of stage, with pre-seed companies competing alongside Series C firms, which made for an interesting but imperfect comparison.
What stood out from the event was the depth of innovation across categories: medical devices analyzing cardiac and brain data, cancer detection from speech, deep tech infrastructure plays. AI agents were everywhere and across every vertical. Healthcare and infrastructure were the areas that made the strongest impression.
It's a good humbling experience. In the day-to-day we are not exposed to that level of builders from all over the world. Seeing them all in one place was very cool.
Ramón Rodrigáñez, Co-founder, Nova Talent
Karpathy joins Anthropic. What it means.
Andrej Karpathy, one of OpenAI's original co-founders and the person who coined the term vibe coding, has joined Anthropic. He joins a growing list of exceptional AI engineers who have left OpenAI adjacent orbits and landed at Anthropic: Dario Amodei himself is an OpenAI alumnus, and in recent months the former CTOs of Workday, Instagram and Box have all taken individual contributor research roles there.
The pattern is striking enough that Andrea and Ramón draw the comparison explicitly: right now, joining Anthropic as an AI researcher is what joining Real Madrid was during the Galácticos era. The best players in the world want to be on the same team.
The Anthropic talent pull
- Dario Amodei: formerly VP of Research at OpenAI
- Andrej Karpathy: OpenAI co-founder, coined "vibe coding", joins as researcher
- Former CTOs of Workday, Instagram and Box: took IC research roles
- Pattern: top talent choosing research impact over leadership titles
The compensation involved is likely extraordinary. Pre-IPO Anthropic equity at current valuations makes stock option packages worth tens of millions at minimum. But the hosts are clear that money alone does not explain the pattern. Anthropic has built a brand among serious engineers that OpenAI has steadily eroded through lawsuits, governance controversies and what increasingly looks like a consumer-first strategic pivot.
For developers and researchers, the signal is clear: Anthropic is where the serious work is happening.
Lovable vs Replit: two very different paths to $300M+ ARR
Two companies, both enabling people to build software without deep coding expertise, both now at remarkable scale. Their stories could not be more different.
Lovable
Lovable was built by a former engineer from Sana Labs who left to pursue a single idea. Three years later, the company has $400M ARR, 150 employees, and revenue per head approaching $3M. Growth has not slowed either: it went from $300M to $400M ARR in just a couple of months.
$400M ARR, reached in roughly 3 years
150 employees, ~$3M revenue per head
8M users on the platform
The product is deliberately aimed at solo founders and small teams: build an MVP, get your URL, launch. It is more accessible than Claude Code for non-technical users, and the company has continued adding features like multiplayer collaboration and, most recently, built-in GEO and SEO tooling so products rank well in AI search from day one.
Replit
Replit's story is almost the opposite of Lovable's in structure, though it lands in a similar place. The company was started in 2015, originally as a Chrome extension for learning to code on low-bandwidth connections, particularly for students in India. For most of its first decade, the company had massive user numbers but almost no revenue.
2015
Founded as a browser-based coding tool for low-connectivity users, particularly students
2023
22 million users but only $2.5M ARR. 7 months of runway remaining. Pivot decision made.
2024
Pivot to agentic coding complete. Former student users now adult developers, ready to pay.
2025
$250M ARR. Raised latest round at $9 billion valuation. Less than 2 years after the pivot.
The pivot worked because the timing aligned with something Replit had accidentally built: a decade of young users who grew up on the platform and were now ready to pay for a serious agentic coding tool. The user base did not need to be rebuilt. It just needed the right product to unlock it.
They had 7 months of runway when they pivoted. Less than 2 years later, they raised at $9 billion. That's the Replit story.
Ramón Rodrigáñez
PLG vs enterprise: when to choose which
The Lovable and Replit stories both point to the same conclusion: if you can crack product-led growth, there is nothing more scalable. No sales team needed. No long procurement cycles. No pilot agreements. The product grows itself.
Product-led growth
- Self-serve, value is immediate
- Scales without headcount
- Fastest path to large ARR
- Requires very strong product intuition
- Works best with wide addressable markets
Enterprise sales
- Higher ACV per customer
- Longer cycles, pilot requirements
- Stronger defensibility once embedded
- Requires a sales and CS motion
- Grows more slowly but predictably
The comparison that lands hardest is Palantir. Peter Thiel's company is one of the great enterprise software success stories. But the timeline to get to meaningful revenue was measured in decades, not years. PLG companies like Lovable compressed that into months.
For Nova Recruiter, the reflection is live and practical. The company built its initial traction through account executive led enterprise sales because the product was not self-serve ready. Now it is. The question is how to run both motions simultaneously without the enterprise legacy slowing down the PLG flywheel. Most companies that reach scale eventually run both, but the thinking, the team, and the metrics are genuinely different.
If you tell me you're able to grow to 10 million ARR in a very short time, most likely it's because we did great on Nova Recruiter product-led growth rather than any other business we have.
Andrea Marino, Co-founder, Nova Talent
The quiet death of middle management
The Karpathy story and the broader Anthropic talent pull feed into a structural point that comes up throughout this episode. Karpathy, like the former CTOs of Workday, Instagram and Box, took an individual contributor role rather than a leadership one. The title went down. The impact went up.
This is not a coincidence. It reflects something changing in how the most capable people in technology think about career growth. The traditional model was clear: build technical skills, move into management, lead bigger and bigger teams. Success was measured in headcount under your name.
AI is putting pressure on that model from two directions at once. Fewer people are needed to accomplish the same output, which means fewer management layers are needed above them. And the most valuable thing a senior person can do is often to contribute directly rather than to coordinate others who contribute.
A lot of us have grown under the idea that professional growth means leading bigger and bigger teams. AI is putting a lot of pressure on that. The best companies are now looking for high-caliber ICs who can do a lot on their own.
Ramón Rodrigáñez
The implication for anyone building a career right now is worth sitting with. The question is not how many people you manage. It is how much you can deliver and how clearly that impact is attributable to you directly.
Cerebras IPO: $66B valuation, 100x ARR multiple
Cerebras, the chip company building an alternative to Nvidia's GPU architecture, completed its IPO. The numbers were striking: shares priced at $185, rose to over $380, and settled around $300. The company did $500M in ARR last year and is now valued at $66 billion, roughly 100 times last year's revenue.
The multiple is extreme even by AI infrastructure standards. But the market conditions for chip companies are exceptional right now. Cerebras' technical edge is specific: their chips co-locate compute and memory on the same silicon, which makes them faster than Nvidia's architecture for training very large models where data transfer between separate compute and memory chips becomes a bottleneck.
Nvidia is also one of Cerebras' major investors. Which raises an interesting parallel for Nova: are you a challenger to the dominant player in your category, or are you actually a complement they would want to have a stake in? The answer shapes everything from positioning to fundraising to long-term strategy.
Takeaways from this episode
- Nova won the Deel global pitch competition in Dubai, placing in the top 10 out of 3,500 regional applicants
- Karpathy joining Anthropic confirms a pattern: the best AI talent is gravitating toward Anthropic the way elite footballers once gravitated toward Real Madrid's Galácticos era
- Lovable crossed $400M ARR in 3 years with 150 people. Replit went from 7 months of runway to a $9B valuation in under 2 years. Both are PLG stories.
- If you can crack product-led growth, nothing scales faster. Enterprise is more defensible but slower. Most companies eventually need both, but the motions are fundamentally different.
- The traditional career model of growing into management of larger teams is under real pressure. High-impact individual contributors who can do a lot autonomously are increasingly what the best companies are optimizing for.
- Cerebras IPO'd at 100x ARR. The chip layer remains the most defensible and most valued part of the AI stack.
- Moat in AI tools increasingly comes from proprietary data and community, not from the model layer. Anyone can buy the same public data sources. Unique data compounds over time.
Nova Membership · Applications open
The network where the most ambitious people in the world connect.
Nova connects over 25,000 top professionals across 80+ countries through a rigorous merit-based selection process. If you want to be around the people building the next chapter of tech, talent and business, this is where to start.
