🍍 Gabriel Jarrosson

Lobster Capital

Managing Partner

Leonis Investissement

Investing in YC startups. $36M deployed since 2020.

Chairman

Founder & CEO

Piña Colada

US stocks recommendation newsletter
Company sold in 2023 (acquired by Agora France)

Founder & CEO [acquired by Agora France]

albert académie

Online e-learning platform. 
Company sold in 2023 (acquired by P2KR)

Co-Founder & CEO [acquired by P2KR]

Le Vin de France

Marketplace for french wine growers.
Company sold in 2016 (acquired by Les Grappes).

Founder & CEO [acquired by Les Grappes]

🔩 Computer Science Engineer
💪🏻 7x founder with 3 exits (4 failures, learned some stuff). Bootstrapped to $1M ARR with 0 employees. 
🦞 VC @ Lobster Capital: Funding the best YC startups to change the world.

How does a YC startup get over $9M ARR in 3 months? 

Everyone thought “friendship apps” were a dead end. Then RealRoots walked into YC and showed $9.4M ARR across 80+ cities.

Dorothy Li explains how they cracked it: 
• Manual-first → AI product design 
• Shifting stigma (friendship today = online dating 10 years ago) 
• Cold DMs that converted into real, paying customers

It’s a masterclass in how consumer is coming back at YC.

🔗 ⬇️  Link in the comments below

Gabriel Jarrosson

VC @ Lobster Capital 🦞🧢 Investing in the top 2% of YC startups | Founder with 3 exits

This Demo Day, I backed a missile defense company and autonomous excavators.
What changed wasn't my thesis - but as Lobster Capital has grown, we have seen 1000s of hardware businesses and can see now the patterns for how they work best.

    -- Software companies live or die by customer acquisition.

You can build the perfect product, but if you can't get users, you fail. It's a game of distribution, 
marketing, and growth loops.

    -- Hardware and deep tech flip this equation.

If you can actually build what you claim to build, the customers are often already waiting.

Take Perseus Defense… Today it costs $100,000 to shoot down a $500 drone. Their solution costs $2,500 to manufacture and works just as well.

The Pentagon's response when they saw the demo = When can you deliver and how many…

Flywheel AI is retrofitting existing machinery for remote operation.

If you can eliminate that risk while making operators more productive, every construction company will want your solution.

The software/hardware difference = "Can you find customers?" vs "Can you actually build it?"

That shifts due diligence from market analysis to team evaluation.

Are these the most credible people to execute this vision?

Do they have the technical chops and operational excellence to deliver at scale?

Perseus drove to the desert every two weeks to test prototypes… They 3D print modifications in their tent between Saturday and Sunday test runs. Four months from zero to working prototype.

When the technical risk is high but the market demand is obvious, the entire investment equation changes.

I had a call scheduled with a hot YC company for Monday morning at 10am - the first official day of their fundraising process.

By lunch, their round was full and oversubscribed.

They'd only taken meetings Saturday, Sunday, and Monday morning. Maybe 15 total conversations. The round closed in 60 hours.

For some companies there's no time for multiple touchpoints or extended due diligence. If you need three meetings to make a decision, you're not getting allocation.

I've seen companies where the first 12 investor meetings were all yeses. When demand outstrips supply by that margin, traditional investment processes break down.

YC trains founders to be fundraising machines. They know exactly how to present their metrics, tell their story, and create urgency. They're optimized for speed.

This creates adverse selection for investors who move slowly. If you're only investing in companies that give you three meetings, you're probably looking at companies that everyone else already passed on.

But this system also rewards pattern recognition over discovery. Investors develop shortcuts and heuristics to make quick decisions. The more obvious the opportunity, the faster it moves.

The companies that don't fit obvious patterns \- even if they're potentially better investments \- struggle to get attention in this environment.

Speed becomes a competitive advantage for both founders and investors.

When competing in this environment you need pattern recognition that screens for companies outside of the regular patterns that can still be very valuable.

YC’s next billion dollar bet is on construction.

On the latest Lobster Talks, I sat down with Jash Mota (Flywheel AI), who’s building Waymo for excavators. Their approach is simple but powerful: retrofit any machine in hours, enable remote tele-op today, and build the data flywheel that leads to full autonomy tomorrow.

Why this conversation matters for investors, operators, and founders:
- Labor shortages in construction are a multi-trillion-dollar bottleneck.
- Safety failures aren’t just tragic — they create massive financial penalties.
- Flywheel’s retrofit works across all machines, not just the 5% with drive-by-wire.
- Autonomy is a data game — whoever collects the most edge cases wins.
YC hardware founders can thrive by following the counterintuitive playbook: sell first, build last.

We also cover the wild story of bringing an excavator to YC Demo Day, and how six hours of data was enough to train their first autonomy model.

If you care about AI, autonomy, and YC’s deeptech wave, this episode is a must-watch
 
🔗 ⬇️ Link in the comments below

Most founders dream of raising VC. Few understand the risks.

Once you take institutional money, there are only 3 outcomes: shut down, get acquired, or IPO. The “profitable lifestyle business” path is gone.

Worse: the wrong VC can push you into a profitability death spiral — chasing vanity growth instead of building a real company.

In this video, I break down:

• The 3 outcomes of VC funding
• How bad investors push startups into death spirals
• The #1 test to run before you take money

🔗 ⬇️ in the comments

Garry Tan gave startups the best news they'll hear all year

AI scaling laws are slowing down..

Massive implications for startups.... For years, the narrative has been that AI will accelerate exponentially until it makes human entrepreneurs obsolete.

The reality is different. Scaling laws - the methods AI labs use to increase model capabilities - are hitting diminishing returns. The easy gains from throwing more compute at problems are tapering off.

When scaling laws were perfect, only companies with massive compute budgets could compete. OpenAI, Google, and Microsoft had structural advantages startups couldn't overcome.

But when raw scaling hits walls, other factors become important. 

-- Domain expertise.
-- Creative applications
-- Customer relationships.
-- Execution speed.

These are areas where startups excel.

This pattern repeats across technology cycles. When Moore's Law slowed, Intel's manufacturing advantages decreased while Apple thrived through design and software integration.

Garry sees this at YC. He's watching entrepreneurs work 80-hour weeks building companies that leverage AI rather than compete with it.

The slowdown doesn't mean AI stops improving... It means improvements come from creativity (rather than brute-force scaling).

Top companies are using existing models to solve specific problems faster than incumbents still optimizing for the previous paradigm.

Most unicorns were never unicorns.

They were billion-dollar valuations built on cheap money, not real traction.

Between 2020 and 2022, dozens of startups raised at 20× revenue, burned $10M/month, and called themselves unicorns.

The term “unicorn” has become broken.

But… Venture capital still depends on rare, fund-returning outcomes.

One big win can return an entire fund. That’s why VCs still chase unicorns… and always will.

In my latest video, I break it all down.
🔗 ⬇️ Link in the comments below

If you’re not building your startup in SF, you’ve already accepted mediocrity.

Every few months, someone declares it dead.

Then you land here, see founders pitching in cafés, YC partners on Market Street, and AI teams raising rounds before lunch.

Here, speed compounds because deals close in days.

In France, ideas wait for conviction.
In SF, conviction finds the ideas.

In this video, I explore the journey from Europe to the Bay, and how San Francisco’s speed, density, and access still make it the ultimate startup hub.

🔗 ⬇️ Link in the comments below

He was going so viral that Garry Tan invited him to YC. 

Stormy AI takes influencer marketing from weeks of spreadsheets and cold DMs to minutes: search creators → AI negotiates → deal closed.

On this episode of The Lobster Talks, Robert Lukoszko (Stormy AI) and I dive into:
• How viral Twitter demos landed him a YC invite
• Why influencer marketing is moving from manual outreach to AI-driven UGC
• What it takes to build an AI startup that compounds with every release

This is one of the cleanest examples I’ve seen of AI turning chaos into workflow. Worth a listen! 

🔗 ⬇️ in the comments

#YC #startups #AI #CreatorEconomy #LobsterTalks

Everyone mobbed the two companies with massive ARR numbers.

$9 million and $15 million in annual recurring revenue at Demo Day gets attention.

But… both companies already had their rounds closed. Everyone chasing them was fighting for scraps in an oversubscribed round at a higher valuation.

This doesn’t always = success.

For example - Motion - now worth $550 million - was unsexy and still raising at their 2020 Demo Day. They'd just pivoted and were still figuring things out.

The pattern repeats every batch. The companies that crush Demo Day often do it with pre-existing traction and closed rounds. The companies that seem "whatever" but have the right fundamentals can often go overlooked.

This creates a systematic bias. Investors flock to obvious winners who don't need more investment, while underestimating companies in earlier stages who might have better risk-adjusted returns.

The best Demo Day strategy isn't to chase the highest ARR numbers but looking into the companies with the strongest fundamentals that haven't yet captured widespread attention.

Klarna went public after 20+ years, raising $222M in primary and seeing $1.2B in secondary sales. The story violates conventional venture wisdom in ways that matter.

Fintech isn't winner-take-all… 

Every major category supports multiple winners: Klarna vs Affirm, Stripe vs Adyen, Ramp vs Brex vs Airwallex. It might actually pay to be second - you draft off the market education of leaders while avoiding their early mistakes.

Market saturation is a myth in fintech…

Klarna has 82% penetration in Sweden but only 10% in the US, despite operating there for years. This isn't a failure - but proof that identical products can have vastly different adoption curves based on cultural, regulatory, and competitive factors.

The same financial need (flexible payments) manifests differently across markets, creating room for localized solutions even when global players exist.

Two implications for investors:

1) look externally for fintech innovation. The US isn't always the source of breakthrough business models.

2) if Klarna represents consumer payments wired for the digital age, what does payments look like wired for the agentic age?

The payment flows, credit decisions, and fraud prevention that work for human commerce won't work for agent commerce.

The companies building payment infrastructure for AI agents might not come from Silicon Valley. They might emerge wherever regulatory frameworks and technical talent align first.

Klarna's 20-year journey reminds us that the biggest opportunities might take longer and come from more unexpected places than venture wisdom suggests.

Everyone says they’re building “AI-native enterprise software.”

But what does AI-native actually mean?

It’s not Salesforce with a chatbot bolted on.

That just gives you the same screens and dashboards with a new interface. AI-native means rethinking what work looks like when you interact less with keyboards and screens in the first place.

Enterprise software today is endless dropdowns, forms, and dashboards. Whole teams exist just to shuttle information between systems. “Using” the software is the job.

An AI-native system flips that. Instead of clicking through ten tabs, you describe the outcome you want - or the system surfaces it before you even ask. Reporting, CRM updates, approvals, reconciliations… they don’t live behind a login page.

They happen auto or humans step in only when judgment is required.

That’s the wedge.

Incumbents like Salesforce, Workday, Oracle still have deep moats - distribution, compliance, relationships and steak dinners.

They won’t be out dashboarded!

But they could be beaten by companies that change the category itself… less time in the system, less screen, less keyboard.

That’s the question founders should be asking. Not “how do we add AI to enterprise software,” but what does enterprise software look like when people barely touch it at all?

I met multiple founders at Demo Day who went straight from high school to SaaS.
This is pretty interesting, going straight from Roblox/ Minecraft server dollars into startups…

** One 18-year-old has 30% week-over-week growth with enterprise customers.
** Another freshman dropout closed her round in 48 hours.
** Another was summarising research papers at 14 and got Lucy Guo to fund it.

Generation Alpha is skipping that ‘life discovery’ phase and building $100K+ ARR businesses as teenagers.

Is it that APIs, cloud services, and AI models democratize what used to require years of technical expertise or that distribution channels that took decades to build are now accessible through social media and developer communities?

The traditional path was - learn to code, get a job, gain experience, then start a company.
Now it's - learn to prompt, find a problem, build a solution, sell it.

This is about access to leverage. When the tools to build and distribute software become this accessible, the constraints that forced previous generations through apprenticeship phases disappear.

Generation Alpha has A+ tools and virtually zero gatekeepers.

Teenagers are building businesses that would have required entire teams five years ago… 
Bullish on this? Why/Why not?

Taplio

Gabriel Jarrosson's Linkedin Analytics

Gabriel Jarrosson

Performance ScoreGabriel Jarrosson's LinkedIn Performance

What is Gabriel talking about?

Gabriel Jarrosson's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

Taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Gabriel Jarrosson's Linkedin Analytics

Gabriel Jarrosson

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Gabriel Jarrosson's LinkedIn Performance

What is Gabriel talking about?

Gabriel Jarrosson's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Gabriel Jarrosson's Linkedin Analytics

Performance ScoreGabriel Jarrosson's LinkedIn Performance

Gabriel Jarrosson's Best Posts (last 30 days)