Alexandre Berriche

Most YC founders don’t have traction. They have a friends-and-family pipeline. The pattern is brutal: • 10–15 warm intros from your network • A handful of pilots (sometimes free) • Deck updated: “early traction achieved” That’s not traction. That’s momentum theatre. The problem? YC optimises you for fundraising, not for selling. What you don’t get in the accelerator pack: • A systematic outbound motion • Qualification filters sharper than “anyone with budget” • Pipeline math (coverage, velocity, CAC sanity checks) • Cadence for running a weekly GTM review So the day after Demo Day, the clock starts. Your warm intros dry up. Investors ask about repeatability. And you’re staring at an empty CRM. Fundraising extends runway. Systems extend survival. YC24/25 founders - when the warm intros end, what’s your GTM scaffolding?

9

The real difference between San Francisco and Paris isn’t the vibe. It’s the intersection between money and « no legacy ». In Paris, the ecosystem splits in two. You have young startups with no legacy but little money. And you have established companies with money but too much legacy. The overlap between both is small. That’s why innovation feels slower. Young companies can’t afford to take big risks. Big companies can’t afford to move fast. In San Francisco and the Bay Area those two circles overlap way more Young companies raise fast, have cash, and no history holding them back. They experiment, iterate, and fail openly. Every day, new startups are born, others raise, others die. But even the ones that fail feed the system through paying people, testing new tools, and ideas. At the ecosystem scale, nothing is really lost. It all compounds.

336

31 Comments

The real question is not "is SaaS dead?" It is which #SaaS companies will #thrive in this environment? A great piece of reflection from our partners Gaetano Zanon and Julien Pache, with a helpful framework to detect the most potential winners. - The debate is misdirected. "SaaS is dead" is a "religious" statement. The real disruption is structural: pricing models, moat sources, and distribution advantages are being repriced simultaneously. - Horizontal point solutions with low switching costs and no proprietary data are the most exposed. They are essentially interfaces, already abstracted away. - Vertical SaaS with deep workflow integration, regulatory complexity, or domain-specific data might be the resilient ones, where switching costs are real. - Outcome-based and consumption pricing is the inevitable consequence of agents doing work that humans used to do per seat. https://lnkd.in/eVMc9_xf

5

We just published our Call for Startups for H1 2026, covering the structural shifts and market opportunities we're tracking most closely. These themes reflect where we're allocating focus and resources over the next six months: ▪️ Compute Starts Acting Like a Commodity – Financial infrastructure for enterprises to hedge GPU capacity and manage compute risk like airlines hedge fuel. ▪️ The Edge Stack Diverges from Cloud – Memory, interoperability, and developer tools purpose built for edge AI deployments. ▪️ AI Infrastructure Shifts from Training to Orchestration – The software layer between models and real work, focusing on governance, cost optimization, and agent operations. ▪️ The Winners of the Infrastructure Supercycle Will Build for Efficiency – Data center innovations around power, water, construction speed, and networking as constraints tighten. ▪️ Energy Continues to be a Bottleneck for AI – Novel power architectures, grid participation, and energy storage solutions. ▪️ Critical Materials As a Constraint to Industrial Power – AI driven modernization across the mining lifecycle from exploration to processing. ▪️ The Legacy Factory Modernization Challenge – Retrofitting legacy manufacturing or building AI native facilities, including robotics and labor force multipliers. ▪️ Full Stack AI Disrupts Professional Services – Outcome based service firms in categories where software historically failed due to workflow fragmentation. ▪️ Infrastructure emerges for managing human-AI hybrid workforces – Platforms for routing work between humans and AI, performance management, and new employment models. ▪️ Acute Care Moves Into the Home – Infrastructure for delivering hospital level care at home via telehealth. ▪️ Medical Malpractice Litigation and Insurance Changes in the AI Era – New underwriting models and prevention layers as AI permeates healthcare. If you're building in any of these areas, we'd love to hear from you. Link to full in comments.

40

6 Comments

ChatGPT5 prompt: Is Stripe's Open Issuance launch a eGolf Village stablecoin a viable alternative to the plan to issue a utility coin for the equity conversion Loyalty Rewards Program Response; Yes — using Stripe’s Open Issuance to issue a stablecoin (or “branded digital dollar”) could be a viable alternative to issuing a more traditional “utility coin” for an equity-conversion-based loyalty rewards program (like your proposed eGolfVillage stablecoin). But there are important trade-offs, risks, and architectural, legal, and business considerations you’d need to work through. Below is an analysis of pros, cons, and key issues to validate viability. link to full response: https://lnkd.in/g75Kajqx

1

The Frontier Tax Everyone repeats the same line: AI keeps getting cheaper. Last week the best model on the market got twice as expensive. Anthropic shipped Claude Fable 5 at $50 per million output tokens — double the price of the model it replaced. At the very top of the market, capability went up and so did the bill. Meanwhile the floor is falling out. DeepSeek, Qwen, Mistral and Llama have pushed open-weight models good enough for most enterprise work to nearly free. Two curves, moving in opposite directions. This is the part executives keep getting wrong. They treat "AI cost" as one number trending down, and they budget like the whole stack is on a discount. It isn't. The commodity middle is collapsing toward zero. The frontier — the small slice of work that actually needs the best model — is getting more expensive, not less. The mistake is running everything on one curve. Either you overpay by sending routine work to a frontier model, or you underperform by forcing your hardest problems onto a commodity one. The skill in 2026 is knowing which work belongs on which curve. Map your workload. Most of it runs fine on cheap, open, self-hosted models. A few high-stakes tasks justify the frontier premium. Price them separately, architect them separately, and stop pretending it is one line on a chart.  It is a race. Time to catch up.  #AI #EnterpriseAI #AITransformation #AIRace

3