× close spc-demo-faire.md South Park Commons India Demo Faire — only actual, working, live demos. Max 5 minutes. No pitch videos. SPC is built for the "-1 to 0" stage, and their Demo Faire proved exactly why that phase is so volatile. When you build live, the chance of a mishap is always non-zero — but that's where you separate signal from noise.

Maya [7.7/10] — Voice AI for local Indian languages + Arabic. Latency felt too smooth for live — my read: fast model pattern, passively passing prompt to backend while still communicating with user. 1.5M Play Store downloads. Japan next as speech-first market. Brilliant technical trick to slash perceived latency.

Fest AI [7.4/10] — Emotionally intelligent AI for venting and social connection. The game isn't memory storage — it's abductive reasoning, implying things from what the user leaves unsaid. But the consumer space is flooded. Survival is purely a distribution play. Nail onboarding and manufacture viral loops immediately, or die slowly.

Armor [7.2/10] — Hedge-fund-grade deterministic quant agent for personal finances. Moving away from generic LLM wrappers is the right play. But SEBI regulations, RIA licensing, backtesting transparency — what's the Sharpe ratio? If everyone gets these models, edge disappears. Trust is the bottleneck, not tech.

Cent AI [6.7/10] — Preventive health and early detection. 1,500 scanned, 28% major findings, 2.3% asymptomatic critical diseases caught early. ₹25K lifetime genetic test targets HNIs. Noble cause, but is their proprietary protocol a strong enough moat against well-funded legacy health tech?

Suchama AI [6.4/10] — Context-aware intelligence for manufacturing SOPs. Scraping SAP data to manage dispatch. Useful for quick-commerce. But manufacturing is slow to adapt. Can they compete on onboarding ease and deliver immediate, undeniable ROI?

GameRamp [6.2/10] — AI replacing strategy consulting. Goal: kill "let's circle back" culture. Having seen consulting internals, I'm sceptical. Consulting relies on capital allocation decisions requiring high conviction. Top management won't outsource core competency to a black box.

SurgeGrowth [5.6/10] — Parallel agentic operations at scale. 100-150 agents in parallel = rapid loss of control. Better proposition: pitch unikernels over cloud infra to cut fixed costs. Too many competitors, too much funding-game dependence for a low score to feel unfair.

Pattern: founders who knew exactly one customer problem cold beat everyone who tried to be broad.

× close yc-startup-school.md Every founder story collapses to the same loop: talk to customers, obsess over the experience, keep one ruthless focus.

Meta-lessons YC kept repeating: Start with the customer, not competitors or trends. Find the most extreme version of what users want, then work backwards. If 30 users love you, you're ahead. 1,000 mildly-unhappy users is a trap. Distribution + retention beats hype. Predict the market; don't react to it.

Zepto — Aadit Palicha: Started during pandemic when grocery procurement was chaotic. The pivot to 10-minute delivery came from reframing: "Ignore competitors. Define the customer ideal." First dark store = KV's apartment. They were "one of the worst companies in the batch" — what helped was singular focus on 30-40 users and removing all noise. Thesis: Zepto isn't a consumer app company — it's a logistics + supply chain company. Every rupee saved in supply chain becomes lower prices or reinvestment into last-mile delivery.

Razorpay — Harshil Mathur: Started with "I just want to code," not a finance obsession. Made zero sales during YC because of regulations. Called every single customer when banks pulled out — even if they got scolded. Quote: "If you build responding to what happens in the market, you're already dead." They believed India is a long-term compounding market and bet early on UPI. When demonetization hit, Zomato signed up — inflection.

Meesho — Vidit Aatrey: First product shut down in 3 months because they spoke to businesses, not customers. Discovered shops selling via WhatsApp groups — built tooling used by 100K businesses. Identified power users: online-native sellers, resellers/dropshippers. Launched Meesho supply. For 10 months: zero marketing spend, massive growth → PMF. COVID reset everything — they had to kill the existing business and restart.

Groww — Lalit Keshre: Started as robo-advisor. Customer question they kept hearing: "Why this product vs that one?" → selection + transparency matter. First month: 600 customers. Doubled down on what customers loved. "Launch litmus test: if users love it or hate it → both are okay. If users are indifferent → you have a problem."

YC observations: Success is about being in front of the wave. Build at the edge of technology. Clarity is the #1 application edge. YC invests in founders, not ideas. Rate of learning matters. The founder qualities YC likes haven't changed much over time.