Jacob Zhao

🚀 A milestone for Pantheon Research! 🌐 pantheon-research.com After almost one year of building, I'm pleased to share that Pantheon Research has been selected as one of 18 finalists in the Global FinTech Hackcelerator 2026! Organised by @gftn_official in partnership with @MAS_sg, the programme brings together FinTech startups, financial institutions, and investors at the @sgfintechfest. 🏦 Our Challenge | @juliusbaer "Reimagining Private Banking for Digitally Native HNW Clients" Pantheon combines systematic investment frameworks and evidence-grounded AI across Equities, Macro, Crypto, and FICC — making professional investment research more systematic, explainable, and accessible. Rules decide. AI explains. Humans make investment decisions. Looking forward to exploring how our technology can serve professional investors and the next generation of wealth management. 📍 See you at @sgfintechfest! 📅 November 18, 2026 | 2–5 PM SGT 🎤 Catalyst Stage, Singapore #GFH2026 #SFF2026 #WealthTech #AI

9/ Scaling doesn't mean quitting Vercel, but unbundling the stack: teams BUILD on @vercel for speed, SCALE by offloading backends to PaaS, and OWN by migrating core compute to hyperscalers.

8/ @Cloudflare attacks from the flank on cost, while @awscloud locks in the infrastructure floor and enterprise ceiling. In a fragmented stack where no single player owns everything, credible alternatives alone are enough to squeeze @vercel margins.

7/ @vercel faces asymmetric competition: @Cloudflare offers a cheap alternative, @awscloud acts as both partner and threat, while @supabase completes the stack. These viable choices cap Vercel’s pricing power and abstraction premium.

6/ Next.js and Git workflows form @vercel solid deployment core. Expanding into AI SDKs and Sandboxes broadens its scope, but commercial viability is unproven. Meanwhile, zero-margin token routing and AWS reselling monetize connections, leaving infra heavily partner-dependent.

5/ @vercel isn't for everyone. Its core sweet spot remains AI-native builders and pro developers who hold their own codebase but deliberately outsource infrastructure complexity.

4/ Despite soaring deployment metrics, a massive gap remains between "distribution activity" and "durable revenue": public evidence is lacking on whether these ephemeral builds convert to surviving apps, paying users, or high-margin expansion.

3/ AI deployment decisions are splitting into two camps: all-in-one platforms (Vercel uses v0 as a wedge) and general Coding Agents (Claude Code, Cursor). Vercel won early mindshare via CLI/MCP integrations, but these tools are easily cloned by Cloudflare—raising default odds, not a moat.

2/ Vercel’s core is a "deployment abstraction"—stripping out complex cloud decisions to swap control for extreme speed and frictionless shipping. It perfectly serves the Vibe Coding wave: teams that want code without managing infra.

1/ Vercel is surging with 6M daily deployments (>50% triggered by AI agents) and $340M ARR. Yet at a $93B valuation, its abstraction advantage faces Cloudflare’s squeeze, while whether agentic deployment spikes can yield long-term margins remains unproven.