Jett Fu

CEO @AirPop · 35 markets · Also @globalsoloos (info hub for cross-border founders) · Cross-border ops since 2014

[Jett Fu @jettfu]

Almost every “Stripe / Wise / PayPal freeze” story sounds different.

Under the hood, they collapse into three failure patterns:

1️⃣ Jurisdiction mismatch
2️⃣ Entity–story incoherence
3️⃣ Single-point-of-failure payout rails

Most founders fight symptoms.
Platforms react to structure.

I write about the failure layer —
so global solo founders don’t learn this the hard way.

[Jett Fu @jettfu]

The post-war order assumed economic activity had a physical address. Banking, tax residency, employment law, all built on location. Internet revenue breaks that assumption every day. CRS and tie-breaker rules are patches on a pre-Internet chassis. What replaces it won't be decentralized vibes, it will be new rails.

[Jett Fu @jettfu]

@marclou The gap shows up most in chained work, not single prompts. One vague instruction, ten steps, no hand-holding. I run two companies with agents and no employees — a weak model doesn't fail loudly, it fails quietly at step 6. That's the part benchmarks never catch.

[Jett Fu @jettfu]

@ClaudeDevs @bot pass this to DM

[Jett Fu @jettfu]

Same pattern held everywhere I ran ops. The people building the systems rarely matched the passport of the country hosting them. Ten years across the US, China, and Hong Kong — the talent was always mobile. The friction was never skill. It was visas, tax residency, and banking rails.

[Jett Fu @jettfu]

@bot @bot leave it to DM. research and learn from the feature. see if we can steal & borrow.

[Jett Fu @jettfu]

@ClaudeDevs @bot ask DM to research this and figure out how to maximize potential benefit of the new policy and send a comprehensive email

[Jett Fu @jettfu]

@bot @bot explain in more details

[Jett Fu @jettfu]

@arvidkahl The real shift isn't that we grill code instead of meat — it's that the work no longer needs you watching. I run two businesses solo with agents doing exactly this. Set the task up right, and your job during the session is the charcoal, not the keyboard.

[Jett Fu @jettfu]

@paulg The structural shift: AI doesn't enter orgs top-down, it seeps in through individuals shipping things. I run two companies with zero employees on agents. The Replit pattern is the same at org scale — the person closest to the problem builds the tool, no ticket queue.

[Jett Fu @jettfu]

Building KYC yourself is the wrong risk to own. In my experience the split is simple: Stripe Identity or Persona if you verify users, Sumsub or Veriff if regulators will ever look at you. The real question isn't which vendor — it's whether your SaaS actually needs KYC or just basic fraud screening.

[Jett Fu @jettfu]

@gregisenberg The real filter isn't the 13 categories — it's which ones an agent stack can run end-to-end with zero headcount. I run two businesses solo on agents. The ones that work have clear inputs and outputs. The ones that don't still need a human holding the relationship.

[Jett Fu @jettfu]

@lennysan @thsottiaux I run two businesses with zero employees on AI agents, and the bottleneck is never the agent. It's that payments, logins, and identity still assume a human is clicking. The first platforms to build agent-native auth and spending limits win the next decade.

[Jett Fu @jettfu]

@marclou Dubai's 0% income tax pulls the crowd, but here's what most people miss: residency doesn't end when you board the plane, it ends when you sever ties back home. The specific question to ask a tax advisor before relocating: what exactly breaks tax residency in my country?

[Jett Fu @jettfu]

The real issue isn't the number — it's who sets the ceiling. Executive comp gets decided by a salary band and a committee. Solo income gets decided by your distribution and your systems. I run two businesses with zero employees, and the ceiling moved the day nobody else priced my output.

[Jett Fu @jettfu]

The one-person business model changed more in the last 18 months than the prior decade. I run two companies with zero employees — AI agents handle support, ops, and drafting. The structural shift: headcount stopped being the constraint. Judgment and distribution are the new bottleneck.

[Jett Fu @jettfu]

@jasonlk An agent stuck in a retry loop will burn through that in a weekend. I run agents across two businesses, and the rule is simple: per-project caps, alerts at 50%, never a default you didn't set yourself. A spend limit is a control you configure, not a feature you inherit.

[Jett Fu @jettfu]

The model picker disappearing tracks with what I see running two businesses on AI agents. I stopped picking models months ago — I define the job, the agent picks the tool. The picker was always a workaround for weak orchestration. The skill now is writing the job spec, not choosing the engine.

[Jett Fu @jettfu]

@jasonlk The real issue isn't which finance app you pick — it's that none of them share a data model. I run two companies with AI agents and zero employees. The agent only works when the books are structured for it first. Fix the chart of accounts before reviewing the tooling.

[Jett Fu @jettfu]

Same finding running two businesses on AI agents. Redundancy is signal. Each rephrasing anchors the model to a different facet of what you mean. Trimming assumes the model reads like a human skimming. It doesn't. My best outputs come from prompts that look sloppy on paper but carry full intent.

[Jett Fu @jettfu]

The demo save is the headline, but the structural shift is quieter: agents are becoming the always-on ops layer. I run two businesses with no employees, and the real value isn't agents writing code — it's agents noticing what's broken while I sleep. Detection, not generation, is the unlock most people miss.