Company of One: New Structure, Sweet Trap, and Vietnam's Special Position

Opening

The previous essay talked about how the software production chain is falling apart. One consequence of that disintegration — which the previous essay didn't address fully — is the emergence of a new kind of business unit: the AI-augmented company of one.

The concept of a "company of one" isn't new. Paul Jarvis wrote Company of One in 2019, developing the philosophy of "better instead of bigger" — keeping a company intentionally small in exchange for freedom and margin, not small because it failed to grow. But by 2024-2026, under the AI wave, this model has shifted from a lifestyle choice to a data-backed business model — and at the same time become the sweetest trap for people in tech.

Sam Altman publicly predicted in 2024 that AI would create "the first one-billion-dollar one-person company." As of May 2026, no company of one has been confirmed to reach 100 million USD ARR, let alone a billion. But there are notable numbers:

• Pieter Levels alone runs a portfolio of roughly 3 million USD ARR (Photo AI, Nomad List, Remote OK, Interior AI)
• Marc Lou — ShipFast crossed 100K USD/month at 91% margin
• Justin Welsh — 4.15 million USD in 2024 revenue, 86% margin, no W-2 employees apart from one part-time virtual assistant
• Tony Dinh (Saigon) — TypingMind crossed 1 million USD in trailing-12-month revenue in November 2024
• Daniel Nguyen (HCMC) — BoltAI at roughly 15-30K USD/month with 7,000+ customers

These are real, verifiable numbers. But this is also where 99% of solopreneur essays stop — and that's where it's most dangerous. Because these numbers are drawn from an extremely selected sample: those who survived and speak publicly. Behind every Pieter Levels are hundreds of indie hackers who shipped 5-20 products, made under $1,000 cumulative MRR, burned out, and went back to employment. They don't blog. They don't go on podcasts. They disappear.

This essay goes in four steps: why this model is more viable than ever (structure, not slogan), why most who try will fail (and that's the most important thing to say honestly), Vietnam's special position in the new configuration, and finally — a decision framework for those seriously considering it.

The tone is the same as the previous essays: reasoning to the end, no boilerplate optimism, no boilerplate pessimism. There's a "where this might be wrong" note at the end.

---

Part 1: Why 2026 is structurally different

Before we talk about who should try and who shouldn't, we need to clarify why this moment is different — not emotionally different, but different in the economic structure of one person making software.

Reasoning from first principles, a software company needs at least five kinds of capability to operate: understanding the customer's problem, designing the solution, writing code, operating the system, closing the feedback loop (the same five layers from the previous essay). In the traditional model, each capability needs at least one dedicated person. The minimum math: 5 people + 1 manager + 1 sales = 7 people. With an average cost of 100-150K USD/year in developed markets, the company needs at least 700K-1M USD in revenue to break even.

This is the minimum barrier of the traditional software business model. Not because of law, not because of culture — because of the cognitive capacity limit of a single human who can't do all five layers well.

What does AI change? It doesn't make one person superhuman. It makes an ordinary person able to access all five layers at good-enough quality. Not excellent — good enough for niche markets.

flowchart LR
    subgraph Old[Pre-2022 — Needs a team]
        direction TB
        O1[Layer 1: Product/BA]
        O2[Layer 2: Architect/Designer]
        O3[Layer 3: 3-5 Engineers]
        O4[Layer 4: DevOps/SRE]
        O5[Layer 5: Support/Marketing]
    end

    subgraph New[2024-2026 — One person + AI]
        direction TB
        N1[Operator<br/>1 person]
        N1 -.uses.-> AI[AI amplifies<br/>all 5 layers]
    end

    Old ==> New

More concretely, a solo operator in 2026 can run:

• Understand the problem: customer interviews → Claude synthesizes transcripts into a prioritized problem list
• Design: v0.dev or Lovable generates interfaces in minutes; Claude proposes architecture and API contracts
• Write code: Cursor + Claude Code does Layer 3 at 5-8x the productivity of a traditional engineer
• Operate: Vercel/Railway handles deploys; Claude analyzes logs; n8n or Lindy runs automated workflows
• Feedback: Intercom Fin or Plain answers 70% of tickets; Claude drafts emails; PostHog summarizes weekly metrics

Total stack cost: 80-120 USD/month. Compared to the minimum payroll of a team of 5-7 (in Vietnam: ~50,000-100,000 USD/year), that's a 95-98% cost reduction.

This isn't theory. This is exactly how Tony Dinh runs TypingMind. How Marc Lou runs ShipFast. How Pieter Levels runs Photo AI. They don't have teams. They have a tool stack.

An important warning before going further: most of the solopreneur data I'll cite in this essay — Pieter Levels' 70 failures, Justin Welsh writing for 6 years before revenue, Marc Lou's 23 projects in 2.5 years — are all pre-AI data, from the 2013-2022 period. These numbers describe the old game. The new game may have very different dynamics — faster try-fail loops, smaller niches that can still support one person, product experiments costing days instead of months.

But at the same time, there's a counterargument that can't be ignored: those who succeeded pre-AI have a double advantage in the post-AI world. They have a built audience (that AI can't shortcut), market judgment skills tempered by dozens of failures (that AI can't replace), and now AI amplifying them on top. Pieter Levels didn't need AI to succeed — he succeeded before GPT-3.5 even launched. Now with AI, he just gets richer faster. Those who already did well will do even better; those who never did well aren't automatically rescued by AI.

This is the spirit in which to read this essay: take the framework, don't rigidly fixate on numbers. When I say "18-24 months to replacement revenue", that's an estimate based on pre-AI people; it could shrink to 6-12 months for a good post-AI operator, or balloon to infinity for someone without a foundation. The variance band is much wider than historical data suggests.

And — this is where most blog posts stop without saying it — reducing payroll cost is not the same as creating revenue. They are two different problems. The next section is about the much harder one.

---

Part 2: Why most who try will fail

This is the part few people write. Because it doesn't sell courses.

Income distribution data for solopreneurs in the US — where public data is best:

This is not a bell curve. This is a power-law distribution — characteristic of markets that anyone can enter, but winners take nearly all the value. Median US solopreneur income is around 50,000 USD/year. The "feels successful" threshold is around 219,000 USD. The gap between the median and "successful" is where 80% of people will land — not terrible, but definitely not the story they imagined when they left their job.

Moving to micro-SaaS data: 92% of micro-SaaS businesses die within 3 years. Leading cause (42%): "no market need". Second (29%): cash flow. This is CB Insights data on SaaS startups in general — solo isn't immune.

Why is the failure rate so high, even though tools are cheaper and more powerful than ever? Because when technical barriers drop, the real barrier shifts to a side where few tech people are prepared to be good.

flowchart TD
    A[Pre-2022:<br/>Main barrier = technical] --> B[Tech people have<br/>natural advantage]
    B --> C{2024-2026:<br/>AI flattens<br/>technical barrier}
    C --> D[Barrier shifts to<br/>distribution + trust]
    D --> E[Tech people<br/>have no edge here]
    E --> F[Failure rate of<br/>solo tech still high]

Specifically four new barriers — not technical:

Barrier 1: Distribution. Justin Welsh wrote on LinkedIn for 6 years before his digital product revenue exceeded his former executive salary. Pieter Levels began building in public in 2013-2014; Photo AI launched in 2023 as the result of 10 years of audience accumulation. Tony Dinh took 20 months from launching TypingMind to 1 million USD in revenue — on top of years of building skills and a Twitter account with followers.

This is pre-AI data, and as mentioned in Part 1, dynamics may differ in the new period. There's reason to believe this curve compresses: content is produced faster with AI, new distribution channels (TikTok, YouTube Shorts) accelerate follower accumulation, and global customers are now in reach. Welsh's 6 years may now shrink to 2-3 years for a strong post-AI operator.

But there's one thing AI doesn't shorten: the accumulation curve of trust. People don't buy software because it's good. They buy because they trust the person behind it. Trust is a person-to-person relationship, not a person-to-technology one. AI writing content faster doesn't mean readers trust faster. People need to see you consistent across many cycles — launch a product, get customers, get praise, get criticism, get things wrong, fix them — before paying. That cycle has its own rhythm independent of AI.

A strong engineer leaving a company to go solo can solve the technical layer in weeks. But they'll take 12-24 months to build a distribution channel strong enough to sell — shorter than pre-AI but still longer than they imagined. And most don't have enough savings for 12-24 months of zero revenue — so they quit at month 6-9.

Barrier 2: Market choice. An engineer inside an organization has a product to build. A solo engineer has to choose what to build. This is a completely different skill — requires understanding a niche deeply, understanding what those customers actually pay for, knowing the difference between "technically interesting problem" and "problem customers pay to solve".

Most engineers going solo for the first time pick the wrong market. They build developer tools (because that's what they know) → the developer market is stingy and crowded. Or they build an AI wrapper for a broad use case → get swallowed by the foundation model providers themselves. Or they build for a market they don't belong to → they don't understand the customers' real pain.

Marc Lou shipped 23 projects in 2.5 years before ShipFast succeeded. Pieter Levels had about 70 failures before AvatarAI/PhotoAI broke through. This is pre-AI data — the cost per attempt was much higher then. Today a solo founder can ship an MVP in 3-5 days instead of 3-5 weeks. The number of attempts per unit time is 5-10x higher. But the failure rate per individual product hasn't dropped proportionally — because the barrier has shifted away from the technical layer. Shipping faster doesn't mean picking niches better. Lower cost per attempt is good news; but don't confuse that with the hit rate also rising.

Barrier 3: Operational loneliness. When solo, there's no one to discuss hard decisions with, no one to share responsibility when things break, no one to push back on your bad ideas. AI sounds like it can replace this — but as the "invisible ceiling" essay analyzed, AI tends to agree and flatter. For someone lacking integrity with themselves, AI isn't a collaborator — it's an echo chamber amplifying mistakes.

35% of solopreneurs report high stress, vs. 26% of business owners with employees. This stat is counterintuitive (fewer people = less stress, supposedly). The reason: stress doesn't come from managing others; it comes from having to decide everything yourself without anyone sharing responsibility. Absolute freedom is a psychological burden many don't anticipate.

Barrier 4: Dream collapse. This is the most subtle trap. When you're a CTO or senior engineer, life has structure: meetings, sprints, OKRs, colleagues, regular paychecks. When you go solo, all of that vanishes — and some people discover that those structures (not freedom) are what gave their work meaning. They thought they hated meetings; actually they needed the social structure. They thought they wanted freedom; actually they wanted recognition in a system with peers.

Discovering this after 12 months of burning through savings is a particular kind of suffering. And it's not rare — just rarely written about because it doesn't sell courses.

---

Part 3: Why this model still deserves attention — and why it's doubly dangerous for Vietnamese people

The doubtful part is done. Now why this model still matters — and matters most for Vietnamese people — despite the high failure rate.

One, this may be the most common business unit type in the next 10 years. The number of US businesses without employees rose from 76% (1997) to 84% (2025). The share of startups founded by a single person rose from 23.7% (2019) to 36.3% (2025). This isn't a slogan. This is a structural trend that has run for 25 years and is accelerating under AI.

Understanding this model doesn't mean having to participate. But if you're a founder or executive at a Vietnamese software company, your future competitor isn't another company with 50 engineers — it may be a solo operator in Thailand, India, or even Vietnam, alone building a product that competes directly with you. They don't need to win — they just need to eat 5-10% market share at the easiest segment, and their 90% margin will permanently squeeze yours down. This is Shift 1 and 2 from the previous essay, applied at the startup layer.

Two, Vietnamese tech people have a special advantage configuration that few other markets have. Specifically:

• Low cost of living — a Vietnamese with 12-18 months of savings can run much longer than an American with the same amount. Tony Dinh once said: "If I lived in Vietnam (my homeland), I could easily run 4 years with no revenue."
• High technical level — Vietnamese engineer training is good enough for Layers 3-4; with AI amplification, output quality matches developers from developed markets in many use cases.
• Sufficient written English — Vietnam ranks 64/123 in the EF EPI 2025, ahead of China, India, Japan. Enough for Twitter, LinkedIn, blogs, SEO. Not enough for large-scale live video or sales calls — but that's where AI is closing the gap (transcripts, auto-translation, voice cloning).
• Favorable time zone — Vietnam sleeps when the US is awake, making it easy to serve EU/US customers in parallel with a Vietnamese life.

Tony Dinh and Daniel Nguyen prove this isn't theory. Both are in Vietnam (or of Vietnamese origin), serving global markets, earning USD, spending VND. This is the exchange rate as inverse leverage — a business model only really viable for people in emerging markets with developed-market skills.

Three, but the same configuration is also a special trap for Vietnamese people. Because:

• Domestic salaries are lower, so the feeling that "going solo is worth it" is stronger. A US CTO earning 300K USD/year has very high psychological resistance to leaving the chair. A Vietnam CTO earning 80-150K USD/year is more tempted — and that's precisely when it's most dangerous. Because lower salary doesn't mean lower personal risk; the absolute opportunity cost is lower but the success rate isn't higher.
• Financial infrastructure is incomplete. Stripe doesn't support businesses registered in Vietnam — this is a real technical barrier that many don't know about until they hit it at month 6. You have to route through merchants-of-record (Lemon Squeezy, Paddle) or register an LLC in the US/UK — both with costs and legal complexity that beginners easily get wrong.
• Vietnam's tax reform is mid-way. Resolution 68-NQ/TW (May 4, 2025) ordered the abolition of the household business lump-sum tax by 2026 at the latest, codified through Tax Administration Law 108/2025/QH15 and Decree 68/2026/NĐ-CP (effective Jan 1, 2026). The consequence: under 500 million VND/year is exempt from both VAT and PIT; 500 million – 3 billion VND can choose between two tax methods; over 3 billion must use the (revenue − expense) method. This is a major shift toward greater transparency but also greater complexity — and many young solopreneurs who have never done bookkeeping will drown in it.
• The "company without a brand name" model has no social precedent in Vietnam. The question "what do you do?" needs a short answer accepted by relatives. "I'm a solopreneur, 30K USD/month revenue" doesn't carry the same social weight as "I'm CTO at company X". Most Vietnamese tech people underestimate the psychological burden of this — until they live with it for 6 months.

---

Part 4: Expanding the scope — one person doing the work of a full SME

The section above said "one person doing the work of 5-7 people on the same software product". But that's just the narrow view. The broader view — which is taking shape — is: one person can run the entire scope of a multi-functional small or medium enterprise (SME), not just a software product.

Think about a traditional Vietnamese SME — e.g., a 15-person print shop, or a 20-person marketing agency, or an e-commerce shop with 30 warehouse-logistics staff. Typical headcount:

• 1-2 sales/customer service
• 2-3 accounting/admin
• 5-10 core operations (printing, content, packaging)
• 1-2 marketing
• 1 IT/web
• 1-2 management
• 1 founder/director

Total 15-25 people, typical revenue 200-500 million VND/month. This is the backbone of the Vietnamese economy — millions of SMEs like this.

Look at that list again. Most of those roles aren't "making software". They do customer service, accounting, marketing, operations, management. Each role is a kind of knowledge work with a concrete process.

The question: in the next 3-5 years, can a solo operator with AI amplification run this entire business?

The cautious answer is "most of it, not all". Specifically:

flowchart LR
    Solo[Solo operator<br/>+ AI agents]

    Solo --> A[Sales/CS<br/>AI chatbot, voice agent]
    Solo --> B[Basic accounting<br/>MISA + AI bookkeeping]
    Solo --> C[Marketing content<br/>AI generates + human reviews]
    Solo --> D[Simple operations<br/>Automation + 1-2 part-time]
    Solo --> E[Data management<br/>AI dashboard]

    Solo -.needs humans.-> F[Complex physical operations]
    Solo -.needs humans.-> G[Strategic B2B relationships]
    Solo -.needs humans.-> H[Heavy legal decisions]

The top five roles — sales/CS, basic accounting, marketing content, automatable operations, data management — are within reach of a strong solo operator with a modern AI stack. In 2026 there's already real data: one-person Shopify e-commerce shops with 500K-2M USD/year in revenue with 1 owner + 2-3 contractors + 5-10 AI workflows. They don't have a "company" in the traditional sense. They have a human + a machine.

The bottom three roles — complex physical operations (large-scale printing, large warehouses, manufacturing), strategic B2B relationships (million-dollar contracts, long-term partnerships), heavy legal decisions — still need humans. But the number of humans needed is much smaller, and they're typically project hires, not permanent employees.

Consequence if this trend is right: a completely new class of niche SME may emerge in Vietnam — 1-3 person scale, 100-500K USD/year revenue, 70-90% margins. This wasn't viable before because of the minimum payroll of a traditional SME.

At the same time — and this is the point Vietnamese tech folks should pay attention to — this model spawns a new generation of specialized roles that don't exist in the current occupational classification.

Role 1 — The SME AI setup person (extension of Layer 0 from the previous essay). Millions of Vietnamese SMEs can't set up internal AI systems themselves. They need someone to do it. This isn't pure IT — it's a person who understands both business operations and how to choose models, write prompts, integrate tools, connect with existing Vietnamese accounting/CRM systems (MISA, KiotViet, Sapo). This is a huge employment opportunity over the next 5 years, with extremely lopsided supply-demand.

Role 2 — The "multi-solo" HR manager. As more solo operators emerge, each with 1-3 part-time contractors, a new market forms: people specialized in matching contractors with solos, managing quality, acting as "fractional HR" for multiple solos simultaneously. This role doesn't exist in Vietnam yet but is emerging in the US (through platforms like Athena, Magic).

Role 3 — Vertical specialist + AI. An accountant who understands MISA + can write AI workflows to automate accounting is worth 3-5x a pure accountant. A marketer who understands SEO + can use AI to scale content is worth 5x a pure marketer. The general pattern: vertical specialist + AI capability = a new role with high premium. This is the opportunity for people in "non-tech" industries — they don't need to become programmers, they just need to add an AI layer to their existing expertise.

Role 4 — The vertical integrator (personal-scale). A specialized form: a person who deeply understands a niche (e.g., real estate, small F&B chains, short-haul logistics) and can also build software + operate services in that niche. They're not a "software company" nor an "industry X company" — they're both in one person. The number of these people in Vietnam is currently very small because it requires rare skill combinations; precisely because they're rare, the value premium is high.

Role 5 — "Operator-as-a-service". A new kind of service: hire someone to operate a part of your SaaS for 6-12 months to shape the process, train the AI agents, then hand back. This is productized consulting but deeper and longer than traditional consulting. Emerging in developed markets; not yet visible in Vietnam but likely soon.

The main spirit: the one-person model isn't "job replacement" — it's redistribution of types of work. Old roles (entry-level accountants, pure content writers, tier-1 customer support) shrink. New roles (vertical AI integration, agent operation, multi-solo contractor matching) emerge. Total jobs may stay the same or grow, but the distribution is completely different.

This is where Vietnamese tech people have a natural advantage — because both wings (solo operating + creating new deep specializations) require strong technical foundations. But at the same time, if you only have technical foundations without developing a vertical dimension, you'll get squeezed in the middle — not specialized enough to compete with vertical integrators, not wide enough to run a full SME.

---

Part 5: Two paths — and why hybrid beats both

People often talk about two paths: Path A (build products — SaaS, AI wrappers, boilerplate) and Path B (services — consulting, fractional CTO, productized services, courses). Both are viable, with Vietnamese examples succeeding on each path.

But actual data shows the hybrid model beats both pure paths. Justin Welsh, Marc Lou, Greg Isenberg, Tony Dinh — all run a hybrid model: a personal brand + audience distributing both digital products (courses, templates) and services (or SaaS). The audience accumulates; each new product inherits the distribution channel of the previous one.

flowchart LR
    A[Path A — Pure SaaS<br/>Build first, channel later] --> X[Problem: 6-24 months<br/>no revenue]
    B[Path B — Pure Services<br/>Channel first, build later] --> Y[Problem: income ceiling<br/>at personal billable hours]

    H[Hybrid — Audience-led<br/>Channel immediately + both product types]
    A -.beats.-> H
    B -.beats.-> H

    H --> Z1[Early revenue from services]
    H --> Z2[Audience distributes<br/>digital products later]
    H --> Z3[Digital products scale<br/>past the hours ceiling]

Specific comparison:

Lesson to draw: for Vietnamese tech people considering this, hybrid with Path B in the lead is the least risky choice. Because:

1. Services generate cash flow quickly (2-8 weeks) — you don't burn through savings before validating there's a market.
2. When doing services, you're close to customers — you hear real problems, not imagined ones. This is the most valuable input for choosing a SaaS product later.
3. Your audience accumulates from day one (writing about doing services). When you launch a product in year 2, the channel is already there.
4. Productized service (services packaged into a product with fixed scope) is a good intermediate step between pure services and pure products — you learn pricing, packaging, marketing skills before investing in software product development.

Important warning: don't confuse "hybrid" with "distracted". Someone doing 3 things in parallel isn't hybrid — they're just unable to decide. True hybrid is one concrete wedge (e.g., AI implementation consulting for Southeast Asian SaaS/AI companies → course packaging the knowledge → eventually a SaaS tool for the same audience). Everything serves the same audience, the same brand story.

---

Part 6: Infrastructure stack for Vietnamese operators — the part few people write about

This is the concrete section, worth bookmarking. Most English solopreneur blogs assume you're in the US/EU/Singapore. Vietnamese operators need a different stack.

Tooling layer (total 80-120 USD/month):

Legal layer — this is more important than people think:

There are three practical configurations for Vietnamese operators:

flowchart TD
    Q{Target revenue source?} --> VN[Vietnamese customers<br/>paying VND]
    Q --> Global[Global customers<br/>paying USD]
    Q --> Mixed[Mixed]

    VN --> S1[Hộ kinh doanh<br/>if &lt;500M VND/year<br/>→ TNHH MTV if more]
    Global --> S2[US LLC or UK Ltd<br/>+ Stripe direct<br/>+ Wise/Payoneer for USD inflow]
    Mixed --> S3[Foreign LLC for global<br/>+ HKD/TNHH for VN<br/>+ Wise Business connecting]

Details per configuration:

Configuration 1 — Serving the Vietnamese market:

• Household business if under 500 million VND/year (per Decree 68/2026, this tier is exempt from VAT and PIT)
• Upgrade to single-member LLC when exceeding 500 million VND/year or needing to sign large B2B contracts
• Payment: VNPay, MoMo, ZaloPay, VietQR via NAPAS 247
• Hire a bookkeeping service (~2-5 million VND/month) — don't DIY
• Limitation: Vietnamese subscription willingness is weak, margins are hard to keep high

Configuration 2 — Serving the global market (recommended for most):

• Delaware LLC via Stripe Atlas (500 USD), doola or Firstbase.io (399-599 USD), Globalgo Fast (249 USD), StartGlobal (599 USD). Or UK Ltd via Incorpuk (~100-300 GBP) if more comfortable with the UK system.
• Stripe direct via foreign entity (Stripe Atlas gifts 2,500 USD Stripe credit)
• USD inflow: Wise Business (receive only, can't send from Vietnam yet) + Payoneer (1% receive + up to 3.5% FX) + Hurupay (1% flat, USDC) + Hawala
• US tax compliance cost ~150 USD/year (Form 5472 + 1120 for single-member LLC of non-US person)
• Vietnam side: still need to file personal PIT on income received from the foreign LLC (see bookkeeping service)

Configuration 3 — Hybrid, advanced:

• Apply once you have revenue from both VN and global, need to optimize tax
• This is where you must have a specialist accountant — don't DIY

One often-overlooked point: if you're a CTO at a company considering going solo, handle this legal layer in the first 30 days, not when you already have revenue. Reason: wrong initial setup then fixing later costs 3-5x correct initial setup. Especially LLC registration takes 2-4 weeks and requires full information before your first customer — you don't want to land a customer and not know how to invoice them.

---

Part 7: Decision framework — should you try?

This section is personal. Every essay arriving here usually lists 10 traits of successful solopreneurs. I won't do that. Instead, six questions you have to answer honestly with yourself (see "The Invisible Ceiling" essay on why this is hard):

flowchart TD
    Start[You're thinking<br/>about going solo] --> Q1{1. Do you have 12-18 months<br/>of savings?}
    Q1 -->|No| End1[Don't go full-time<br/>Start as side venture]
    Q1 -->|Yes| Q2{2. Can you handle 18 months<br/>without a 'title'?}

    Q2 -->|Not sure| End2[Come back after trying<br/>side venture for 6 months]
    Q2 -->|Yes| Q3{3. Willing to write in English<br/>4-5 times/week<br/>for 12-18 months?}

    Q3 -->|No| End3[Pick a different path —<br/>senior employee + side project]
    Q3 -->|Yes| Q4{4. Family agrees<br/>with specific conditions?}

    Q4 -->|Not yet| End4[Have that conversation<br/>before deciding]
    Q4 -->|Yes| Q5{5. A specific niche<br/>where you have<br/>asymmetric advantage?}

    Q5 -->|Not clear| End5[Validate with 15<br/>customer interviews first]
    Q5 -->|Yes| Q6{6. Baseline health and<br/>energy are okay?}

    Q6 -->|No| End6[Recover first —<br/>solo is not the time<br/>to fix burnout]
    Q6 -->|Yes| Ready[Try side venture<br/>10-15 hours/week for 90 days<br/>before deciding]

Each question above can trigger self-deception — especially question 1 (counting savings) and question 5 (assessing personal advantage). To counter this, a specific technique: write your answers on paper, hand them to someone who won't flatter you, and ask them "do you believe me?". If they hesitate, that's a signal. Don't go to AI for confirmation — as said in the previous essay, AI will agree, and that's exactly where it's most dangerous.

Threshold for "qualified to leave the CTO chair" (all 5 conditions simultaneously, not 4):

1. Side venture has generated 60-80% of current CTO salary with <20 hours/week, sustained for 3 consecutive months
2. Has ≥6 months of living expenses even if revenue drops to 0 immediately after leaving
3. Pipeline (signed retainers, MRR, signed projects) covers ≥4 months ahead
4. Has a clear re-entry plan — specific list of 3-5 people who can refer you back to a CTO/Head of Engineering position within 60 days if needed
5. Spouse and parents have agreed with specific milestones, not general agreement

What to note: there's no "leave the chair to have time to build the product" in this threshold. If you need to leave the chair first to have time to build, you're not ready. Why? Because side venture while full-time employed is the real test of the self-discipline that solo life requires every day. Someone who can't carve out 10-15 hours/week with structure won't be able to build structure on their own without structure.

---

Part 8: Warning for those with special advantage — CTO/Head of Engineering

There's a reader group that needs to be addressed separately: those currently CTO or Head of Engineering at Vietnamese companies.

You're in the sweetest position to be tempted by this model. Because:

• You have high technical skills
• You have a network (clients, peers, vendors)
• You have credibility — the "former CTO" story sells productized services immediately
• You see Tony Dinh, Daniel Nguyen — and think "I'm technically better than them"

And you're right — you might be technically better than them. That's not the deciding variable.

The deciding variable for the solopreneur path is distribution + market choice + persistence through the 18-24 month death valley. None of these three relate to your technical skill. Tony Dinh didn't succeed because he coded better than everyone — he succeeded by persistently writing build-in-public in English on Twitter for years, and choosing TypingMind (an abstraction layer for ChatGPT) at exactly the right time the market needed it.

A CTO in Vietnam has a special temptation: the feeling "I'm managing people, I want to go back to making products". The feeling is real. But it doesn't necessarily lead to a solo solution — it might lead to "become an IC engineer again" (IC engineer track at a large company), or "find a Head of Product/Tech position at a smaller startup", or "leave the CTO chair to be CTO at a founder/CTO startup". Solo is one of many choices — not the default.

And there's another, subtler temptation: believing that because you successfully managed teams of 50-100 engineers, you'll also run solo well. These are different skills. Running a large team teaches you communication, systems, prioritization. Running solo teaches you marketing, sales, customer support, accounting, content writing — mostly things you've outsourced to teams before. You're not going back to being an IC engineer; you're becoming CMO + CFO + COO + IC engineer simultaneously.

If you still want to try (and this is a personal decision, no one has the right to judge), specific tactical suggestions:

1. Don't quit in the first 90 days. Run a side venture 10-15 hours/week.
2. Start with Path B (productized consulting/audits for Vietnamese and Southeast Asian SaaS/AI companies, priced 5-25K USD per engagement). Time to first revenue: 2-8 weeks.
3. Write in English on LinkedIn (primary) + X (secondary), 4-5 posts/week from week 1. Topic: the intersection of "CTO lessons in scaling engineering teams" and "what AI is doing to the same problems". This is the distribution channel — minimum 18 months before it becomes an inbound channel.
4. Set up the legal/payments layer in month 1 — not when you have customers. Delaware LLC via doola or Stripe Atlas. Lemon Squeezy account. Wise + Payoneer.
5. Set specific 90-day goals: (a) one paid pilot of 3-10K USD delivered, (b) 1,000+ LinkedIn followers, (c) 15 customer interviews synthesized, (d) one concrete offer (product or service) tested with ≥3 prospects.
6. Only leave the CTO chair when all 5 conditions in Part 7 are met. Before that, this is a portfolio bet, not a career change.

---

Closing: Four questions for yourself

Continuing the structure of the previous essays — four questions you don't need to answer to anyone, only to yourself:

One, do you want to go solo for real reasons, or for reasons you think you should want? There's a difference between "I've been trying side projects for 2 years, there's market signal, and this is the natural next step" and "I'm tired of my current company, read about Tony Dinh, and I think I can too". The first reason is a signal; the second is a reaction. The reaction might be right — but needs to be tested carefully.

Two, if you go solo and fail — a realistic scenario in 30% of good cases and 60% of bad cases — can you accept it? Not financially (that's handled in the decision framework). But in terms of reputation, identity, relationships. People will know. You'll have to tell that story many times. What's the story? If the answer is "I'd be ashamed", maybe you're not ready.

Three, who will give you real feedback throughout the journey? Not AI (see "Invisible Ceiling"). Not family (they want you to be okay). Not Twitter fans (they want a good story). One or two specific people, with relevant expertise, not financially dependent on you, willing to speak frankly. If you don't have anyone like that, that's the work to do before going solo, not after.

Four, are you looking at the solopreneur model as an escape, or as a tool? This is the most important question. If it's an escape — from a bad boss, from company culture, from managing people — then the root solution isn't solo, it's solving what you're escaping. If it's a tool — you're building a long-term career path, and solo is a specific phase with a reason — then the conversation is completely different. Ask yourself this at 6 AM when no one is flattering you, no AI to ask, no audience to perform an answer for.

---

Note: Where this essay might be wrong

As with every essay that strays far from solid data, there are places worth doubting:

One, the supporting data is US-biased. Most numbers (29.8 million solopreneurs, 77% profitable in year one, 92% micro-SaaS dying in 3 years) come from the US. Vietnam doesn't publish equivalent stats at this level of detail — only household business data (~5 million households nationwide, affected by the 2026 tax reform). Actual success rates in the Vietnamese market with different dynamics may be significantly different — perhaps lower because financial infrastructure is incomplete, or perhaps higher because the USD niche is less competitive from the Vietnam side. There's no data to assert either way.

Two, the "low cost of living + USD income" advantage may erode faster than predicted. VND/USD exchange rate can fluctuate. Cost of living in major Vietnamese cities is rising fast — downtown apartments in HCMC or Hanoi are approaching Bangkok/Kuala Lumpur prices. If this trend continues, Tony Dinh's "4 years runway" advantage may shrink to 2 years in the next 5 years.

Three, AI changes faster than this essay can keep up. The specific tool stack (Cursor, Claude Code, Lemon Squeezy) is accurate as of May 2026 — but specific tool names may change within 12 months. The category-level argument (a solo founder can coordinate the workload of 5-8 people through AI) is more durable. Read this essay for framework, don't rigidly hold to tool names.

Four, inverse survivorship bias. This essay actively warns about survivorship bias — but the essay itself may suffer from another kind of survivorship bias. Cases like Tony Dinh and Daniel Nguyen are known because they succeeded. There may be hundreds of Tony Dinhs who failed silently, and the sample for this analysis is still skewed despite my attempts at balance. Read numbers like "60-80% of CTO salary", "18-36 months to replacement revenue" with ±50% uncertainty.

Five, most data cited is pre-AI. As mentioned in Part 1, the numbers "Welsh's 6 years on LinkedIn", "Marc Lou's 23 projects", "Levels' 70 failures" all describe the old game. Post-AI operators can try 5-10x faster, but may also hit ceilings elsewhere we don't yet know about. When I write "18-36 months to replacement revenue", treat it as an estimate with ±50% bandwidth — and the bandwidth may be even wider for post-AI operators with not-yet-long-enough actual data.

Six, the conclusion "most readers of this essay shouldn't try" may be too cautious. Some readers will find themselves fitting. They may be right. My warning leans toward "don't decide quickly" more than "don't do it" — and that lean may be too heavy for some people at the right moment in their lives. Read this essay as a cautionary film, not a prohibition. The final decision is yours.

Seven, the predictions about one-person SMEs and new specialized roles (Part 4) are the farthest-reaching reasoning. Patterns like "accountant + AI", "marketer + AI", "personal-scale vertical integrator" have a few examples in developed markets, but in Vietnam they're embryonic. Maybe these roles won't be as common as predicted — because of organizational culture barriers ("a business must be a company"), because the law isn't favorable yet for the form of one person running a multi-function SME, because traditional customers still prefer transacting with "companies" over individuals. Read Part 4 as a scenario, not a certain roadmap.

Read this essay as a scenario worth thinking about, not a confident forecast — just like every other essay.