I. Before 1995 - Knowledge Was A Privilege

Picture the world in 1990. What would a student in Hai Phong do to learn programming? Buy a hand-copied, illegally translated Pascal textbook at Trang Tien. Go to the nearest university and beg to photocopy a C reference. If lucky enough to know someone at the Bach Khoa computer science department, borrow a few textbooks. A problem you didn't understand - no one to ask. A C library you needed - no way to download it. A bug - you could only guess.

Before the internet, highly technical knowledge sat behind three locked doors: money (an English-language textbook cost half an engineer's monthly salary), geography (only a handful of major cities had a university library worth the name), and network (mentors and family connections decided who got access to which information). A self-learner's ceiling depended almost linearly on what their parents did for a living, where they lived, and whether they had enough money.

This wasn't unique to Vietnam. In the US in 1990, breaking into finance required a Wharton or Stanford MBA - not because the knowledge didn't exist elsewhere, but because of signaling and networks. Becoming a doctor, a software engineer, a lawyer - all required passing through a top school, a high tuition bill, an admissions process that favored people who already had a foundation. A degree isn't knowledge; a degree is a certificate that you passed the filter. When knowledge is scarce, the filter is the only way to allocate it.

II. The First Revolution - The Web, 1995–2015

The first revolution didn't start on a single day. It was a twenty-year chain of small tools stacking on top of each other - each one chipping a hole in the wall of the barrier, until the wall could no longer stand.

Result: a self-taught generation enters Google without a degree

Within the 20 years after 1995, a new kind of profession appeared that had never existed before in human history: the self-taught software engineer. In 2024, the Stack Overflow Developer Survey reported that roughly 51% of professional developers learned to code mostly outside of formal schooling - through YouTube, freeCodeCamp, Udemy, books, and side projects; about 34% held no university degree. That ratio was unimaginable in any highly technical field before this - you can't become a self-taught doctor, a self-taught civil engineer, a self-taught lawyer. But a self-taught software engineer, you could.

Jobs that didn't exist before 1995 - now high-paying careers

Worth noting: not one job in the table above had a stable school curriculum before it appeared. Every one of these fields grew out of an open community - Stack Overflow, GitHub, Reddit, Hacker News, Twitter - and hiring processes gradually accepted a portfolio in place of a degree. This is exactly what traditional university education couldn't keep pace with: the time from "a technology is born" to "you can earn a living from it" shrank from ~20 years (mechanical engineers in the 1850s) down to ~3–5 years.

III. The Second Revolution - AI, From 11/2022

On November 30, 2022, OpenAI launched ChatGPT. Within 5 days, the product had 1 million users. Within 2 months, 100 million - faster than any consumer app in history (TikTok took 9 months, Instagram 2.5 years, Facebook 4.5 years). This wasn't a marketing explosion; it was a signal that a new barrier had just collapsed.

This barrier collapse is fundamentally different in nature from 1995. The internet broke down the barrier to accessing knowledge. AI is breaking down the barrier to executing knowledge. Before AI, knowing Python didn't mean you could write an app - you still had to type every line, debug for hours, deploy every config by hand. After AI, the line between "knowing a concept" and "shipping a working product" has thinned dramatically.

Two kinds of compression: knowledge (like 1995) + execution (new)

IV. Opportunity - Who Wins This Time

1. Those who already knew how to teach themselves - a multiplier, not a replay

An easy mistake when looking at wave two is thinking each revolution resets the playing field, giving newcomers a fresh chance to start from the same line. Wrong. Every time a barrier collapses, people who already have the self-teaching meta-skill don't get reset - they get multiplied.

Concretely: in wave one, people who knew how to teach themselves (reading technical books, asking on forums, doing projects) beat out people waiting for a school to teach them. When Stack Overflow launched, people already used to digging up information on their own learned 3–5 times faster than those who weren't. In wave two, that same group now has AI as an assistant - their learning and building speed multiplies another 3–5 times. Not 3–5× over zero; 3–5× over the level already multiplied in wave one.

This is the counterintuitive part of democratized knowledge: it does not lead to equal opportunity. It leads to a new kind of inequality - one that sorts people by self-direction ability. Because once the internet and AI have stripped away the external barriers (tuition money, geography, network), the barrier left standing is the internal one: self-study discipline, taste, the ability to ask the right question. That part doesn't get automatically compressed by tools. Right now, it's become the main differentiator.

Given the same AI model, two users can produce outputs that differ by 10×. Someone who doesn't know how to ask gets generic answers from AI. Someone who knows how to ask well, knows when AI is wrong, knows when to stop and think for themselves - gets their own ability multiplied by AI. This is why, in recent YC batches, the people getting the most out of AI tools aren't curious junior developers - they're senior builders who already shipped products before AI and now ship five times faster.

2. Solo founder + AI = the company that used to need 5 people

In 2024, Y Combinator published a notable figure: the share of "1-person teams" in each batch rose from ~3% (2019) to ~15% (2025). On the Lex Fridman podcast in 2024, Sam Altman predicted the "one-person billion-dollar company" would appear within a few years - not fantasy, but a straight-line extrapolation of the rising AI-leverage-per-person ratio.

3. Non-technical backgrounds enter software

Wave one broke the barrier for self-learners - but it still demanded a certain kind of technical mindset. You had to be patient enough to sit for 8 hours debugging a regex, comfortable enough with syntax. Most designers, marketers, teachers, and lawyers never got past that barrier.

Wave two lowers that barrier for this exact group. A product designer can now build a Next.js prototype without a developer. A marketer can write their own automation script instead of waiting on engineering. A lawyer can spin up a SaaS product serving a small niche in their field. Not beautiful code - code that runs. That's the point.

4. Countries outside the core: the English barrier drops

One of the most invisible barriers of wave one was English. Stack Overflow in English, GitHub READMEs in English, Hacker News in English, and higher developer pay meant selling to English-speaking American clients. Vietnam, China, Japan, Brazil - all had to clear the language barrier.

AI's wave two lowers this barrier too. Claude and GPT translate in real time, write English-language content from Vietnamese ideas, fix emails, write product copy. A Vietnamese developer in 2026 can publish an app for the US market with polished product copy, without needing an English copywriter. This is a new door - especially for indie founders in countries with low costs, high technical skill, and a small domestic market.

V. Risk - Who Gets Left Behind

Every time a barrier collapses, it creates an opportunity for one group and digs a hole for another. Wave one dug a hole for anyone who didn't teach themselves. Wave two digs a hole that's deeper and wider - because the compression speed this time is roughly 10× faster than the first wave. There are at least six risk zones worth watching.

VI. Vietnam - Where It Stands On The Map

Vietnam was one of the biggest beneficiaries of wave one. The question for 2026 is: will it benefit just as much from wave two, or will it fall into the group being compressed instead?

The bright side - three advantages

The dark side - three risks

A comparative lesson: who saw wave one coming?

In 1995, no one predicted that a self-taught Vietnamese developer would end up at Google. There were people who forecast that "the internet will change everything" - but exactly how it would change things, exactly which opportunities would open for which people, which industries - no. The winners of wave one weren't the ones who forecast correctly; they were the ones who jumped in early, while the opportunity was still unclear.

In 2026, we're at an equivalent stage: wave two is clearly underway, but exactly "which jobs will exist in the next 10 years" - nobody knows. "Prompt engineer," "AI app builder," "context manager," "AI safety auditor" might be job titles in 2030. Or they might not be. What's highly likely: many of the high-paying jobs of 2030 don't have a name yet today - just as "DevOps" had no name in 1995.

VII. Conclusion - Two Lessons, One Question

Two barrier collapses teach us two durable lessons:

1. Every wave creates an opportunity for early self-learners. In wave one, people got into Google without a degree. In wave two, people ship six-figure-revenue products with a team of one. Every compression event lifts a new generation up - and they usually don't start from a position of privilege.
2. Every wave leaves a group behind. In wave one, it was people who didn't believe the internet was real, people who lacked the patience to teach themselves, people working in industries that couldn't digitize fast. In wave two, it's juniors stuck at the level of tasks AI now does instead, people who learned only to get a degree, and companies selling "cheap" instead of "value." The mechanism is the same; the group it hits is different.

There's still an open question with no answer yet:

Some candidates for the barrier that's left: taste (knowing what's worth building), distribution (getting a product in front of the right people), trust (who believes your product when AI spam is everywhere), energy (compute isn't free - it needs electricity, land, chips), and physical labor (manual trades haven't been compressed yet - robotics may be the third barrier to collapse within the next 10 years).

Concrete advice, for Vietnamese readers in 2026, especially those under 35:

• Don't learn just to type code. AI already types well. Learn to understand the product, the user, the market. Domain knowledge + a bit of technical skill + AI leverage - these three things combined matter more than mastering syntax.
• Ship, don't just keep studying. In wave one, someone who read 100 books lost to someone who did 10 projects. Wave two widens that gap even further - because AI shortens build time. One MVP a month for two years gives you 24 shots. A master's degree gives you zero.
• Distribution > coding. Learn to write, learn to make content, learn SEO, learn to sell. Someone who ships a great product nobody knows about loses to someone who ships an average product with an audience.
• Don't bet everything on the hype. Build a product with real value - one that doesn't depend on any single LLM existing. If your product is just a "wrapper around GPT," the day OpenAI or Anthropic builds that exact feature into their own product, you're dead.
• English still matters - just in a new way. Not for coding; AI handles that. But for talking to international users, writing product copy, participating in the builder community. The places that decide who hears about your product are still Twitter, Reddit, Hacker News - and they still speak English.

The internet knocked down one wall. AI is knocking down a second. In the next 10 years, a third wall may fall - robotics, BCI, AGI, or something that doesn't have a name yet. The central question isn't "should I learn" - of course you should. The question is: how fast can you adapt when the next barrier falls?

Whoever answers that question will win wave three. Just as they won wave one, and wave two.