The steepest adoption curve consumer technology has ever seen.
Nothing has ever spread like this — not the telephone, not the smartphone, not social media. And the spending behind it is only accelerating.
Time to reach 100 million users
Everyone uses it. Almost nobody can direct it.
Usage went vertical; skill stayed flat. The research consensus is blunt: most people cannot prompt effectively, prompt quality measurably predicts output quality — and there is still no standard way to measure or prove the skill. A billion users, no credential. That is the empty middle of the entire AI economy.
The gap between knowing and doing
The market already pays a premium for people who can direct AI.
Not a projection — measured on real job postings, over a billion of them, and it grew every single year it has been measured.
Advertised wage premium for AI skills
Two honest caveats, because the numbers deserve them: the premiums are measured on what employers advertise, which is strong evidence but not the same as individual raises. And at the company level, a widely-cited MIT report found most enterprise AI pilots delivered no measurable profit impact — the value shows up at the skilled-individual level first. Trillions in, little out, until someone in the room can actually direct the tools. That unevenness is not a hole in the story. It is the story.
The same tool raises one worker and lowers another.
Both findings are rigorous. They don’t contradict each other — they describe two different ways of using the same machine, and the dividing line is exactly one skill.
Direct it — value rises
In complex work, AI raises preparedness, control and self-efficacy: the human decides, the machine drafts, the human judges. That is the use pattern behind the 62% premium and the 4× productivity number.
The mechanism matters: the skilled user keeps conceptual engagement — they know what they want before the model speaks, and they reject what misses it.
PwC · Lightcast · task-complexity studiesOffload to it — skill falls
MIT Media Lab put writers under EEG: those who let the model write showed lower cognitive engagement — and later performed worse without AI than people who had never used it. The authors named it cognitive debt.
A 666-participant study links heavy passive use to weaker critical thinking, mediated by cognitive offloading. And output homogenises — run everyone’s ideas through the same model and everyone starts sounding the same.
MIT Media Lab · Gerlich 2025 · Doshi & Hauser 2024AI raises your value when you direct it and keep thinking.
It lowers your value when you hand it your thinking.
A billion people are doing the second thing. The premium goes to the first.
Numbers worth repeating at dinner.
ChatGPT reached 100 million users in two months. Instagram needed two and a half years.
public launch metricsIndia became ChatGPT’s second-largest market in early 2026 — the growth is coming from Asia, not Silicon Valley.
regional usage reportsThe gap between workers who say AI skills are critical (54%) and workers training them (4%). Half a population knows; a rounding error acts.
workforce surveys 2025In UK job postings, the advertised premium for AI skills now exceeds the premium for a master’s degree — and approaches a PhD’s.
UK postings analysisThe AI-skills wage premium doubled in a single year — 25% to 56% — then kept climbing to 62%.
PwC AI Jobs BarometerStandard, accepted ways to measure how well a person directs AI. The skill commands a premium and still has no credential. That vacuum is what a verifiable record fills.
measurement research, 2025From the papers.
Worldwide AI spending forecast to hit $2.52 trillion in 2026 — up 44% year over year.
The AI Jobs Barometer update finds a 62% wage premium for roles requiring AI skills, and near-fourfold productivity growth in AI-exposed industries. Jobs grew despite automation fears.
EEG study introduces “cognitive debt”: LLM-assisted writers showed lower cognitive engagement and reduced unassisted performance afterwards.
Over 70% of organisations now use generative AI in at least one function — while only 1% rate themselves AI-mature.
44% of US adults have used ChatGPT; 54.6% of the working-age population had adopted generative AI by August 2025.
The gap is the opportunity. The record is the proof.
For you
The ten skills of directing AI are learnable — the same ones a founder runs a company on: know what you want, say it precisely, judge what comes back, ship it to a real person. Makery teaches them from the dough up and keeps the verifiable record as you go. The premium is the market’s; the proof can be yours.
Start free — no cardFor companies
Trillions are going in and most pilots return nothing, because prompts are easy and direction is rare. The fix is not another tool — it is people whose record shows they reject bad output, verify claims, and ship things real users kept using. That record is searchable here.
Find directed talentsources · Statista · Microsoft · McKinsey · Gartner · PwC AI Jobs Barometer 2025/2026 · Lightcast · Pew Research · MIT Media Lab · Gerlich 2025 · Doshi & Hauser 2024 — premiums are measured on advertised postings; Makery makes no income or job-outcome promises. More context on the Recipes blog.