Ten people used to be the smallest team that could ship software. One person can now do that work. Here is what changed, what it costs, and the part the pitch leaves out.
For twenty years, ten people was the smallest team that could build and run a software product. That was not a decision anyone questioned. It was the entry fee.
Serkan Inci is what happens when the fee stops being mandatory. He is 39, Turkish, and trained as a marine engineer. He quit. In December 2009 he co-founded İnci Sözlük with İsmail Alpen, one of Turkey's wildest social platforms, and helped turn a local image-meme format called caps into a mass habit that spilled onto Facebook and Twitter. Then an ad agency. Then he sold his businesses, moved to the United States, and started from zero. In America he went the opposite direction from software: a used car dealership, contracting, automotive businesses, even firearms importing. Physical, local, unglamorous work.
Then AI happened. His portfolio today is a handful of demand-capture sites for local trades: septic tank pumping, mobile auto glass, stump grinding, car acquisition, and a website platform for independent used car dealers.
His claim is narrow and specific. He did not say teams are dead. He said none of those products needed a ten-person team to get built and launched. One person could now do the work.
The claim sounds like founder Twitter. The arithmetic behind it does not.
Start with the bill. A fully loaded ten-person US software team runs about $2.0M to $2.4M a year, and that is the middle of the market. The Bureau of Labor Statistics puts the mean software developer salary at $148,100, and employer burden, meaning payroll taxes, health coverage, retirement and paid leave, adds roughly 40 to 45 percent on top. A senior team in a major hub runs $2.7M to $3.5M. Excluded from all of it: office rent, and the recruiting pipeline you refill every time someone leaves.
For a founder, that number was a gate. You raised it, you hired it, you shipped. Which means the interesting question is not whether one person can build a product. It is what happens to the gate when the work behind it stops requiring ten salaries.
The clearest evidence is what companies produce per head. Read these as run-rate estimates from private companies, because none of them publish audited revenue.
| Company | Milestone | Per employee |
|---|---|---|
| Cursor (Anysphere) | $500M ARR / 60 people | ~$8.3M |
| Midjourney | ~$200M revenue / 40 people | ~$5.0M |
| Lovable | $100M ARR / 45 people | ~$2.2M |
| Gamma | $50M ARR / ~30 people | ~$1.7M |
| Public SaaS median | 2025 cohort | ~$393K |
| Private SaaS median | 2026 benchmark | ~$141K |
Private AI-native companies at their matched milestone, against the public SaaS median. Run-rate estimates.
Bars are scaled to the $8.3M maximum. The public SaaS median sits at about 4.7 percent of the top bar.
Read the bottom two rows first. A normal public software company produces roughly $400K per employee. The AI-native names run an order of magnitude above that. Some of it is usage-based revenue and early-stage luck, and none of it is audited. But if one person can carry a share of the output that used to take ten, this is where it shows up in a number.
Speed moved the same way. Stripe found the top 100 AI companies on its platform reached $1M in annualized revenue in a median 11.5 months, against 15.5 months for the top SaaS cohort of 2018. And Carta reports that 36 percent of startups formed on its platform in 2025 had a single founder, a record, up from 31 percent a year earlier. Carta calls that correlation rather than proof. It is still the direction everything is moving.
Here is the detail that matters, and it is not a unicorn.
Take the septic pumping site in that portfolio. It is a booking front end that quotes by tank size, $249 to $449 depending on capacity, plus a lid dig-out add-on. It holds a deposit until the job is complete. It sells across all fifty states and territories, including American Samoa, one generated page per location. And it publishes an llms.txt file, the machine-readable index that tells AI crawlers what the site contains.
That is not a startup. That is a person who understood that the ten-person team was never only a tech-industry thing. The local service economy, the plumbers, the tree crews, the glass installers, was built entirely on small firms with no CTO and no product team. Their front office was a phone and a spreadsheet. Now one operator can run the front office of a dozen trades at once.
The adoption numbers agree. The US Chamber found 58 percent of small businesses used generative AI in 2025, up from 40 percent the year before. Among AI-using small businesses, CNBC and SurveyMonkey found 20 percent had already cut headcount, and 35 percent said they might.
That is the quiet version of the story. Not a billionaire founder. A septic pumping site that prices your tank online.
Now the other half, because the first half is the sales deck.
ChartMogul looked at roughly 200 AI-native companies above $250K in annual recurring revenue. Median gross revenue retention was 40 percent. Below $50 a month, products retained just 23 percent of gross revenue. Above $250 a month, retention climbed back to 70 percent. Cheap tools attract tourists who cancel inside a quarter.
Median by price band, roughly 200 companies above $250K ARR. Higher is better.
The product gets cheaper to sell and harder to keep.
The moat did not get cheaper. Bessemer's read on the sector is blunt: foundation-model vendors will absorb generic features, and the products that survive own a workflow, proprietary data, distribution, or a measurable outcome. An API call behind a nice interface is not a business.
Google makes the same point from the other side. Mass-generated pages built mainly to manipulate search rankings are classified as scaled content abuse, and AI-assisted content gets no ranking credit. The specific playbook in that portfolio, generate a page per state and rank, is the exact shape of what Google polices. Durable rankings still come from first-hand evidence and a strong Business Profile.
And someone has to answer the support ticket. AI output is probabilistic, model behavior shifts under you, and every use carries inference cost. A solo founder covering engineering, evaluation, sales, and support has no slack. A survey of more than a thousand solopreneurs found 53 percent reported burnout or emotional struggle.
There is one more ledger, and it belongs in the room.
For the founder, the ten-person team became optional. For the ten people who used to be on it, the team was the ladder. Indeed's August 2025 data shows 30.7 percent of software postings used senior titles against 2 percent junior. New graduates were 7 percent of Big Tech hires in 2024, less than half their 2019 share. Stanford's analysis of payroll data found software developer employment among 22 to 25 year olds down nearly 20 percent from its late-2022 peak. The authors describe the link to AI as suggestive rather than proven, and they are right to, because post-pandemic overhiring and interest rates sit in the same picture.
But the pattern holds across industries. The first rung is the cheapest one to remove.
So the title holds, with a qualifier nobody puts on the T-shirt.
Ten people is optional now because one person can carry the work, and the cost of assembling a team collapsed faster than the cost of competing. What AI removed was headcount. What it left standing was everything that was ever hard: retention, distribution, trust, and the willingness to keep answering the phone after the interesting part is done.
Serkan Inci built on the internet, walked away into physical businesses, and came back when the tools caught up to a solo operator. He is not proof that teams are dead. He is proof that the entry price fell far enough that a 39-year-old ex-marine-engineer with a septic pumping site is now a legitimate data point.
One person can do the work of ten. The ten are still the ones who used to learn how.
Primary source: https://x.com/srkninci/status/2100302884080431397
Serkan Inci's own public account, 17 September 2026. His earlier US trade businesses and the claim that none of the portfolio products needed a ten-person team are self-reported. No revenue figures were given for any site.