AI Engineer World's Fair 2025 — Tiny Teams: How Ultra-Small Teams Dominate the AI Era

Overview

This session at the AI Engineer Summit showcases how ultra-small teams (around 10 people or fewer) achieve outsized results in the AI era through fast execution, high context, AI leverage, and community power.

1. Opening (Britney Walker, CRV)

CRV has operated 19 funds over 55 years, investing in early-stage infrastructure and AI startups.
"The era where a handful of people create projects at a scale previously unimaginable is here."

2. StackBlitz & Bolt.net: 7 Years of Craft, Under 20 People (Eric Simons)

Bolt launched with under 20 team members after 7 years of building StackBlitz with limited commercial success. "When we launched Bolt, we were actually preparing to shut down."
The initial product was a "racing car with all the seats stripped, bare metal MVP."
"Fewer people, more context, greater ownership." — Each team member holds more context, enabling faster decisions without permissions.
"The most important thing at a startup is maximizing 'shots on goal.' More people means higher burn rate and fewer opportunities."
AI tools handle 90% of support queries. "That's work we'd need 50 people for without AI."
"Don't recruit an army. Recruit Spartan elites."

3. Oliv: 4 People, $6M ARR (Sid Bendre)

Just 4 people achieved $6M ARR, 500M views, 1M users.
Three principles: "Hire great or don't hire at all," "Profit first," and relentless process improvement.
Organizational structure: "Harvesters" (full-stack product owners) and "Cultivators" (AI infrastructure / automation).
AI turns 10x engineers into 100x engineers.

4. Gum Loop: 2 to 9 People, Enterprise Customers (Max Broer Herbas)

Post-YC, grew from 2 to 9 people with enterprise customers like Instacart, Webflow, and Shopify.
"Super picky" hiring: "If it's not a clear 'this is the one,' we don't hire."
Product-led hiring and work trials (hacking together in an Airbnb for several days).
Almost no meetings: "We hired great people, so we let them build."

5. Gamma: 30 People, 50M Users (Grant)

Built an "anti-PowerPoint" AI presentation tool with ~30 people serving 50 million users over 4 years.
Three design principles: Generalists, Player-Coach model (leaders who also code), Brand and culture as competitive advantages.
"Why pursue a solo billion-dollar startup? Building as a team is way more fun!"

6. Data Lab: 3 People, 7-Figure ARR (Vic Paruturi)

3 (now 4) people: 40K GitHub stars, 7-figure ARR, training cutting-edge models.
Four productivity secrets: No specialist overload, no excessive process/meetings, hire only senior generalists, use AI and internal tools to fill gaps.
"A 500M parameter, 90-language, 99% accurate OCR model was built end-to-end by just 2 people."
"More people doesn't mean more productivity!"

7. EveryY: Benchmarks Are Memes (Alex Duffy)

"Benchmarks spread like memes and determine AI's direction."
The lifecycle: Someone has an idea → it spreads → model providers train on it → saturation → new benchmark emerges.
"The benchmarks you create could drive AI's development for the next 5 years."

Conclusion

Small teams, high context, fast execution, AI and community power — proven across multiple real-world cases.
Generalists, low ego, high trust, user obsession, and execution are essential.
Culture, brand, and community investments matter.
Those who create benchmarks define AI's future.

Key Terms: Tiny Teams, Generalist, Profit First, Player-Coach, AI Automation, Community, Culture, PLG, Super Tools, Benchmark as Meme, Execution, Low Ego High Trust, Rapid Feedback Loop, Hiring Bar