Skip to main content

the ai activation flywheel

the competition is literally building games to upskill themselves. if you're not doing this, you're already behind.

what ramp proved​

in april 2026, Ramp (a financial payments infrastructure company, 700 employees) published how they built an internal AI suite called Glass. the numbers:

  • AI usage up 6,300% year over year
  • 99.5% of employees active on AI tools
  • 84% using coding agents weekly
  • 1,500+ apps shipped in six weeks from 800+ builders
  • non-engineers doing 12% of all production pull requests
  • 350+ shared skills in an internal marketplace called the Dojo

they built this with a team of four in under three months. then they gamified the whole thing: leaderboards tracking every team and individual, visible to everyone. competitive dynamics that did more for adoption than any training program ever could.

the key insight: the people who got the most value were not the ones who attended training sessions. they were the ones who installed a skill on day one and got a result. the product taught them faster than any workshop.

why this matters for us​

this is the new competitive bar. not for Ramp specifically, but for every organization in the economy. there are now two kinds of companies:

  1. companies where every employee has an AI partner, shares skills, competes on leaderboards, and compounds each other's capabilities daily
  2. companies where people are still manually robot-moding through their work

the gap between these two types of companies is widening every week. and it's not closing.

the flywheel pattern​

the pattern Ramp validated is what we're calling the AI Activation Flywheel:

  1. someone discovers a useful AI workflow
  2. they package it as a reusable skill (markdown instruction file)
  3. the organization distributes it (marketplace, recommendations)
  4. everyone who installs it levels up
  5. from the higher baseline, new discoveries happen
  6. repeat

each cycle raises the floor for the entire organization. the flywheel spins faster with every cycle because each iteration starts from a higher baseline.

this is not a technology pattern. it is a cultural pattern enabled by technology. the leaderboards, the celebrations, the competitive dynamics, the peer recognition: these are what make the flywheel spin. without the culture, you have a skill library nobody uses.

what this means for Travis and OpenTeams​

we've been advising Travis on what to build (Project Meta-Flywheel, documented separately). the Ramp case study crystallized the thesis:

step one: apply it to your own organization first. before selling anything to anyone, OpenTeams needs to run the flywheel internally. shared skills, leaderboards, the full game. if you can't demonstrate the transformation on your own team, the product pitch is hollow.

step two: productize the playbook. most companies cannot do what Ramp did (dedicate four engineers for three months to build a custom internal AI suite). the opportunity is packaging the Ramp playbook as open-source, sovereign infrastructure that any company can install. the product category is "AI Activation Flywheel as a Service."

step three: the meta. this is why we call it Project Meta-Flywheel. OpenTeams' growth engine is engineering flywheels for other companies. but the meta is that OpenTeams runs the same flywheel internally. the act of transforming OpenTeams IS the product development, IS the first case study, IS the strongest sales proof. build it for yourself. prove it works. install it everywhere else.

the race​

someone is going to productize this. could be Ramp spinning up a division. could be Anthropic building "Claude Code for Teams." could be a well-funded startup. could be OpenTeams.

the open-source, sovereign version wins because companies do not want their organizational intelligence living on someone else's servers. Travis built NumPy and SciPy. the Python community trusts him. the "Linux of organizational AI" is his to build if he moves now.

implications for Imagos portfolio companies​

every Pegasus team we incubate needs an internal flywheel. not eventually. from day one. the skill marketplace, the shared context, the adoption dynamics. it should be part of the standard Imagos operating infrastructure that every portfolio company gets.

if Nebari becomes the open-source standard for this, Imagos portfolio companies are early adopters and case studies. the network effects compound: skills built for one portfolio company can be shared across the ecosystem (with proper scoping). the Imagos network becomes a flywheel of flywheels.