LLenny's Podcast
← All frameworks
LeadershipKeith Coleman (VP of Product) and Jay Baxter (ML Lead)

The Thermal Team

Run a startup inside the company: one owner, one senior sponsor, total focus, self-set pace.

Difficulty
Moderate
Time to result
~months to results
Steps
5
Confidence
95%

A 'thermal' team is an internal team deliberately insulated from corporate process so it can move like a startup. It has one clear founder-like driver, one senior decision-maker outside the team, 100% dedicated members, no mandated planning rituals (OKRs, Jira), and it sets its own milestone-based pace. Coleman credits this structure as a reason Community Notes exists at all.

Origin

Coined at Twitter by co-founder Kavon Beykpour (referenced as 'Kavon') as a program for isolated, startup-like teams; the bird/'thermal' name is a Twitter bird-analogy for the thermals that lift birds in flight. Coleman made Community Notes the first thermal project circa 2020.

Core principles

  • 01More people managed does not equal more impact — the myth that scope equals impact gets in leaders' way.
  • 02A one-to-two-week delay from a distracted collaborator kills a project's momentum.
  • 03Quarterly/annual goal cadences are arbitrary; set goals at the pace the work actually moves.
  • 04Self-selection beats assignment — everyone should opt in and be interviewed for fit.

How to run it

  1. 1

    Appoint one founder-like driver

    Name a single clear owner of the project who operates like its founder, plus one clear senior decision-maker outside the team they escalate to.

    Pro tip The external decision-maker must be senior enough to make whatever call you need, so decisions happen in one hop.

  2. 2

    Demand 100% focus

    Every member is fully dedicated to the project — no split attention across other teams. Total focus is what lets iteration cycles collapse from weeks to a day.

    Pro tip With full focus you can propose an algorithm change in the morning and review results that afternoon or the next day.

    Watch out Borrowing part-time help from other teams reintroduces the one-to-two-week delays that break momentum.

  3. 3

    Keep it tiny — one of each function

    Start as small as possible. Community Notes launched build with ~5: backend, frontend, ML/scoring, design, and a researcher to understand the customer.

  4. 4

    Let people self-select in

    Don't assign anyone. Let people apply or reach out, and interview every member so both sides are fully bought into the mission and each other.

    Pro tip Recruit internally the way you'd recruit for an external startup — an exciting vision is the draw.

    Watch out Assigning random people to a crazy new project makes success much harder.

  5. 5

    Set your own milestone-based pace

    Skip OKRs and heavyweight task tools. Set the goal for the next milestone that matters, hit it, then set the next — whether that's two weeks or three months. Coordinate in one lightweight running doc reviewed in a daily meeting.

    Pro tip A long-running Google Doc where irrelevant items just fall off beats a backlog you have to groom.

    Watch out The OKR planning process itself can take longer than picking a goal and finishing it.

In the wild

Daily doc, dynamic roadmap

The team runs off one very long Google Doc (long enough it periodically broke Google Docs) and a daily meeting, setting goals dynamically around 'what helps people most' or 'the biggest problem right now.' They may change the roadmap multiple times in two weeks.

Shipped every week through the entire acquisition period without getting distracted; Coleman doubts the product would exist without this structure.

Scaling Spaces in weeks, not a year

After the acquisition Elon asked a tiny team to scale Spaces to host huge audiences. At Twitter 1.0 this would have taken a year if it happened at all.

The small, motivated team delivered it in two to three weeks.

Common mistakes

Equating headcount with impact

Chasing a bigger org or wider scope for career reasons produces '16 more pages of OKRs' rather than real output; a small focused team can have far bigger impact.

Imposing standard corporate process on the team

Mandating OKRs, Jira, and quarterly planning slows a zero-to-one team; the planning overhead can exceed the time to just pick and finish a goal.

Is it for you?

Best for

A senior leader who can secure autonomy and a senior sponsor, launching an ambitious zero-to-one product inside a larger company.

Not ideal for

Mature products needing broad cross-team coordination, heavy compliance, or predictable quarterly commitments to stakeholders.

From the transcript

there's one clear driver of the project who's effectively like founder

42:00

also there's one clear decision maker that they go to oh outside of the team

42:30

another was 100% Focus so everyone on the project is expected to be totally focused on it

43:00

people are self- selecting to join we did not assign anyone to this project

56:30

we set our own pace and and goals at that pace

44:30

From the episode

An inside look at X’s Community Notes

Keith Coleman (VP of Product) and Jay Baxter (ML Lead)