How Basecamp Runs a Global Remote Company Without Full-Time Managers

Snapshot

37signals was founded in 1999. Its employee handbook says the company has been remote since it started and is fully distributed across five continents. The communication guide says 98% of internal communication happens in Basecamp and that the company does not use Slack or Teams. In August 2024, 37signals ended its experiment with full-time management roles and returned to having no full-time managers. (Founding, remote model, communication stack, management model)

Jason Fried reported 62 employees in a March 4, 2026 account. This is a newer dated observation, not a verified October headcount. (Source)

The company documents communication, planning and onboarding in public guides. [Inference — these sources make several parts of the operating model inspectable; they do not establish that every internal practice is public or that it is the strongest case in the dataset.]

Founded: 1999 in Chicago by Jason Fried (with Carlos Segura and Ernest Kim) as a web-design firm called 37signals. David Heinemeier Hansson (DHH) joined in 2003 and co-created Basecamp, which launched in February 2004 and eventually became the company’s primary product.
Products: Basecamp (project management), HEY (email), Shape Up (free methodology book).
Structure: Privately held and funded from operating revenue. Jeff Bezos bought a minority stake from the founders in 2006; the founders say the deal supplied them personal liquidity without board control. (Founders’ explanation)

Core Philosophy

37signals believes that calm, focused work is a competitive advantage. Their operating principles follow from that belief:

• Async by default, with exceptions. The handbook recommends written requests first and allows real-time collaboration when async is inefficient. (Handbook)
• Writing over speaking. “Writing solidifies, chat dissolves.” Every important decision, discussion, and direction gets written down — not because they are obsessive documentarians, but because it is the only thing that scales across time zones and future employees. (Source)
• Urgency is a default that should be overridden. “ASAP is poison.” Their communication guide explicitly names the assumption of urgency as toxic to focus and decision quality. (Source)
• Managers of one. Every hire is expected to set their own direction, determine what needs doing, and do it — without waiting to be told. (Source)

Communication Model

Basecamp holds nearly all internal communication. The company says it does not use internal email, Slack or Teams. “98% of our internal communication happens inside Basecamp.” (Source) Zoom is used occasionally for 2–3 person calls. GitHub is used occasionally for pull request review.

Four required async mechanisms keep the company in sync without meetings:

Daily and weekly questions use automatic check-ins. Heartbeats and Kickoffs are separate write-ups assigned to team leads and posted for each cycle. The handbook makes the response and posting requirements explicit; it does not establish that these mechanisms eliminate all-hands or all routine meetings. (Handbook)

Spatial context matters. If you are discussing a task, the discussion lives in the comments under that task. If you are discussing a document, comments attach to the document. Communication does not detach from the thing it is about. This is what makes the “everything in one tool” rule actually workable.

Group chat (Campfire) exists but is treated as a hazard. 37signals published a full guide on why group chat is harmful as a default: it creates false urgency, makes casual thoughts feel like decisions, and interrupts the focus blocks that deep work requires. (Source) Campfire is scoped explicitly to the equivalent of office small talk — “stuff that disappears, and that’s fine.” It is not for decisions, not for project work, not for anything that matters. The anti-chat rule is enforced by culture and policy, not by locking the tool.

The core argument: real-time communication is a tax on the whole company, not just on you. Every message that arrives during focused work is a request for someone to context-switch. The default should be to write something thoughtful and asynchronously, and accept that a response will come hours later, not minutes.

Planning and Cadence

37signals documents six-week cycles followed by two weeks of cooldown, with all teams on the same six-week cadence. Cooldown includes bugs, write-ups and choosing next work; the handbook acknowledges that large projects sometimes extend into it. (Handbook)

The published Shape Up book separates shaping and building into parallel tracks. Shaped pitches feed a betting-table meeting during cooldown. The book describes a designer with one or two programmers, with QA joining later; it also allows several smaller projects within one cycle. These are published method details, not a guarantee that every current team has that exact composition. (Shaping, betting and team sizes)

Appetite limits the investment. Choose the time worth spending and adjust scope within it. The handbook describes six yearly opportunities for major work choices; it does not cap the number of bets or projects at six. (Betting, handbook)

Decision-Making Model

The published Shape Up betting table includes the CEO, CTO, a senior programmer and product strategist. It is a meeting during cooldown, with pitches read beforehand and the CEO described as having the final product decision. This contradicts a blanket claim that important decisions happen only in documents and never in meetings. (Betting table)

The handbook expects self-management and starts requests with written messages, to-dos or documents, while allowing real-time collaboration when async is inefficient. DHH’s August 30, 2024 account says senior staff share management alongside their own work; conflict and persistent performance problems may need escalation to the top. No full-time managers does not mean no management. (Handbook, management account)

Org Structure

• Jason Fried reported 62 employees on March 4, 2026. Preserve this as a dated founder report, not a current census. (Account)
• DHH’s August 30, 2024 account says the company had returned to no full-time managers. Management work remains distributed among senior staff alongside individual contribution. (Account)
• That account assigns one mentee to each lead and principal programmer; it does not establish that every senior employee manages exactly one person or provide a current complete organization chart.
• Automatic questions support status communication; the same evidence does not establish the absence of all PMs, VP titles or management layers across every function.

Tools and Stack

(Sources, handbook, internal systems)

Rituals

In-person meet-ups happen twice per year, one week each, different city every time. Loosely structured. Strategic team sessions, 1:1s, and social exploration. This is where in-person relationships that sustain remote culture get built. (Source)

All Hands happens at the end of each cycle. The whole company meets on Zoom or in person at a meetup. (Source)

Everyone on Support (EOS) — the handbook encourages everyone to do periodic customer-support shifts. The stated purpose: “Talking directly to customers helps us realize what’s wrong and what’s right with our products.” (Source)

Social check-ins are sent automatically every few weeks: “What books are you reading?” “What did you do this weekend?” These are optional but designed to build the connective tissue that in-office teams build by proximity. (Source)

Onboarding

The current handbook describes remote setup through a personalized Basecamp welcome project, with tasks for the newcomer, Ops buddy and manager. A separate 37signals buddy comes from another team and checks in during the first months. The manager meets on day one to set a recurring 1:1 schedule and expectations. This does not establish that the company never uses cohorts or in-person orientation. (Getting Started)

Before starting: the People team orders role-appropriate computer equipment and accessories. The guide specifies Omarchy for Ops, SIP and Ruby product programmers; others may choose Omarchy or Mac. (Guide)

Early work and ramp: the guide says supported real projects begin in the first week and most newcomers feel up to speed and part of the team in about three months. That is a company expectation, not a measured universal 90-day productivity guarantee. [Inference — the transferable design is named support and explicit expectations rather than a fixed number of buddies or a headcount cutoff.]

Benefits and Working Conditions

The living handbook documents the following policies; access was November 1, 2026 and the policy-change dates are not established here.

(Benefits, working hours, hardware)

How They Hire

The handbook expects employees to self-manage; DHH’s 2024 account says the company seeks and develops people who can set their own agenda and work autonomously. Those are stated expectations, not proof that every employee always works independently. (Handbook, founder account)

The sources reviewed here do not establish a current company-wide hiring timeline, absence of headcount targets or a universal paid-trial process for every design and engineering role. Verify a specific current role before presenting its assessment process as policy.

[Inference — application guidance.] Hiring and onboarding should assess the level of self-direction a role requires and provide clear routes for help. The founder’s escalation account suggests a tradeoff for distributed management; it does not establish a quantified cost of a bad hire.

AI

37signals’ 2026 founder accounts describe internal investigations and how AI changed the building of Basecamp 5. They are not measured adoption rates or headcount studies. (January account, July account)

Internal investigations preceded the July development account. In a January 21, 2026 episode, DHH credits Jeremy with an agent-assisted first pass on HackerOne reports, using report history and submitter context to prepare review. He also describes successful agent assistance with on-call and performance investigations across logs and monitoring tools, and discusses applying AI to the company’s biweekly production-access log reviews. These are founder-reported internal workflows; the episode supplies no independently measured time saving, company-wide use rate or universal approval policy. (Source)

Security work includes adversarial checks and production feedback. In an October 5, 2026 engineering account, Mike Dalessio describes building Hot Cell with an agent from written requirements, running adversarial review on almost every commit, and working with Jeremy to have agents attack the system. He says they found real gaps. The system was running in production at 37signals, and outage lessons fed back into the library. This is one engineer’s development and maintenance account, not a workforce adoption rate or proof that agent checks eliminate security risk. (Source)

Basecamp 5 was the first fully AI-accelerated product. Fizzy, launched the previous December, had little direct AI-generated code. Acceleration started around December 2025 and covered about half of Basecamp 5’s development. DHH says that, as of July 2026, a majority of fixes and feature upgrades start with a prompt, except when a developer already knows the change.

Designers can get a feature working before it is allowed into the codebase. Designers, product managers, or anyone with a feature idea may make it work on a branch and show it on a beta with real data. Changes to Ruby or JavaScript from the design side or from a junior programmer need a more senior review before merge. Some prompted code was discarded after it created technical debt. DHH says quality code still takes effort even when producing code is cheap.

A larger menu of work is treated as a reason to say no. Fried compared the current six-week cycle, about 20 candidate items, with maybe six a year earlier. He treated the larger menu as a reason to limit how much of the product moves at once.

Faster implementation can reduce useful challenge. On October 16, 2026, Head of Product Strategy Brian Bailey described how agents let one or two people take an idea to a working pull request much faster. He contrasted this with the feedback accumulated through shaping, betting, design, programming and QA, and asked how to retain that challenge. This is an employee’s unresolved operating question, not evidence that 37signals abolished Shape Up or settled on a replacement process. (Source)

Agents can use Basecamp; native AI features that were not good enough were not shipped. In March 2026, DHH said 18 months of native AI-feature experiments in Fizzy and Basecamp were not shipped. The company launched an API, CLI, and agent skill instead. In the July account, Fried says agents already work in the company’s own Basecamp, and that the product stays usable for people who do not want AI as the front door. (Agent access, July account)

Token spend is described as a subscription, and headcount is not the adjustment. DHH says the company buys frontier-model subscriptions that can run to hundreds of dollars per person per month, rather than metered enterprise token bills. He says 37signals is not changing its size because of AI, and that shipping more features would not linearly increase revenue. That is a founder statement, not a published spend ledger.

Inference: the manager-of-one model now includes a review gate for AI-produced code. The July account does not establish that every employee uses AI, or that the absence of full-time managers caused the acceleration.

What They Do Well

[Inference — interpretation of the documented mechanisms, not measured comparisons.]

• Written visibility: check-in answers and cycle write-ups give colleagues records of progress. This does not mean all tacit knowledge has been captured.
• A place for work choices: cycles and betting make work selection explicit; the sources do not prove there is never unfinished work.
• Protected attention with exceptions: written defaults coexist with calls when useful and with operational on-call work.
• Management responsibilities remain named: self-management and shared mentoring do not remove the need for escalation.

(Handbook, betting table, management account)

Tradeoffs and Weaknesses

Writing is part of the work. [Inference:] recurring questions and cycle write-ups create a real communication obligation. The daily prompt requires answers at least twice weekly, not every day. (Handbook)

Conflict and poor performance still need escalation. DHH’s 2024 account identifies those as the hardest management burden to distribute. He cautions that the arrangement could be difficult at hundreds or thousands of employees; it does not establish a failure threshold at 80, 100 or 200. (Account)

Cross-time-zone norms are incompletely documented in this profile. The written default can reduce immediate-response pressure, but the reviewed sources do not establish every team’s overlap or escalation arrangements.

Lessons Learned: What They Tried and Changed

Full-time management: on August 30, 2024, DHH described returning to no full-time managers after experimenting with such roles over the previous few years. He explained the preference for management alongside individual contribution and acknowledged the difficulty of conflict and performance work. The source does not precisely establish a 2021 start or the earlier profile’s detailed claims about management-track motives and worsened decision quality. (Account)

Political discussion: Jason Fried’s April 26, 2021 announcement ended societal and political discussion on the company Basecamp account. The Verge reported substantial departures afterward. Keep the employer’s reasoning separate from employees’ responses; the announcement does not prove those discussions produced no value or caused every departure. (Primary announcement, contemporaneous reporting)

Planning evolved into Shape Up: the book explains the reasoning behind shaping, betting and time-limited building. Its published method is useful evidence without asserting that the company tried and rejected every mainstream planning system. (Shape Up)

What Founders Can Copy

[Inference — design options from the sources, not measured savings or validated headcount ranges.]

1. Specify recurring written updates. Distinguish automated daily/weekly prompts from manually written cycle posts.
2. Choose a commitment period. 37signals uses six-week cycles and cooldown; other teams should evaluate a cadence against their work.
3. Assign mentoring work. The founder’s account gives lead and principal programmers one mentee alongside their own work; it does not establish an 80-person ceiling.
4. Make work selection explicit. A betting table is one mechanism, with shaped choices and people authorized to decide.
5. Set an appetite. Decide the investment before building, then use scope as the variable.
6. Define where work context lives. Basecamp’s near-complete communication consolidation is an example, not proof every specialist tool should disappear.

(Handbook, betting, management, communication)

Where This Model Breaks

[Inference — risks to inspect, not verified failure thresholds.]

• Growth can require more explicit management and escalation; the founder cautions about grand scale without a precise cutoff.
• Persistent performance or conflict cases require someone with authority and time to respond.
• Live operational work needs coverage and escalation alongside an async default; the handbook explicitly acknowledges on-call work.
• New teams need clear expectations and relationship-building rather than an assumption that self-direction will appear without support.

Related Practices

• Async Communication — the foundation that makes everything here possible
• Planning Cycles — how structured cadences replace perpetual roadmap management
• Documentation Systems — the written culture that makes async durable
• Decision Ownership — how “managers of one” distribute authority

Related Frameworks

• Sync vs Async Decision Matrix — when to use real-time vs written communication

Sources

37signals — Benefits and Perks (living policy, accessed November 1, 2026): https://basecamp.com/handbook/benefits-and-perks

Shape Up — Principles of Shaping (parallel shaping and building tracks): https://basecamp.com/shapeup/1.1-chapter-02

Shape Up — The Betting Table (method and cycle selection): https://basecamp.com/shapeup/2.2-chapter-08

1. 37signals Employee Handbook: How We Work — https://basecamp.com/handbook/how-we-work
2. The 37signals Guide to Internal Communication — https://basecamp.com/guides/how-we-communicate
3. Our Rituals — 37signals Employee Handbook — https://basecamp.com/handbook/our-rituals
4. Getting Started — 37signals Employee Handbook — https://books.37signals.com/3/the-37signals-employee-handbook/7/getting-started
5. DHH: We once more have no full-time managers at 37signals — https://world.hey.com/dhh/we-once-more-have-no-full-time-managers-at-37signals-f8611085
6. Shape Up (37signals) — https://basecamp.com/shapeup
7. Historical background only — Jason Fried: Why I Run a Flat Company (Inc., April 2011): https://www.inc.com/magazine/20110401/jason-fried-why-i-run-a-flat-company.html — current management claims use the separately cited August 2024 founder account; direct retrieval of this older interview was restricted during the November 1 review.
8. Group Chat: The Best Way to Totally Stress Out Your Team — https://basecamp.com/guides/group-chat-problems
9. Jason Fried, “Changes at Basecamp” (April 26, 2021) — primary source for the political-discussion policy change: https://world.hey.com/jason/changes-at-basecamp-7f32afc5
10. The Verge, “Basecamp implodes as employees flee” (April 30, 2021) — strong-secondary for the ~33% departures following the policy change: https://www.theverge.com/2021/4/30/22412714/basecamp-employees-memo-policy-controversy
11. REWORK, Outside Money is Plan Z (December 7, 2021) — founders distinguish the 2006 secondary stake sale from business funding: https://37signals.com/podcast/outside-money-is-plan-z/
12. Bezos Expeditions invests in 37signals (July 20, 2006): https://signalvnoise.com/archives2/bezos_expeditions_invests_in_37signals
13. REWORK, One-way Doors (August 26, 2026) — reaffirmation of the historical sale: https://37signals.com/podcast/one-way-doors/
14. REWORK, AI challenges in software development (July 1, 2026): https://37signals.com/podcast/ai-challenges-in-software/
15. DHH, Basecamp becomes agent accessible (March 25, 2026): https://world.hey.com/dhh/basecamp-becomes-agent-accessible-3ae6b949
16. REWORK, AI Revisited (January 21, 2026): https://37signals.com/podcast/ai-revisited/
17. REWORK, AI Revisited – Part 2 (March 4, 2026): https://37signals.com/podcast/ai-revisited-part-2/

37signals — Our Internal Systems (undated living handbook, accessed November 1, 2026): https://basecamp.com/handbook/our-internal-systems

Mike Dalessio — Hot Cell v1.0 (October 5, 2026): https://dev.37signals.com/hot-cell-1-0/

Brian Bailey — Slow burn (October 16, 2026): https://world.hey.com/bb/slow-burn-044c1416

Inferences

• The total written obligation across all required async mechanisms (daily check-ins, weekly previews, heartbeats, kickoffs, in-context project documentation) is substantial. The “fewer meetings = more freedom” framing is accurate but incomplete — the freedom is purchased with disciplined documentation culture. Companies that adopt the tools without the writing discipline will get the overhead without the benefit.
• The twice-yearly in-person meet-ups appear to do more cultural maintenance work than 37signals explicitly claims. Purely async remote teams that skip synchronous time tend to lose trust calibration. The meet-up cadence is likely load-bearing for the manager-less model.
• The two-week cooldown reserves reflection time after each six-week build period. This gives reflection a regular place in the calendar; it does not mean the combined sequence repeats every six calendar weeks.
• The July 2026 development account adds a review step to a manager-less company: prompted work can prove an idea, and a senior still decides what enters the codebase. It does not show that AI removed the need for that judgment, or that a larger cycle menu should be shipped in full.

Work with Alex

If your company is running into the coordination problems this operating system is designed to solve — fragmented communication, missed commitments, unclear ownership, meetings that replace decisions — Alex helps leadership teams design the operating system that allows execution to scale.