COMPANY PLAYBOOK
Sourcegraph
System principles
- Make the person doing the work the default decision owner.
- Cut scope, not quality.
- Own customer value, not just output.
Tensions and trade-offs
- High agency shortens decision distance but can pull teams apart without shared context.
- Early shipping accelerates learning but temporary solutions can become permanent debt.
PRACTICES
How Sourcegraph builds its operating system
The doer owns the decision
The default owner of a technical decision is the person or team that owns the work it implies.
- How it works
- Decision authority starts with the person doing the work; unresolved disagreement uses a clean escalation.
- Why it is used
- Keeps decisions close to information and reduces waiting.
- Trade-off
- An owner who gathers too little context can optimize locally.
- Best fit
- Teams with explicit accountability and company goals.
Smallest useful release
A product is made useful first, then made smooth, fast and scalable.
- How it works
- Under uncertainty, a narrow scope that can generate feedback ships first.
- Why it is used
- Validates the solution before making a large investment in the wrong direction.
- Trade-off
- Temporary shortcuts need visible follow-up and cleanup.
- Best fit
- Reversible product decisions and early discovery.
Move async context into writing
Decision and work context stays written and accessible for time-zone-independent participation.
- How it works
- Processes, principles and decisions are made durable in shared handbook surfaces.
- Why it is used
- Reduces oral-history and visibility inequality in a remote team.
- Trade-off
- Reading load and document freshness become operational work of their own.
- Best fit
- Teams with strong writing across multiple time zones.
Weekly managed-cloud updates with monthly self-hosted releases
A 5 January 2026 announcement schedules weekly Sourcegraph Cloud updates from 25 February 2026 while self-hosted releases remain monthly; it says each cloud update retains a changelog, test/QA and staggered rollout.
- How it works
- Company-managed cloud and customer-operated installations use different release frequencies according to their control boundaries.
- Why it is used
- Allows smaller, more frequent managed-cloud changes while giving self-hosted customers planning time.
- Trade-off
- Weekly releases depend on sustained testing and rollout discipline.
- Best fit
- Products offered as both managed cloud and customer-operated installations.
Let an agent pre-triage security reports first
Sourcegraph’s security team receives every new vulnerability report from its bug bounty platform through a signed webhook to a small service, which verifies the signature and passes the report to the Deep Search agent to investigate it in the relevant codebase. The agent judges validity, severity and affected code, suggests fixes, and posts the result as a team-only comment on the report; the engineer finds the investigation done in the morning and can fork the conversation to ask follow-ups.
- How it works
- Every new report goes to the agent automatically; its findings (validity, severity, code links, suggested fixes) are posted as a team-only comment and human review continues from there.
- Why it is used
- To cut the hours spent manually validating raw reports and to shorten response time; the first read happens in the machine while review and remediation stay with people.
- Trade-off
- The author names three risks: reporters consuming the agent’s usage quota (limited by entitlements and an invite-only program), wrong repository matches (service-to-repo mapping written into the prompt) and prompt injection (entitlements, provider defenses and untrusted-input tags). The gain is described as “hours”, not quantified.
- Best fit
- Teams with a heavy, repetitive inbound report stream where first triage is mostly reading; portability is limited because Sourcegraph uses its own product.
Learn continuously through retrospectives
Teams review completed work, process and collaboration through retrospectives and turn lessons into improvement actions.
- How it works
- Teams review completed work, process and collaboration through retrospectives and turn lessons into improvement actions.
- Why it is used
- Makes the practice repeatable and visible.
- Trade-off
- It requires coordination and maintenance capacity.
- Best fit
- Teams with explicit responsibilities.
Use 1:1s as a distinct communication system
Every manager holds a weekly recurring 25-minute 1:1 with each direct report, using a private shared notes document containing the agenda and goals.
- How it works
- Every manager holds a weekly recurring 25-minute 1:1 with each direct report, using a private shared notes document containing the agenda and goals.
- Why it is used
- Makes the practice repeatable and visible.
- Trade-off
- It requires coordination and maintenance capacity.
- Best fit
- Teams with explicit responsibilities.
Spend a yearly development allowance without approval
Eligible education and professional-development spend within the USD 1,000 annual allowance requires no prior Finance or People Ops approval.
- How it works
- Eligible education and professional-development spend within the USD 1,000 annual allowance requires no prior Finance or People Ops approval.
- Why it is used
- Makes the practice repeatable and visible.
- Trade-off
- It requires coordination and maintenance capacity.
- Best fit
- Teams with explicit responsibilities.
Sixteen weeks of flexible paid parental leave
Paid parental leave provides 16 weeks, usable consecutively or in non-consecutive monthly blocks within 12 months of the initial leave date.
- How it works
- Paid parental leave provides 16 weeks, usable consecutively or in non-consecutive monthly blocks within 12 months of the initial leave date.
- Why it is used
- Makes the practice repeatable and visible.
- Trade-off
- It requires coordination and maintenance capacity.
- Best fit
- Teams with explicit responsibilities.