Open source trust: how Bloomberg embedded engineers in OpenTelemetry

What happened

Bloomberg placed 48 of its engineers inside the OpenTelemetry project for ten weeks earlier this year, with seven project maintainers acting as mentors. Diginomica reports that the arrangement was designed as a practical response to a growing strain on open source contribution, and that the maintainer leading the OpenTelemetry side of the programme now wants to see far more of this kind of collaboration.

The maintainer's case rests on the effect that AI tools are having on how contributions arrive and how they are judged. Diginomica's headline frames the wider problem plainly: AI is making open source contribution harder to trust. The embedded model was positioned as a way of rebuilding that confidence through sustained, person-to-person working relationships rather than through unsolicited changes arriving from outside.

The source material available for this briefing is limited to the programme summary and headline, so details such as the specific work completed, the tooling used and the outcomes measured are not covered here.

Why it matters

For organisations that rely on open source software, the story points to a shift in how trust between contributors and maintainers is built and maintained. When AI lowers the cost of producing code, the expensive part becomes verifying that a change is sound, understood and maintainable. An embedded programme puts that verification work where it can be done well: inside the project, with context and mentorship attached.

For leaders in digital transformation, the example is a useful reminder that technology adoption changes working relationships as well as workflows. Teams that consume and contribute to shared infrastructure will need new ways of working together if AI is to speed delivery without eroding the standards that make shared platforms dependable.

By the numbers

• 48 Bloomberg engineers embedded in the OpenTelemetry project
• 10 weeks duration of the embedded programme
• 7 OpenTelemetry maintainers acting as mentors

The Renascence take

Most commentary on AI and open source will focus on volume: more pull requests, more code, more noise for maintainers to sift. The more useful question is what happens to trust when the effort of producing a contribution no longer signals its quality. Treating that as a tooling problem misses the point; it is a relationship problem, and relationships are built through shared time and context.

The behavioural principle underneath this programme is that trust is earned through visible, repeated interaction rather than through the quality of a single output. Engineers who have spent ten weeks inside a project understand its norms, its history and its reviewers, and that understanding is what AI cannot shortcut. Any operator that depends on shared platforms should ask how it is investing in those relationships, not only in the tools that generate the work. Embedding people is slower and costlier than adding a plug-in, which is precisely why it is worth doing deliberately.