Enterprise Account Executive

Honeycomb releases second edition of bestseller Observability Engineering to redefine the practice for an AI world. New insights include why most companies cannot safely validate AI-generated code in production, how shipping faster leads to organizations learning slower, and more. | June 17, 2026 Whitepapers March 4, 2026 The Director's Guide to the Future of Observability: AI, OpenTelemetry, and Complex Systems. SAN FRANCISCO - June 17, 2026 - Honeycomb.io, the observability platform for the new shape of software, today announced the publication of Observability Engineering: Achieving Production Excellence, 2nd Edition by Charity Majors, Liz Fong-Jones, and George Miranda with Austin Parker. Majors is Co-founder and CTO of Honeycomb, Fong-Jones is Honeycomb's Technical Fellow, and Parker is Honeycomb's Director of AI Strategy. The book is produced in collaboration with O'Reilly and was almost entirely rewritten to reflect the new challenges facing today's engineering teams. "The core argument hasn't changed from the first edition: fast feedback loops are the beating heart of every high-performing engineering organization, and observability is what makes them possible," said Charity Majors. "But the stakes have gotten a lot higher. Most teams are shipping faster than ever, accelerated by AI, yet their production feedback loops haven't kept pace. This means they're not learning faster, they're just accumulating risk faster. That tension is why we wrote this second edition. The only way to close that gap, turning production signals into understanding at the speed of AI-assisted development, is observability." The Foundations of Observability in the Age of AI The first edition, published in 2022, established the principles of modern observability for distributed systems. This second edition is not a revision of that book, but rather a near-complete rewrite, as the assumptions the first edition was built on have changed. The multi-stage software development lifecycle that organized the first edition's argument - write, test, deploy, observe - is compressing into rapid loops of intent and validation, with most of what used to live in pre-production now happening live. The book had to change because the world it described no longer exists. At more than 600 pages, nearly twice the length of the first edition, the second edition extends the same foundational principles across the full scope of modern software engineering: instrumentation for AI-assisted development, debugging LLM-powered applications in production, telemetry pipeline management, ontologies as a shared language for humans and agents, organizational learning speed as a competitive constraint, and the strategic and financial decisions facing engineering teams navigating the shift. "The core questions haven't changed since the first edition: what is your system actually doing, why, and does your code do what you think it does in production? What's changed is how much of that code humans wrote," said Liz Fong-Jones. "When agents generate most of your diffs, you can't validate by reading every line; the proof has to come from production telemetry. So that's what we wrote: instrumentation for AI-assisted development, debugging LLM applications live in production, telemetry pipeline management, and the organizational decisions that follow from all of it. Martin Fowler told us to make the book shorter. We made it nearly twice as long, because that's how much the field has grown." The latest edition features contributed chapters from practitioners at the frontier of this work, including Boris Tane of Polylane on why observability agents succeed or fail based on context quality; Phillip Carter on building continuous improvement loops for LLM applications using production telemetry; Kesha Mykhailov and Darragh Curran from Fin on the organizational and engineering realities of operating AI at scale; and Hazel Weakly, an Architect at ING and Fellow at the Nivenly Foundation, wrote the foreword and contributed guidance on instrumentation for regulated environments. The ClickHouse engineering team also contributed a chapter that provides an in-depth look at how their open-source datastore is architected and tuned specifically for observability workloads. Meeting the Moment and Continuing the Conversation The book's publication coincides with a period of significant product development at Honeycomb. In May, the company launched Agent Timeline, giving engineering teams full visibility into agentic workflows in production, alongside a redesigned Canvas investigation workspace, all built on the same high-cardinality, high-dimensionality foundation the book describes. The thesis of the second edition and the direction of the product are, at this point, the same argument. Early-release editions of the book were shared with practitioners at O11yCon San Francisco in May and LDX3 London earlier this month, where Liz Fong-Jones keynoted and signed copies ahead of the full release. Charity Majors will keynote LDX3 New York in September. Visit the Honeycomb booth to get a signed copy. Additional author appearances will be announced at https://www.honeycomb.io/events. Observability Engineering, 2nd Edition is available now through O'Reilly Media in print and digital formats. Get your free copy of the book here: https://www.honeycomb.io/observability-engineering-oreilly-book. About Honeycomb Honeycomb is the observability platform for the new shape of software. Built on a decade of distributed tracing leadership and deep roots in the OpenTelemetry community, Honeycomb gives engineering teams - and the AI agents now operating alongside them - real-time, high-cardinality answers about any production system, with no pre-aggregation and no cardinality limits. Learn more at www.honeycomb.io and follow Honeycomb on LinkedIn. Media Contact Ciri Haugh [email protected]

Honeycomb and Embrace have launched a strategic partnership integrating Embrace's real user monitoring for mobile and web into Honeycomb's observability platform. The integration enables engineering teams to correlate frontend performance data with backend system health, addressing a longstanding gap where backend teams lacked visibility into user experiences whilst frontend developers used disparate tools. Built on OpenTelemetry, the partnership allows session data, crash signals, network insights and Core Web Vitals from mobile and web applications to flow directly into Honeycomb. This creates a unified view across the entire stack using a shared data standard. The integration is available through AWS Marketplace. Embrace customers include AllTrails, Best Buy, Ford and The New York Times. The company has raised Series B funding from NEA, Allianz, Greycroft and Eniac.

Embrace partners with Honeycomb to unite frontend performance depth with backend observability precision. partnership Here's the big news: embrace.io has launched a strategic partnership with Honeycomb to bring web and mobile RUM to the industry-leading services and infrastructure reliability platform its friends at Honeycomb are building. The reasons go deeper than the product. embrace.io built Embrace on OpenTelemetry because embrace.io believe the future of observability is open, composable, and high-fidelity. embrace.io has always believed that frontend and mobile performance and reliability aren't a separate discipline from observability, they're just the part of the system that legacy platforms didn't reach. Honeycomb has pioneered the same philosophy embrace.io has been building on at Embrace on the backend: that high-cardinality, context-rich data lets engineering teams understand systems in ways that weren't previously possible. That shared foundation, around high-fidelity data and OpenTelemetry, makes this partnership much more than a typical integration, and will help entire engineering organizations better understand highly complex systems and deliver better experiences for every user on every screen. Scattered tooling and separate workflows. Backend and reliability teams have had strong observability options for a while now. Honeycomb has led the pack with a best-in-class high-cardinality model for services observability, and a genuine commitment to OpenTelemetry. But engineers across teams kept telling embrace.io the same thing: they wanted to know what end-users were experiencing on the other side of the API call. Mobile and web developers didn't want shallow RUM tools bolted onto an expensive "all-in-one" platform that wasn't built for them, so they'd seek out point solutions with zero ties to the reliability teams. Client-side environments like mobile apps are notoriously complex to troubleshoot. The result is predictable: no shared data, duplicate investigations, and a structural disconnect between the teams who need to collaborate the most. The pain isn't subtle. An SRE gets paged at 9am on a Monday, narrows a latency spike to a specific endpoint, and then has to stop and ask: is anyone actually feeling this? They have no way to know without calling someone on the mobile team. Or, a mobile engineer sees an ANR spike on the checkout screen and spends twenty minutes trying to rule out a client bug before they can even ask for backend help. A frontend engineer watches a Core Web Vitals metric deteriorate for three weeks because it never crossed the threshold that would have paged someone (yet the users felt it). These aren't edge cases. They're the standard experience for engineering teams working on user-facing software. And they all share the same problem: the ability to get answers stops at the API boundary. What changes when the data is connected and tooling depth matches expertise. Because both platforms are built on OpenTelemetry, user telemetry from Embrace flows into Honeycomb as standard, enriched, composable data with the same format and same dimensions, queryable with the same tools. No proprietary schema, no translation layer. In practice, this changes how three kinds of engineers work: * The SRE on-call starts in Honeycomb and ends in Embrace. When a backend issue fires, Honeycomb narrows the problem to a specific service and traffic slice. From there, one click surfaces the affected real user sessions in Embrace, including who they are, how many, and what they were doing when things went wrong. That's usually enough to decide whether it's worth waking up the frontend engineering team. Honeycomb tells them the "where," Embrace tells them the "who," "how many," "what customer activities are affected," and "how bad." * An app engineer starts in Embrace and ends in Honeycomb. When a checkout flow in the app is freezing (and frustrating users), Embrace quickly shows whether the problem is on the device or somewhere upstream. If it's upstream, one click opens the matching backend trace in Honeycomb, and the engineer can hand off a specific, evidence-backed ticket instead of a hunch. * The web performance / frontend engineer loops between both. Honeycomb catches a Core Web Vitals metric drifting in the wrong direction before it crosses an alert threshold. Embrace gives per-page and per-interaction breakdowns along with Synthetic analysis. They identify a recently-shipped vendor SDK as the cause, ship a fix, and can confirm the improvement in Honeycomb within a day. What this means for composable observability, and for its AI-driven future. The handoff points work because the data is open and the philosophy is shared. The session ID in a Honeycomb span drives into Embrace where users can explore populations of affected users, and the trace ID in an Embrace network span is a link to Honeycomb where you can explore the dynamics of service behavior. The data doesn't stop at the API boundary anymore. Today's integration is the foundation, not the destination. Where embrace.io is headed is a single observability surface where any question about a service, a user, a revenue line, or a cohort on a specific device build has one place to live. That means BubbleUp running on the joint event set, SLOs that incorporate real user experience scores, and web and mobile apps treated as first-class services. Drift on either side gets cross-checked automatically before any threshold fires, and the impact gets surfaced in language that product and leadership can actually act on, not just the engineering team. Engineering teams worry less about the tooling and more about their end-users. That also matters for where AI-driven development is heading. When an AI agent can pull backend trace data from Honeycomb and jump directly to the affected user sessions in Embrace, it can traverse the full stack in one step and surface the context that tells a developer whether to page the team or close the ticket. When Honeycomb's querying capabilities and Embrace's MCP server are working together inside an agentic workflow, the on-call engineer supervises rather than assembles. The quality of that agent's answer is only as good as the data it can reach, of course. Connected data means connected reasoning, and embrace.io'll have a lot more to share on that soon. Engineering teams that have been treating web and mobile performance as separate from their observability practice can now bring it into the same workflow, the same data, and the same tools. That's the promise, and embrace.io is excited to deliver it with Honeycomb.