@honeycomb.io
Honeycomb is the observability platform built for the new shape of software.
@honeycomb.ioAI mandates have had a mixed track record. Most engineering leaders know something isn’t working, but don’t know what to do about it.
Charity Majors (@charity.wtf) and Dr. Cat Hicks are doing a live AMA on exactly that. go.hny.co/47dTv4L
@honeycomb.ioNew from Honeycomb: Canvas Connectors for code, incidents, runbooks and tickets. AI Ecosystem for fleet-wide agent performance and cost. Anomaly Detection now GA. And onboarding via MCP from your coding agent.
honeycomb.io/blog/agents-need-context-canvas-connectors-ai-agent-visibility
@lizthegrey.comReposted by @honeycomb.ioYour SLOs and telemetry are working. Now leadership wants to know why the observability bill went up. My last session was about designing SLOs that mean something. This week's Masterclass is about the bill that follows, and most of it comes down to inefficient sampling.
@honeycomb.ioToday, we released a series of new features including AI Ecosystem, which provides a fleet-wide view of your AI agents’ performance and cost. We also launched 12 Canvas MCP Connectors to offer even more context beyond your telemetry data.
See our announcement: www.honeycomb.io/blog/honeyco...
@honeycomb.ioA fixed 1-in-100 sample rate keeps 1% of your errors, 1% of rare routes, and 1% of health checks, all at the same rate.
The traces you need during an incident are the most likely to be gone.
How adaptive tail sampling fixes that: honeycomb.io/blog/how-adaptive-tail-sampling-works
@honeycomb.ioWatch the full conversation of the first episode of Leading With Observability here. www.honeycomb.io/blog/fin-cto...
@honeycomb.ioMost teams building AI agents are still in reaction mode.
A real feedback loop: Instrument → investigate → fix → prove it worked
Canvas powers the whole loop. We know because we used it to build Canvas.
Full walkthrough: honeycomb.io/blog/how-canvas-powers-ai-agent-development-feedback-loop
@lizthegrey.comReposted by @honeycomb.ioYour SLO breaks. Can your tooling tell you why, or just that something is wrong and customers are suffering? SLOs were popularised by the SRE book a decade ago. Most teams *still* aren't using them well.
@honeycomb.ioNew O11yCast — Mike Goldsmith, Martin Thwaites, and Ken Rimple on adaptive sampling, including Honeycomb's new Tail Sampling Processor for the OpenTelemetry Collector.
Have a listen here: honeycomb.io/resources/podcasts/ep-93-adaptive-sampling-strategies-with-mike-goldsmith
@honeycomb.ioClosing out our AI norms & values series.
Part 3: Ethics, ownership, and how we actually use AI day to day at Honeycomb — energy use, IP, bias, wages, and the places we still don't have good answers.
Worth a read if you're thinking through this for your own team.
www.honeycomb.io/blog/ai-norm...
@honeycomb.ioFin set out to 2x productivity. They nearly 3x'd it.
Darragh Curran, CTO at Fin, talks with Charity Majors about how observability became the trust mechanism that made it work.
honeycomb.io/blog/fin-cto-building-great-engineering-organizations-ai-era
@honeycomb.ioWolfgang Therrien and Jamie Danielson are speaking this week at ShipItCon 2026 in Dublin.
This year's theme is friction: where it slows teams down, and where it's actually doing its job.
More here: shipitcon.com
@honeycomb.ioTraditional monitoring catches the failures you already expected. AI workloads fail in stranger ways, which is why they need real observability instead of another dashboard built for problems you can already name.
@honeycomb.io93% of HiPages' job-posting API calls were succeeding. In a marketplace where a missed post is a missed job, that gap mattered. Here's how they closed it.
@honeycomb.ioBoris: the PR is a fossil. Charity: it's carrying way too much, "too much for one humble conversation." David: it's not can the agent do it, it's whether a person still adds judgment & collaboration. 2/2
@honeycomb.ioBoris Tane, David Poll, & Charity Majors debated whether AI killed the SDLC at O11yCon, and didn't agree. 1/2
