Checkly

💥 AI Agents at work meetup this Wednesday. I'm in San Francisco this week together with Daniel Paulus & Sergii Bezliudnyi, and hence it's event time. 2027.dev (Mika Sagindyk) and Checkly are hosting a meetup. Come join us if you are in town and want to talk AI agents and see demos. https://luma.com/g8pxvf6u

Your coding agent is wasting 99% of what it fetches on junk. At least most of them are. A reality check.. We tested 7 popular AI agents to see how they fetch web content. Only 3 out of 7, Claude Code, Cursor, and OpenCode, request token efficient Markdown. The rest, Codex, Gemini CLI, GitHub Copilot, Windsurf, send standard headers and get full HTML every time. Why does this matter? Agents do not see what you see. They do not render pages. They ingest raw HTML. That means thousands of lines of nav bars, footers, div wrappers, and class names that carry zero value for an LLM. We measured it on our own docs. HTML fetch: 180,000 tokens. Markdown fetch: 478 tokens. That is a 99.7% reduction for the exact same content. And the fix has existed since 1997. It is called content negotiation. Agents can set an Accept header and ask for what they actually need, for example Markdown instead of HTML. The server then decides whether to honor it. If you ask the Checkly docs for Markdown, you get Markdown. Simple. Some companies like Vercel and Cloudflare are starting to serve agents properly. Most of the ecosystem still treats agents like browsers. If your agent does not request Markdown, you are burning tokens, slowing responses, degrading output quality, and paying more than necessary. The web has new visitors now. They do not have eyes. They have context windows. In an agentic world, token efficiency is not a micro optimization. It is infrastructure. We already serve Markdown at Checkly. This is only the beginning. The web will adapt. Or agents will route around it. Link is in the comments below.

I built a Datadog synthetics to Checkly checks migration tool, and it's rad. I'm already using it with customers and prospects and it's a hit. 🦝 It's a hit because; - Datadog is expensive ( we save our customers 50% off their bill or more at scale) 💰 - Migrating is expensive (having an entire team work on a migration for a month or more... whew) - Every org is resource constrained (I think 2025 is year of most layoffs after 2022) 😓 - When you want to replace a solution, you mainly want to know is some base layer of functionality covered (Anyone doing pilots based on price takeouts knows this) ✅ - Knowing what attributes map to what is annoying and manual process that honestly no random team learning a new tool should have to do on their own when moving fast matters most (why didn't I build this sooner?) 🧓 Creating a migrator for DD -> Checkly mainly just required looking at DD docs and being a Checkly expert (after almost 3 years, I think I am at this point, maybe not). AI was involved (it was Cursor, but I'm not going to renew after February in favor of open source tools. Was cute while it lasted, but not that impressive for how I want to leverage AI) DD or New Relic, or any other solutions synthetics exported are just arrays of objects w/ attributes. There are no attributes that are totally foreign to Checkly, just a matter of how resources relate back to each other. For example, DD doesn't assign alerts to a synthetic, it defines alerts within a synthetic. Aside from little weirdness like this (which seems objectively not great by the way) their synthetics pretty nicely map to Checkly checks. This means that AppDynamics, New Relic, and other synthetic monitoring solutions we compete against are now also on my list to create migrators for following the same approach. 🎯 We already have customers and prospects raising their hands requesting the migrator be available for other these other competitors. This is where we are in tech. An SE or a PM or a CEO (if a CEO can do it really anyone can) can build a solution in a few days that works and creates value and immediately just based on word of mouth that solution can become a signal of value, of direction to drive toward, and of how you can win better conquer a market and help people save money fast from being taken advantage of by legacy monitoring solutions. I think after my migrators phase, I'm going to return to my Generative Customer Health App that essentially wraps around Hubspot or Salesforce or whatever CRM you're using to actually be useful for account teams. 🥹

# Signals Conference Berlin 2026 - Reliability in the Age of AI I’m thrilled to announce Signals Conference 2026! A community event at the intersection of SRE × AI × MLOps x Data, in the heart of Europe. Signals is a curated, two-day, single-track event: one room, one stage, with lots of space for conversations between people who are building and operating systems at the frontier. Intentionally limited in size and scope. Built for hallway-track quality, and real community. Speakers include: - Alejandro Saucedo - Zalando - Alex Palcuie - Anthropic - Max Deichmann - LangFuse - Charity Majors - Honeycomb - Sylvain Kalache - Rootly - Maria Vechtomova - Cauchy - Niall Murphy - Stanza Find us at: - 💻 signalsconf.io - 📍 Berlin, Kulturbrauerei - 📅 Sept 10–11, 2026 👉 Visit signalsconf.io and subscribe to the newsletter to get notified when tickets become available. We will have a call for participation if you want to speak at the event! 👉 Follow Signals Conference on LinkedIn Huge thanks to everyone who helped behind the scenes to make this real! Special thanks to our founding sponsors Dash0, Checkly and Chronosphere who co-developed the concept, and brought this to life! Thanks to our sponsors honeycomb.io, OllyGarden, Polar Signals, Coralogix, NOFire AI and Braintrust, for the support! Shoutout to our stellar program team, which is putting together the most exciting reliability event on the 2026 calendar: Mirko Novakovic Hannes Lenke Martin Mao Charity Majors Stig Soerensen Alejandro Saucedo Luis Mineiro Shery Brauner Jeremy Colin Juraci Paixão Kröhling Matthias Loibl 🧊 Dennis Pfändler Sven Johann Alexander Heusingfeld Robert Comeau Marcel Sim Salvatore Furino Spiros Economakis Kyle Forster Severin Neumann Eric Kontargyris Made with ❤️

It's vanity metrics Tuesday! Last week the Checkly CLI reached 100K+ downloads per week. We released v.0.1 of the CLI three years ago and now most of our large customers use it every day from their IDEs and CI pipelines to build and manage their testing and synthetic monitoring pipelines: kinda like if Terraform and Jest had a baby, but then in TypeScript... It was quite a winding road to get there: We started of with a YAML based approach — killed that thank god — then did nothing for a while. Then we went all in on TypeScript, trusting that not only devs, but also SREs and infra engineers would find a way to adopt it. We also had to jump through 719 hoops to claim the "checkly" name from NPM as it was squatted by a user that had vanished from the internet. Now, it's front and center on our main landing page: `npm create checkly`. And, coincidentally, with the AI age blooming, we are finding all kinds of new opportunities as the Linux-style CLI + TS code are extremely familiar ground for most coding agents like Claude, Codex and the rest. https://lnkd.in/e2_Ed8N9