Pranay Prateek

We've been so consumed by the excitement of meeting our most important team member ever in person at KubeCon Atlanta that we didn't communicate as clearly as we should have.

We've been working with Olly for about 2 months now.

We'll be ready to share more about Olly very soon.

I know there's been a lot of confusion.

Since yesterday, my team and I have received numerous messages from customers and well-wishers asking about Olly.

The truth is, things have moved very fast.

AI x Open source : Community evening with Hatchet, Medplum and SigNoz · Luma As part of Open Source AI week, Join the teams behind SigNoz, Hatchet, and Medplum for an evening on how open source is shaping the future of AI—from…

Seats are limited; approval required. See you there!

👉 Register here: luma.com/2w9p5qzn

Mixer with builders, maintainers, and founders

When & where
🗓 Wed, Oct 22, 5:00–7:30 PM PT
📍 98 Battery St, San Francisco

🤓 Who should join

AI/ML engineers, infra/platform teams, OSS maintainers, and founders who care about reliability & open standards

If you’re building AI in production, come hang with the folks shipping the tooling behind it.

💡 What to expect

Candid panel: Open Source in the Age of AI (Hatchet × Medplum × SigNoz)
Lightning talks on agentic AI reliability, healthcare agents, and orchestration at scale

🚀 AI x Open Source — Bay Area community evening (Oct 22)

Open source and open standards are essential for transparency, & interoperability in the AI

We’re teaming up with Hatchet and Medplum to host an in-person meetup during Linux Foundation’s Open Source AI Week.

If you’re building Agentic AI, check out what’s possible with SigNoz + OpenTelemetry. Check out our docs or DM me.

signoz.io/docs/llm-obs...

- LLMs are non-deterministic → you need end-to-end traces to see what actually happened.

- Trace agent loops, tool calls, vector store lookups, and external API hops—along with latency, errors, retries, and token costs.

- Use OpenTelemetry for a standard, portable instrumentation model.

Loved the conversations with the cloud-native community and seeing how fast teams are moving Agentic AI from demos to production.

Why this matters:

Agentic AI is powerful—and unpredictable. That’s why observability (and traceability) is non-negotiable.

This week our SigNoz team was invited to Nutanix HQ in Bay Area to present “Taming Agentic AI with Observability.”

🎥 Clip from the talk below.
Big thanks to the Nutanix team for hosting!

Query Builder v5 - Two Years of Technical Debt, 80 Closed Issues, and a Fundamental Rethinking Query Builder v5 brings familiar SQL like syntax to observability data with expression-based querying that works across logs, metrics, and traces. Write complex queries using the syntax you already…

Query Builder v5 - Two Years of Technical Debt, 80 Closed Issues, and a Fundamental Rethinking | SigNoz signoz.io/blog/query-b...

great to see that you are liking it. If you have any queries/need any help - reach out to our slack community - signoz.io/slack

Installed SigNoz on my server in 5 minutes (literally, git clone && install bash) and now all my apps are sending logs automatically!

Really liking the look of signoz though... I hate how scattered everything in grafana is

LLM Observability in the Wild - Why OpenTelemetry should be the Standard LLM Observability breaks in wild today as new libraries like OpenInference don't adhere to OpenTelemetry standards leading to silos and suboptimal performance for users who want to tie LLM observability with rest of their observability stack

OTel is generally ahead in terms of how code meets metrics, nowadays, as far as I can see. Works for me

https://signoz.io/blog/llm-observability-opentelemetry/

LLM Observability in the Wild - Why OpenTelemetry should be the Standard LLM Observability breaks in wild today as new libraries like OpenInference don't adhere to OpenTelemetry standards leading to silos and suboptimal performance for users who want to tie LLM observabili...

Original blog post - signoz.io/blog/llm-obs...

And many such libraries like OpenInference which claim to be based on OpenTelemetry don't strictly adhere to it's conventions. This introduces problems for users who are trying to get better observability across their stack.

We discuss how building, debugging, and improving AI agents in production gets messy fast. There's multiple competing standards for default libraries for LLM observability.

🚀 Our article on issues with LLM observability currently and the lack of standards there is in top 10 in HN.

Inspired by my conversation with Pranav from Chatwoot, it seems lots of folks are facing such issues

when was the first time in your career you heard the term “observability’?

Also, if you are just getting started with observability and have any specific question, feel free to ask in comments below.

Now all observability practices you see in the market does some permutation and combination of above 4 steps

Now this post is already getting long, so I would continue this discussion in next post of this series, but I have a quick question -

- analysis and visualization UX where you can slice and dice data, explore root causes, set alerts, etc.

(We can argue that LLM powered agents can do 3&4 for us, but as of Sep 2025 we are not there yet, and IMO humans would always like the option of diving deeper manually if they want )

So, here are some basic concepts to be learnt.

Observability consists of:
- generating telemetry from your applications (e.g. Python service, etc) and infrastructure (k8s, fargate, VMs)
- storing it in a datastore
- querying the data via backend/engine

Let’s face it shipping code is more interesting than making it run reliably :P

(One can argue that this might change with the coming of vibe coding and claude code, writing code is now easy and actually running it in prod is where the real shit is, but I digress)

I realise this is because most engineers who become CTOs are software developers ( as YC says, the founding team should be able to ship your product), so unless they are senior engineers they are not exposed to the boring world of observability, application monitoring.

What I realised was that the audience, primarily composed of YC CTOs, who are most likely super technical don’t have basic understanding of questions around observability.

- How to balance monitoring everything vs cost. What is a good 80/20 here?
- Monitoring of desktop apps
- Traceability across infrastructure
- Standardization - different teams using different type of tools
- How to choose an observability backend
- How to get started with o11y? Do I need it ?

🤯 Recently I was present a talk in a YC CTO meetup and came back with a surprising realization that observability is still a foreign concept to even the most technical among devs

It was an Unconference talk, so I started with down all the questions which participants were asking.

That’s a principle we built our business on, and we’ve heard excellent feedback from others who have tried it our way instead.

If you want to learn more about SigNoz, or LLM observability in general, I’d always love to chat.

These are just some of the reasons an all-in-one tool for o11y is the right path.

At SigNoz, we believe that for observability, there is a lot of value in a single pane of observability.