←
Oodle AI89 posts
Oodle AI avatar

Oodle AI

@oodleai

Observability built for agents. AI debugs agents, applications and infrastructure at 80% lower cost.

◷ Joined March 2024▧ 24 Photos and videos
Oodle AI avatar
Oodle AI@oodleai·12h

Your Slack agent pings too much, forgets the thread, and dies when the process does. Ours doesn’t. It survives crashes, lets teammates jump into each other’s sessions, and knows when to shut up. On Thu, our founding engineer @sujeetgholap breaks down how we built it. Join us luma.com/8sq44kth

Post attachment
Sujeet avatar
⟳ Oodle AI retweeted
Sujeet@sujeetgholap·Nov 17

"ClickHouse doesn't suffer with cardinality, so neither should you." @mvijaykarthik has been cooking banger of a detailed benchmark, and unlike this announcement, he's gonna bring receipts. Game on!

Post attachment
Gaurav Maheshwari avatar
⟳ Oodle AI retweeted
Gaurav Maheshwari@gaurav25d·Nov 17

My next talk will be titled "Kuch khoya, kuch paaya" (Lost something, found something) - covering impact of AI in my software engineering career. It will be a story woven with Poetry and Engineering. Open invitation to any conference/meetup organizers who want a fun topic for their next event.

Sujeet avatar
⟳ Oodle AI retweeted
Sujeet@sujeetgholap·Nov 17

Oof! That's one spicy graph! 🥵 We ain't no benchmaxxing here pinky promise! @mvijaykarthik can't wait for your article! That'll help me keep the pinky promise.

Post attachment

"ClickHouse doesn't suffer with cardinality, so neither should you." @mvijaykarthik has been cooking banger of a detailed benchmark, and unlike this announcement, he's gonna bring receipts. Game on!

Post attachment
Gaurav Maheshwari avatar
⟳ Oodle AI retweeted
Gaurav Maheshwari@gaurav25d·Nov 17

That's what we are seeing in observability space as well, 93% of queries on @oodleai is from agents. Query patterns in observability have fundamentally changed with agents becoming three primary persona.

akhilesh krishnan avatar
⟳ Oodle AI retweeted
akhilesh krishnan@real_akhilesh_k·Nov 13

A common pattern I see nowadays. You start in a single cloud, quickly expand to multiple regions, and then to multiple clouds. Somewhere along the way, observability starts breaking. 🙅‍♂️

Post attachment
akhilesh krishnan avatar
⟳ Oodle AI retweeted
akhilesh krishnan@real_akhilesh_k·Nov 14

Blog: oodle.ai/blog/multi-cloud-mo…

A common pattern I see nowadays. You start in a single cloud, quickly expand to multiple regions, and then to multiple clouds. Somewhere along the way, observability starts breaking. 🙅‍♂️

Post attachment
Kiran Gollu avatar
⟳ Oodle AI retweeted
Kiran Gollu@kkgollu·Nov 14

🚀 Starting soon! @gaurav25d takes the stage for his keynote at Google Cloud Community Days! Excited to see @oodleai represented! @oodleai

Today, I’ll be delivering the Keynote at Google Cloud Community Day Kochi 2026, followed by a talk on Building Self-improving Loops for AI Agents. Keynote: 11:40–11:50 AM, Main Hall Technical session: 2:00–2:30 PM, Hall 1 📍 November 14, Holiday Inn, Kochi I’ll also be at the Oodle booth. If you’re building AI agents, come say hello.. let’s talk observability and what it takes to keep improving your agents.

Post attachment
Gaurav Maheshwari avatar
⟳ Oodle AI retweeted
Gaurav Maheshwari@gaurav25d·Nov 14

Today, I’ll be delivering the Keynote at Google Cloud Community Day Kochi 2026, followed by a talk on Building Self-improving Loops for AI Agents. Keynote: 11:40–11:50 AM, Main Hall Technical session: 2:00–2:30 PM, Hall 1 📍 November 14, Holiday Inn, Kochi I’ll also be at the Oodle booth. If you’re building AI agents, come say hello.. let’s talk observability and what it takes to keep improving your agents.

Post attachment
Sujeet avatar
⟳ Oodle AI retweeted
Sujeet@sujeetgholap·Nov 13

Meta point: Every time debugging something, you or your agents end up pulling data from multiple sources and correlating, know this: you deserve better! your agents deserve better!!

New @checklyHQ integration in @oodleai - debug your Checkly test failures with a single unified trace covering Checkly + Frontend + Backend

Post attachment
Gaurav Maheshwari avatar
⟳ Oodle AI retweeted
Gaurav Maheshwari@gaurav25d·Nov 12

Totally agree! I rolled out feedback tools in @oodleai MCP 2 weeks back giving agents ability to - report_bug - raise_feature_request It is incredibly helpful to improve the MCP. On top of this, I have a daily automation to query bug reports/feature requests in last 24 hours and raise PRs. Have been working really well to have a self-improving MCP server. All bug reports in last 7 days, all of them leading to fix PRs, so extremely high quality:

Post attachment

We added a tool for this in our MCP. Agent feedback very interesting so far! Recommend!

Gaurav Maheshwari avatar
⟳ Oodle AI retweeted
Gaurav Maheshwari@gaurav25d·Nov 12

Many SaaS companies hit the same wall when they sell to large enterprises or regulated industries like banks, fintech, or healthcare: "our data can't leave our cloud." The usual answer is self-hosted or BYOC. That leads to splitting the product into two parts: control plane and data plane. How do you observe a data plane you don't own? 🧵

Kiran Gollu avatar
⟳ Oodle AI retweeted
Kiran Gollu@kkgollu·Nov 10

@denalidb is sharing a series of technical deep dives on our journey: ClickHouse → DuckDB → DenaliDB (our own logs engine). lots of lessons from processing petabytes of telemetry for real customers in production. what worked, what didn't, and why we ended up building our own engine. fun journey, with plenty of surprises along the way...

Hey Opus 4.6, move us off ClickHouse. Nine months later, Opus 4.6 woke up and found his little brother, Opus 5.5, had gotten the job done. lnkd.in/p/g4THgcNw

Gaurav Maheshwari avatar
⟳ Oodle AI retweeted
Gaurav Maheshwari@gaurav25d·Nov 10

"𝗪𝗲 𝘄𝗶𝗹𝗹 𝗵𝗮𝘃𝗲 𝘁𝗼 𝗿𝗮𝗶𝘀𝗲 𝗮 𝗳𝘂𝗻𝗱𝗶𝗻𝗴 𝗿𝗼𝘂𝗻𝗱 𝗶𝗳 𝘄𝗲 𝗵𝗮𝘃𝗲 𝘁𝗼 𝗮𝗳𝗳𝗼𝗿𝗱 𝗗𝗮𝘁𝗮𝗱𝗼𝗴" This is what a prospect told me on a call last week. Such a fitting comment while we were running Datadog Detox Week. Here is a recap of what happened during the Datadog Detox week:

2:09