"How do we make kids learn? We send them to school. They take exams, get graded, and when they pass, they move up a grade."
@gaurav25d showed how to build self-improving agent loops at Durable AI by @temporalio.
Silent tool call failures were slowing a support agent down. One
youtube.com/watch?v=dFCFpT…
@gaurav25d showed how to build self-improving agent loops at Durable AI by @temporalio.
Silent tool call failures were slowing a support agent down. One
youtube.com/watch?v=dFCFpT…
0
2
3
32
Hello Bay Area 👋,
@gaurav25d is presenting how to build self-improving agents with @temporalio and Oodle on Oct 21 in SF.
It's the October edition of the Durable AI meetup series, hosted by Temporal & Kernel.
See you there: luma.com/durable-ai-oct…
@gaurav25d is presenting how to build self-improving agents with @temporalio and Oodle on Oct 21 in SF.
It's the October edition of the Durable AI meetup series, hosted by Temporal & Kernel.
See you there: luma.com/durable-ai-oct…
0
1
3
100
Yesterday, we put Jev to the test against GPT-5.4-mini, Claude Haiku 4.5, and Gemini 3.5 Flash.
Not another generic model benchmark. We focused on one question: Is Jev good enough for YOUR workload?
We walked through how to:
- Quickly generate representative datasets from your als across different models.
- Compare accuracy, latency, and cost to make the right trade-offs.
My biggest takeaway: YOUR WORKLOAD IS THE BENCHMARK.
If you're building agents and care about price-performance, this session is for you.
Missed it? DM me and I'll share the recording.
Not another generic model benchmark. We focused on one question: Is Jev good enough for YOUR workload?
We walked through how to:
- Quickly generate representative datasets from your als across different models.
- Compare accuracy, latency, and cost to make the right trade-offs.
My biggest takeaway: YOUR WORKLOAD IS THE BENCHMARK.
If you're building agents and care about price-performance, this session is for you.
Missed it? DM me and I'll share the recording.
0
0
2
53
Oodle AI retweeted
The most important user of observability is no longer a human. It’s an agent.
Agents debug production software. Agents debug other agents.
And AI needs full context across both.
We believe full-stack observability and agent observability will converge into one system — and the future of observability will increasingly be headless.
That changes the workload completely. Humans click through five or ten dashboards. Agents don’t.
Sampling doesn’t work for agent traces because failures are non-deterministic. The one trace you need may be the one you dropped. Agents can generate 1,000x more queries than humans. They query raw, unaggregated data. And they expect instant answers.
Thus, agentic workloads require a fundamental rethink of observability architecture.
youtu.be/JZiwlZSPTTc?si…
Agents debug production software. Agents debug other agents.
And AI needs full context across both.
We believe full-stack observability and agent observability will converge into one system — and the future of observability will increasingly be headless.
That changes the workload completely. Humans click through five or ten dashboards. Agents don’t.
Sampling doesn’t work for agent traces because failures are non-deterministic. The one trace you need may be the one you dropped. Agents can generate 1,000x more queries than humans. They query raw, unaggregated data. And they expect instant answers.
Thus, agentic workloads require a fundamental rethink of observability architecture.
youtu.be/JZiwlZSPTTc?si…
2
3
3
173
Oodle AI retweeted
I loved the internal session @real_akhilesh_k had conducted for the @oodleai team. After some convincing, he's finally agreed to also do a public session.
Drop in, you'll leave with good mental model for Jev, and a reality-check cutting through the hype. Or, maybe the hype is totally justified! Join us to find out!
luma.com/5oe9e06h?tk=IN…
Drop in, you'll leave with good mental model for Jev, and a reality-check cutting through the hype. Or, maybe the hype is totally justified! Join us to find out!
luma.com/5oe9e06h?tk=IN…
0
2
4
156
A strong Q2 at Oodle: 9 PB+ of production telemetry processed last quarter and 120M+ agent traces/month. Still early.
9 PB of production telemetry.
Processed last quarter.
Zero sampling.
That number made me pause.
Oodle is only three years old. I still remember when "scale" meant something very different.
Proud of our incredible team.
And grateful to the customers trusting us.
0
1
3
154
Could you build faster + cheaper logs engine than ClickHouse? checkout talk by Sujeet, our founding engineer for the trade offs.
Heard "news" that ClickHouse won the observability war. So, did a talk on how "winner" may not be the best choice when you're an user.
Shout out to @hasgeek: you made me realize, after 20-year gap, I can still do "public speaking" without breaking down in shivers and jitters 😅
Being excited to share what we've built at @oodleai definitely helped with the debilitating stage-fright for sure!
youtube.com/watch?v=etsjnZ…
0
1
2
92
Oodle AI retweeted
Heard "news" that ClickHouse won the observability war. So, did a talk on how "winner" may not be the best choice when you're an user.
Shout out to @hasgeek: you made me realize, after 20-year gap, I can still do "public speaking" without breaking down in shivers and jitters 😅
Being excited to share what we've built at @oodleai definitely helped with the debilitating stage-fright for sure!
youtube.com/watch?v=etsjnZ…
Shout out to @hasgeek: you made me realize, after 20-year gap, I can still do "public speaking" without breaking down in shivers and jitters 😅
Being excited to share what we've built at @oodleai definitely helped with the debilitating stage-fright for sure!
youtube.com/watch?v=etsjnZ…
0
6
7
631
Oodle AI retweeted
9 PB of production telemetry.
Processed last quarter.
Zero sampling.
That number made me pause.
Oodle is only three years old. I still remember when "scale" meant something very different.
Proud of our incredible team.
And grateful to the customers trusting us.
Processed last quarter.
Zero sampling.
That number made me pause.
Oodle is only three years old. I still remember when "scale" meant something very different.
Proud of our incredible team.
And grateful to the customers trusting us.
0
1
1
230
Oodle AI retweeted
Excited to see our founding engineer, Anirud Samala, speaking at the next @ReactBangalore meetup.
He'll share how we used OpenTelemetry + AI to track down and fix Next.js build time spikes. If you're building with React or Next.js, come check it out!
lnkd.in/gF_xvXuu
He'll share how we used OpenTelemetry + AI to track down and fix Next.js build time spikes. If you're building with React or Next.js, come check it out!
lnkd.in/gF_xvXuu
1
4
3
348
Oodle AI retweeted
@oodleai is live on @auth0 marketplace!
Auth logs are logs too.
Let logs live together 🤗
Let logs live longer 💪🏽
Auth logs are logs too.
Let logs live together 🤗
Let logs live longer 💪🏽
At @oodleai we have been using some features, and so have our customers. No launch post, no announcements, just docs, and happy users.
Yesterday, it hit me: It's not about the features, but a broader point: "let logs live together".
So I wrote a launch post for @auth0 logs
1
2
2
111
Oodle AI retweeted
These .new TLDs are pretty cool!
Got one for OTel for @oodleai (otel dot new) 🥳
Got one for OTel for @oodleai (otel dot new) 🥳
1
2
3
117
Oodle AI retweeted
New integration! ✨
@supabase + @oodleai
Our customers are sending us terabytes of logs a day, and now they can send Supabase logs too!
@supabase + @oodleai
Our customers are sending us terabytes of logs a day, and now they can send Supabase logs too!
Log Drains are now available on Pro. Send your Postgres, Auth, Storage, Edge Functions, and Realtime logs to Datadog, Grafana Loki, Sentry, Axiom, S3, or your own endpoint.
> supabase.com/blog/log-drain…
1
5
11
401
Oodle AI retweeted
A prospect at @oodleai, (now a customer) had a problem that is hilarious and frustrating at the same time: engineers always worried "If I log in to look at logs, will I end up logging out a colleague?"
Come to @oodleai, we don't have crazy per-user pricing! (full story in reply)
Come to @oodleai, we don't have crazy per-user pricing! (full story in reply)
1
3
3
104
The year is 2026.
AI is writing your production code. Your observability tool expects you to write a PromQL query. You are correlating timestamps manually across browser tabs. It doesn't make sense.
That's why we built Oodle AI. #DebugLikeCursor play.oodle.ai
AI is writing your production code. Your observability tool expects you to write a PromQL query. You are correlating timestamps manually across browser tabs. It doesn't make sense.
That's why we built Oodle AI. #DebugLikeCursor play.oodle.ai
0
0
1
63
Synthetic monitors are great at telling something's broken. Not at telling why.
So we built them into Oodle. When a check fails, it's next to your logs, metrics, and traces. Just ask in plain english to find the cause. Setup your first monitor in a minute. shorturl.at/jPxNw
So we built them into Oodle. When a check fails, it's next to your logs, metrics, and traces. Just ask in plain english to find the cause. Setup your first monitor in a minute. shorturl.at/jPxNw
0
0
0
46
Fuel Labs runs $1B+ in annual transaction volume on their blockchain infrastructure.
25% of their DevOps time used to go towards managing their #Prometheus, #Grafana, #Cloudwatch & #OpenSearch setup. With Oodle, it's down to Zero.
Case study - oodle.ai/case-studies/f…
25% of their DevOps time used to go towards managing their #Prometheus, #Grafana, #Cloudwatch & #OpenSearch setup. With Oodle, it's down to Zero.
Case study - oodle.ai/case-studies/f…
0
1
1
52
Debugging across multiple tools and 10s of dashboards? Introducing AI-Native Unified Observability.
blog.oodle.ai/meet-oodle-uni…
#observability #devops #sre #AI
blog.oodle.ai/meet-oodle-uni…
#observability #devops #sre #AI
1
1
1
103
50% cost savings. <3 hour migration. Unified Experience ( 3 tools to 1 ). Hear how Fello AI unified their #observability and empowered their entire engineering team.
oodle.ai/case-studies/f…
#prometheus #cloudwatch #grafana #DevOps #SRE
oodle.ai/case-studies/f…
#prometheus #cloudwatch #grafana #DevOps #SRE
0
2
1
138
Prometheus #OOMKilled 5th time this week!
As you scale, it breaks more frequently. You end up with dedicated engineers, overprovisioned systems, and limiting metrics instead of building cool features.
This blog explores and compares various strategies shorturl.at/4PFMd
As you scale, it breaks more frequently. You end up with dedicated engineers, overprovisioned systems, and limiting metrics instead of building cool features.
This blog explores and compares various strategies shorturl.at/4PFMd
0
0
0
87