Notes on remediation-first infrastructure
Engineering deep-dives, product updates, and field notes from teams shifting off reactive observability.

When They Go Wide, We Go Deep
Every pod was Ready, every log was clean, and the alert named the wrong service. A real Kubernetes incident, and why an autonomous SRE needs depth, not more integrations, to fix it.
Nati ShalomCo-founder & CTO
Engineering9 minOne Slow Disk, Every GPU Waits: Remediation First for GPU Clusters
On a GPU training cluster, one slow disk stalls every GPU while Kubernetes reports Ready. Three faults from our lab, what our agent got wrong on each, and what it took to fix them.
Nati Shalom • Nov 9, 2026
Engineering13 minJev Named the Layer. Now Somebody Has to Measure It.
A model that only makes decisions is the right idea for autonomous operations. But the number that autonomous action depends on is still unmeasured, by the vendor, by the press, and by the first paper built on it.
Nati Shalom • Nov 1, 2026
Engineering10 minNobody Can Tell You What Observability Really Costs
Every observability vendor bills in a different unit. We priced one fixed Kubernetes workload with each vendor's own model to show the real cost, why the cheapest option changes as you grow, and why the curve doesn't scale.
Nati Shalom • Nov 1, 2026
Engineering13 minWho Is the Dashboard For?
A self-driving car has no dashboard and more sensors than ever. The same inversion is coming for cloud observability: less data, a sharper picture, and a new answer to multi-cloud.
Nati Shalom • Nov 1, 2026
Engineering7 minAGNTCon + MCPCon 2026 Recap: Lots of Noise, Little News. And That’s Good News
My takeaways from AGNTCon + MCPCon 2026: agent security dominated, gateways and sandboxes are multiplying, MCP beat A2A on stage, and the real question is no longer whether agents will fix incidents on their own, but how.
Nati Shalom • Oct 26, 2026
Engineering14 minThe Causality Skew: Why a Centralized AI Diagnoses Your Incident Out of Order
Metrics land in seconds, cloud audit events in minutes. An AI agent reasoning on a central store sees symptoms before causes and diagnoses your incident out of order. A worked OOMKill example and two fixes: run the agent where data is born, and watermark every source.
Nati Shalom • Oct 22, 2026
Engineering8 minShip the Agent to the Data. Here Is What That Looks Like.
Autonomous SRE flips observability: ship the agent to the data. How DataAgent runs in your cluster, turns your backend into long-term storage, and fixes through your own pipeline, acting only within the fault classes you enable.
Nati Shalom • Oct 21, 2026
Engineering10 minThe Simulator, Not the Archive: Your AI SRE Learns Somewhere Else
Training an AI SRE on your incident archive clones what tired humans did. Nati Shalom on why agents should learn the way pilots do: replaying engineered faults in a sandbox, graded against an answer key, with trust earned per fault class.
Nati Shalom • Oct 20, 2026
Engineering20 minTwo Specialist Agents, One Incident: What an A2A Handoff Takes
We wired a Kubernetes agent to the AWS DevOps Agent over A2A. The protocol handled the delegation; authorization, cost and accountability we had to build. Measured over 63 runs on a real AWS account, against real faults.
Nati Shalom • Oct 18, 2026
Engineering7 minWhat You Won’t Hear at ObservabilityCON
The stage shows an AI agent on top of your data cloud. The people who carry the pager, when no vendor is in the room, describe something else: an agent that goes to the data, a backend that becomes storage, and a bill that stops charging for a reader who has left.
Nati Shalom • Oct 15, 2026
Engineering14 minWho Pays for the Dashboard?
We priced one Kubernetes-on-AWS workload across observability backends. Copying telemetry into a SaaS platform costs several times the open-source floor, and an autonomous agent uses none of what the premium buys. Build the substrate cheap; fund the loop.
Nati Shalom • Oct 11, 2026
Engineering11 minIs AI Making Your SRE Team Better? Ask the Scale, Not the Mirror
AI makes engineers faster at producing changes and makes production less stable. Which effect wins depends on your platform. Four production numbers tell you whether AI is actually making your SRE team better.
Shahar Harap • Oct 11, 2026
Product12 minDataAgent vs. Dynatrace: Breadth and Causal AI, or the Fix Applied and Verified
Looking for a Dynatrace alternative for the fix? Compare Dynatrace Intelligence and DataAgent on six axes: autonomy, proof, and what each does to your bill.
DataAgent • Oct 6, 2026
Product7 minThey Told Me the Market Is Saturated. That's Exactly Why I Co-Founded DataAgent.
Everyone agrees observability costs are out of control, and almost nobody believes there is a fix. That gap is why I co-founded DataAgent and serve as its CTO.
Nati Shalom • Oct 5, 2026
Engineering8 minHow to Reduce MTTR: The Slowest Part of an Incident Is a Human Reading a Graph
Detection takes seconds; MTTR lives in diagnosis, action, and verification. Remediation-first systems reduce it faster than any dashboard or assistant.
Nati Shalom • Oct 5, 2026
Product13 minDataAgent vs. Splunk: Cross-Domain Correlation, or the Fix Applied and Verified
Looking for a Splunk alternative for the fix? Compare Splunk AI SRE and DataAgent on six axes: autonomy, proof, and what each does to your bill.
DataAgent • Oct 2, 2026
Engineering10 minAI Is Breaking Production, and Observability Can't Fix It
AI made us faster at producing plausible code, not at keeping production healthy. More observability gives you a better thermometer; a smarter assistant gives you more memos to verify. Neither closes the loop. A remediation-first Digital Immune System does.
Nati Shalom • Sep 27, 2026
Product11 minDataAgent vs. Elastic: Full Comparison
Keep Elastic for search and security. See how DataAgent adds autonomous K8s and cloud remediation on top, and helps cut your Elasticsearch storage bill.
DataAgent • Sep 1, 2026
Product12 minDataAgent vs. Datadog: From Finding the Problem to Fixing It
Running Datadog? See how DataAgent adds autonomous remediation alongside it, and how its read-in-place model compares to Datadog's per-host billing.
DataAgent • Sep 1, 2026
Product13 minDataAgent vs. Grafana Cloud: Where Observability Stops and Remediation Begins
See where Grafana Cloud's observability stops and where DataAgent's autonomous remediation picks up, with a fault-by-fault breakdown.
DataAgent • Sep 1, 2026
Engineering13 minLeft of Boom, Right of Boom: Prevention and Remediation Are One Loop
The industry split reliability into two disciplines: prevent it, and survive it. Both ask the same causal question — which change produces which failure — and neither can answer it alone. Why the postmortem is a broken transfer mechanism.
Nati Shalom • Aug 26, 2026
Engineering14 minThe Great Inversion: AI Writes the Code, Production Pays the Review Bill
Generation got two orders of magnitude cheaper; human verification capacity did not move. GitClear's 2026 data shows duplication at its highest recorded level while refactoring falls — and the defects that reach production are structural, not syntactic.
Nati Shalom • Aug 26, 2026
Engineering12 minPrecision vs. Recall: Why Observability Costs So Much
Observability bills are shaped by one unstated assumption: a human might need to scroll back. That is collection optimized for recall. An agent present at the moment of the fault optimizes for precision — and precision scales with your incident rate, not your estate.
Nati Shalom • Aug 26, 2026
Engineering11 minWhat Is Remediation in Production Systems? Mitigation, Remediation and Resolution, Defined
Remediation means correcting the condition that caused a fault — not reducing its impact, not restoring service, not closing the ticket. The difference between mitigation, remediation, resolution and recovery, and what makes an action safe to automate.
Nati Shalom • Aug 26, 2026
Engineering13 minAutonomy Is a Body Problem, Not a Data Problem
In 2026 the biggest observability vendors conceded that fixing, not watching, is the point. But an agent bolted onto a billed-by-ingestion data lake inherits the lake's ceiling. Autonomy is a harness problem: tools, context, state, guardrails and verification. Not a data problem.
Nati Shalom • Aug 24, 2026
Engineering7 minA Year of DevOps With an AI Agent
A year of running infrastructure with an AI agent in the loop, before and outside DataAgent. Not a 10x story: what actually changed was when my mistakes surfaced, what the work cost, and how much of it I had to remember myself.
Shahar Harap • Aug 24, 2026
Product3 minIt's time to let the dog out
For decades, observability has worked like a watchdog: it barks when something's wrong and waits for you to come running. That stopped making sense the moment we could build a system that acts on its own. Why I named the company DataAgent.
Ishay Yaari • Aug 24, 2026