Topic
AI Agents
My writing, talks, podcasts, and projects about AI Agents.
Selected work
Eduardo Ordax — Principal GTM GenAI at AWS
The reality of GenAI in production — why organizational culture is the biggest blocker, scaling non-deterministic LLM systems, and what separates AI winners from experimenters.
LLMs Broke the SRE Runbook. Now What?
AI-generated code is outpacing traditional runbooks. How SRE teams are adapting their incident response playbooks for the LLM era.
From Vibes to Outages: Riding the AI Code Wave
AI-assisted coding is exploding — but acceleration doesn't mean reliability. Real examples of hard-to-trace LLM bugs, hallucinated dependencies, and operational fallout for lean SRE teams.
Will LLMs and Vibe Coding Fuel a Developer Renaissance?
Exploring whether AI-assisted coding tools will democratize software development or create new categories of hard-to-debug production issues.
Why Are Agent Protocols Like MCP and A2A Needed?
Breaking down the Model Context Protocol (MCP) and Agent-to-Agent (A2A) standards — why interoperability matters for the next wave of AI agents.
Is AI-assisted coding an incident magnet?
AI-generated code ships faster but introduces subtle bugs that are harder to trace. What engineering leaders need to know about the reliability trade-offs.
Anthony Alcaraz — GTM Agentic Engineering Lead at AWS
Your AI agents are lost — why context engineering and knowledge graphs give agents the structure they need for retrieval, memory, planning, reasoning, and learning from business outcomes.
Nir Soudry — Head of R&D at 7AI
AI vs. AI — how autonomous attackers are scaling phishing and malware, how defenders can use agents to cut alert noise, and where accountability belongs when AI investigates and remediates incidents.
Eran Kampf — VP of Engineering at Twingate
Why Twingate stopped shipping features to rebuild reliability — from active-active multi-region architecture and smaller blast radiuses to preserving human ownership as agentic coding accelerates delivery.
AIOps Summit
Meta-hosted summit in Menlo Park focused on applying AI, LLMs, and agents to software incidents and response.
Alexey Grigorev — Founder of DataTalks.Club
How a chain of reasonable-sounding decisions led an AI coding agent to run terraform destroy against a live production database — and the guardrails that separate moving fast from losing everything.
AI Made Developers 25% More Productive. It Also Tripled Our Incident Rate.
At AI DevSummit New York: how AI-assisted coding is reshaping software delivery — the productivity gains, the surge in incident rates, and what reliability teams need to do about it.
Maria Vechtomova — Co-founder at Cauchy
LLM observability — how monitoring principles from MLOps apply to large language model applications, the gaps that went overlooked for nine years, and best practices for observing AI systems.
Ganesh Datta — Co-Founder & CTO at Cortex
AI didn't change the game, it just exposed your bottlenecks — how platform engineering and SRE teams solve identical human problems through influence rather than authority, and why AI amplifies existing bottlenecks instead of transforming operations.
Reliability Rebels Podcast — Guest Appearance
Reliability Rebels Guest appearance to discuss reliability engineering, AI, and modern incident response.
Dana Lawson — CTO at Netlify
Fear, identity, and flaky tests — why SRE resistance to AI agents stems from identity and control concerns rather than the technology itself, and practical strategies for adopting AI-driven reliability tools starting with low-risk tasks.
Will Wilson — CEO at Antithesis
The incident you never had — deterministic simulation testing, why conventional testing misses bugs that cause real outages, and how simulation-based approaches improve software reliability.
Swizec Teller — Bestselling Author
Code is cheap, reliability isn't — owning production in the AI era, the hidden complexity of SRE work, and why human ownership remains essential.
OpenClaw Demo Night w/ Rootly AI, Convex, Sentry & DigitalOcean
An evening of AI demos and networking in Toronto — presenting Rootly AI alongside teams from Sentry, Red Brick Labs, Convex, and DigitalOcean.
Dileshni Jayasinghe — VP of Technology at commonsku
Democratizing reliability — empowering non-engineers with operational power, incident management as a muscle, and AI-powered postmortems.
Tomás Hernando Koffman — Co-founder at Not Diamond
99%+ accuracy on a moving target — model deprecation, reliability with LLMs, and treating prompts as architectural components.
Developers lose focus 1,200 times a day — how MCP could change that
Context switching kills developer productivity. How the Model Context Protocol (MCP) can reduce tool fragmentation and keep engineers in flow.
Julien Simon — VP and Chief Evangelist
A conversation about developer advocacy, AI evangelism, and building reliable systems at scale.
Tea, Pipelines, and Retries: A Practical Guide to MLOps
Discussion at SREcon EMEA 2025 on how AI is transforming the software development lifecycle — CI/CD pipelines, deployments, scaling, monitoring, incident management, reliability tooling, and emerging disciplines like LLMOps.
AI Meets Reliability
Panel exploring AI-driven automation and observability with leaders from NVIDIA, OpenAI, Baseten, Replit, and Weights & Biases on scaling operations and reducing MTTR.
Rob Zuber — CTO at CircleCI
The end of good code, AI throughput, and what reliability means at CI/CD scale.
MCPs and the Next Wave of Reliability w/ Rootly AI, WorkOS, Block, Microsoft & Groq
Panel at the AWS GenAI Loft in San Francisco on MCPs, incident automation, observability, and generative AI tooling for reliability.
AI-First Platform Engineering: 3 Signals From PlatformCon
Three emerging patterns from PlatformCon that signal how AI is reshaping internal developer platforms and platform team workflows.
Frontiers of AI: Building with Rootly AI, Zscaler, CircleCI, Fireworks AI & Google DeepMind
Panel at Google HQ with 300+ attendees exploring real-world Gemini models, reinforcement learning, next-gen agent systems, and AI reliability.
AI Security Demo Night w/ Rootly AI, Okta, Panther, Tailscale & More
Live demos at Okta HQ in San Francisco — presenting Rootly AI alongside cybersecurity startups tackling identity, endpoint security, and threat response.
Exploring AI's Role in Incident Management
Interview with Alan Shimel at PlatformCon NYC 2025 on how AI applies to incident management and reliability engineering — triage, root cause analysis, and why AI enhances rather than replaces engineers.
How AI is Fueling the Developer Renaissance
Panel at the AWS GenAI Loft with leaders from a16z, AWS, Dagger, Braintrust, Baseten, and Arize AI on how AI is transforming developer workflows.
The Future of AI-Driven Reliability
Panel with leaders from a16z, Y Combinator, and Google Cloud on how MCP servers and agent-to-agent communication are revolutionizing developer tools. Demos from Anthropic, Sentry, Postman, and Browserbase.
Incident Vibing: The Self-Healing System
Tracing the arc from ingesting logs at LinkedIn/SlideShare to LLM-driven RCA today. How fine-tuning, MCP, and incident vibing are reshaping SRE.
Vibe Coding Is Here — But Are You Ready for Incident Vibing?
If developers are vibe coding, SREs are now incident vibing. What happens when AI-generated code meets production reality.
Rootly Roundtable: The State of AI in Incident Management
Invite-only roundtable examining AI's role in incident response — separating practical applications from hype with industry leaders.
Rootly-MCP-server
An MCP server for Rootly — enabling AI agents to interact with incident management workflows via the Model Context Protocol.
SRE-skills-bench
A benchmark suite for evaluating AI agents on real-world SRE tasks — incident diagnosis, runbook execution, and infrastructure troubleshooting.
MCP-Sylvain-Kalache
A personal MCP server that exposes my portfolio data — articles, podcasts, talks, panels, GitHub projects, news, bio, timeline, and live weather — as queryable tools for AI agents.
Introducing Canyon — an AI for developer self-service
Canyon lets developers provision resources, troubleshoot issues, and interact with internal tooling via natural-language prompts.
From hype to impact: How AI is reshaping platform engineering
Moving past the buzzwords — where AI is actually changing developer workflows, internal developer portals, and what it means for platform teams.
EU AI Act Secrets Revealed
What the EU AI Act actually requires — risk classifications, compliance timelines, and what it means for AI-driven products.
Intelligent Document Processing Compliance, from Stone Tablets to Digital Docs
The evolution of document processing compliance — from physical records to AI-powered intelligent document processing systems.
AI-Driven Incident Resolution — Hype or Reality
A grounded look at what LLMs can and can't do in an incident response workflow today, drawn from real production experiments at Rootly AI Labs.
Navigating LLMs challenges in data security & compliance
The unique data security and compliance challenges that large language models introduce — from training data to inference outputs.
For companies that use ML, labeled data is the key differentiator
Why Tesla leads on ADAS — and what it teaches every company about the strategic value of training data, annotation pipelines, and the $6B labeling market.