# AI Agents — Sylvain Kalache

URL: https://www.sylvainkalache.com/topics/ai-agents

My writing, talks, podcasts, and projects about AI Agents.

## Selected work

- [Eduardo Ordax — Principal GTM GenAI at AWS](https://www.sylvainkalache.com/content/eduardo-ordax-principal-gtm-genai-at-aws.md) (2025-12-14; Humans of Reliability): 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?](https://www.sylvainkalache.com/content/llms-broke-the-sre-runbook-now-what.md) (2025-11-05; The New Stack): 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](https://www.sylvainkalache.com/content/from-vibes-to-outages-riding-the-ai-code-wave.md) (2025-10-08; SREcon EMEA 2025): 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?](https://www.sylvainkalache.com/content/will-llms-and-vibe-coding-fuel-a-developer-renaissance.md) (2025-07-09; The New Stack): 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?](https://www.sylvainkalache.com/content/why-are-agent-protocols-like-mcp-and-a2a-needed.md) (2025-05-21; The New Stack): 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?](https://www.sylvainkalache.com/content/is-ai-assisted-coding-an-incident-magnet.md) (2025-05-15; LeadDev): 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](https://www.sylvainkalache.com/content/anthony-alcaraz-gtm-agentic-engineering-lead-at-aws.md) (2026-07-21; Humans of Reliability): 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](https://www.sylvainkalache.com/content/nir-soudry-head-of-r-d-at-7ai.md) (2026-07-15; Humans of Reliability): 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](https://www.sylvainkalache.com/content/eran-kampf-vp-of-engineering-at-twingate.md) (2026-07-06; Humans of Reliability): 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](https://www.sylvainkalache.com/content/aiops-summit.md) (2026-06-23; 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](https://www.sylvainkalache.com/content/alexey-grigorev-founder-of-datatalks-club.md) (2026-06-16; Humans of Reliability): 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.](https://www.sylvainkalache.com/content/ai-made-developers-25-more-productive-it-also-tripled-our-incident-rate.md) (2026-06-09; AI DevSummit): 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](https://www.sylvainkalache.com/content/maria-vechtomova-co-founder-at-cauchy.md) (2026-05-14; Humans of Reliability): 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.
- [Doom Agent Arena](https://www.sylvainkalache.com/content/doom-agent-arena.md) (2026-05-11; GitHub): An MCP-native arena that pits AI models against each other in real-time Doom matches to benchmark spatial planning, adaptation, speed, accuracy, and cost.
- [Ganesh Datta — Co-Founder & CTO at Cortex](https://www.sylvainkalache.com/content/ganesh-datta-co-founder-cto-at-cortex.md) (2026-04-09; Humans of Reliability): 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](https://www.sylvainkalache.com/content/reliability-rebels-podcast-guest-appearance.md) (2026-04-08; Reliability Rebels): Reliability Rebels Guest appearance to discuss reliability engineering, AI, and modern incident response.
- [Dana Lawson — CTO at Netlify](https://www.sylvainkalache.com/content/dana-lawson-cto-at-netlify.md) (2026-03-31; Humans of Reliability): 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](https://www.sylvainkalache.com/content/will-wilson-ceo-at-antithesis.md) (2026-03-19; Humans of Reliability): 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.
- [MCP-Sylvain-Kalache](https://www.sylvainkalache.com/content/mcp-sylvain-kalache.md) (2026-02-28; GitHub): 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.
- [Swizec Teller — Bestselling Author](https://www.sylvainkalache.com/content/swizec-teller-bestselling-author.md) (2026-02-17; Humans of Reliability): 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](https://www.sylvainkalache.com/content/openclaw-demo-night-w-rootly-ai-convex-sentry-digitalocean.md) (2026-02-11; OpenClaw): 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](https://www.sylvainkalache.com/content/dileshni-jayasinghe-vp-of-technology-at-commonsku.md) (2026-01-14; Humans of Reliability): 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](https://www.sylvainkalache.com/content/tom-s-hernando-koffman-co-founder-at-not-diamond.md) (2025-12-23; Humans of Reliability): 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](https://www.sylvainkalache.com/content/developers-lose-focus-1-200-times-a-day-how-mcp-could-change-that.md) (2025-12-22; VentureBeat): 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](https://www.sylvainkalache.com/content/julien-simon-vp-and-chief-evangelist.md) (2025-11-19; Humans of Reliability): A conversation about developer advocacy, AI evangelism, and building reliable systems at scale.
- [Tea, Pipelines, and Retries: A Practical Guide to MLOps](https://www.sylvainkalache.com/content/tea-pipelines-and-retries-a-practical-guide-to-mlops.md) (2025-10-07; SREcon EMEA 2025): 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](https://www.sylvainkalache.com/content/ai-meets-reliability.md) (2025-09-23; Rootly AI Labs): 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](https://www.sylvainkalache.com/content/rob-zuber-cto-at-circleci.md) (2025-09-10; Humans of Reliability): 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](https://www.sylvainkalache.com/content/mcps-and-the-next-wave-of-reliability-w-rootly-ai-workos-block-microsoft-groq.md) (2025-08-28; Rootly AI Labs): 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](https://www.sylvainkalache.com/content/ai-first-platform-engineering-3-signals-from-platformcon.md) (2025-08-25; The New Stack): 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](https://www.sylvainkalache.com/content/frontiers-of-ai-building-with-rootly-ai-zscaler-circleci-fireworks-ai-google-deepmind.md) (2025-07-25; Rootly AI Labs): 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](https://www.sylvainkalache.com/content/ai-security-demo-night-w-rootly-ai-okta-panther-tailscale-more.md) (2025-07-22; AI Security Demo Night): 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](https://www.sylvainkalache.com/content/exploring-ai-s-role-in-incident-management.md) (2025-06-30; Techstrong TV): 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](https://www.sylvainkalache.com/content/how-ai-is-fueling-the-developer-renaissance.md) (2025-06-05; Rootly AI Labs): 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](https://www.sylvainkalache.com/content/the-future-of-ai-driven-reliability.md) (2025-04-24; Rootly AI Labs): 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](https://www.sylvainkalache.com/content/incident-vibing-the-self-healing-system.md) (2025-04-23; Adventures in DevOps): 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?](https://www.sylvainkalache.com/content/vibe-coding-is-here-but-are-you-ready-for-incident-vibing.md) (2025-03-19; The New Stack): 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](https://www.sylvainkalache.com/content/rootly-roundtable-the-state-of-ai-in-incident-management.md) (2025-03-06; Rootly): Invite-only roundtable examining AI's role in incident response — separating practical applications from hype with industry leaders.
- [Rootly-MCP-server](https://www.sylvainkalache.com/content/rootly-mcp-server.md) (2025-03-01; GitHub): An MCP server for Rootly — enabling AI agents to interact with incident management workflows via the Model Context Protocol.
- [SRE-skills-bench](https://www.sylvainkalache.com/content/sre-skills-bench.md) (2025-02-01; GitHub): A benchmark suite for evaluating AI agents on real-world SRE tasks — incident diagnosis, runbook execution, and infrastructure troubleshooting.
- [Introducing Canyon — an AI for developer self-service](https://www.sylvainkalache.com/content/introducing-canyon-an-ai-for-developer-self-service.md) (2025-01-01; Platform Engineering): 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](https://www.sylvainkalache.com/content/from-hype-to-impact-how-ai-is-reshaping-platform-engineering.md) (2024-06-01; PlatformCon): 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](https://www.sylvainkalache.com/content/eu-ai-act-secrets-revealed.md) (2024-05-15; Data Defenders Forum): 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](https://www.sylvainkalache.com/content/intelligent-document-processing-compliance-from-stone-tablets-to-digital-docs.md) (2024-05-08; Data Defenders Forum): The evolution of document processing compliance — from physical records to AI-powered intelligent document processing systems.
- [AI-Driven Incident Resolution — Hype or Reality](https://www.sylvainkalache.com/content/ai-driven-incident-resolution-hype-or-reality.md) (2024-05-01; DevOpsDays Austin): 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](https://www.sylvainkalache.com/content/navigating-llms-challenges-in-data-security-compliance.md) (2023-10-03; Data Defenders Forum): 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](https://www.sylvainkalache.com/content/for-companies-that-use-ml-labeled-data-is-the-key-differentiator.md) (2021-05-18; TechCrunch): 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.

[All work](https://www.sylvainkalache.com/index.md)