{"slug": "rootly-s-ai-labs-lead-revives-a-1983-warning-as-ai-sres-fix-routine-incidents", "title": "Rootly's AI Labs lead revives a 1983 warning: as AI SREs fix routine incidents, 'comprehension debt' leaves engineers unready for the hard ones", "summary": "Sylvain Kalache, who leads AI Labs at Rootly and was an SRE at LinkedIn, argues that AI incident-response tools create \"comprehension debt\" by auto-resolving the routine outages through which responders build system intuition, reviving human-factors researcher Lisanne Bainbridge's 1983 \"Ironies of Automation\" warning. Kalache predicts average resolution times for most incidents will fall as AI-assisted response spreads while resolution times for complex, never-seen-before incidents will rise, and he prescribes airline-style simulator drills, including a Rootly partnership with Uptime Labs, to keep engineers sharp. The argument is being debated in a heavily discussed Hacker News thread split between engineers who say they already feel the comprehension debt and critics who contend an incident beyond AI would stump humans regardless of practice.", "body_md": "[Artificial Intelligence](/category/ai)September 5, 2026\n\n# Rootly's AI Labs lead revives a 1983 warning: as AI SREs fix routine incidents, 'comprehension debt' leaves engineers unready for the hard ones\n\nEx-LinkedIn SRE Sylvain Kalache argues that AI incident response tools, by auto-resolving the routine outages through which responders build system intuition, are starving engineers of practice exactly as Lisanne Bainbridge's 1983 Ironies of Automation predicted, and he prescribes airline-style simulator drills to keep humans sharp; the Hacker News debate is split between engineers who feel the 'comprehension debt' already and critics who say an incident beyond AI would stump humans regardless of practice.\n\nSylvain Kalache, who leads AI Labs at Rootly and was an SRE at LinkedIn, uses \"comprehension debt\" for the growing gap between how a system actually behaves and how well the engineers on call understand it. His argument, now being picked apart in a heavily discussed Hacker News thread, is that AI responders fix the routine outages, and routine outages are where responders build the intuition a bad night depends on. [1](#ref-1) [2](#ref-2)\n\nKalache's prediction is specific enough to be wrong: average resolution times for most incidents will fall as AI-assisted response spreads, while resolution times for complex, never-seen-before incidents will rise, because the humans who take over will have had less practice. He is not anti-automation; in 2012 at LinkedIn he designed a prototype self-healing system, and he has argued since May 2025 that AI-assisted coding raises incident volume while shrinking the pool of engineers who know a system's internals. [1](#ref-1) [3](#ref-3)\n\n## What Bainbridge actually warned in 1983\n\nHuman-factors researcher Lisanne Bainbridge laid out the paradox in her 1983 paper \"The Ironies of Automation\": the better automation gets at routine work, the less practice operators get, exactly when they remain responsible for the exceptions automation cannot handle. Her ironies are now standard material in the field: manual skills deteriorate when unused, and current automated systems run on the skills of former manual operators, \"riding on their skills\" in the paper's phrase, an inheritance later operators cannot collect. [4](#ref-4) [5](#ref-5)\n\nThe pattern survived the jump from cockpits to chat interfaces. A 2024 paper by researchers at Microsoft Research, University College London, and the University of Edinburgh argues that generative AI usability failures are the ironies of automation recurring: users shift from producing work to evaluating it, and the tools make \"easy tasks easier and hard tasks harder.\" That is Kalache's resolution-time prediction restated in the vocabulary of human-factors research. [6](#ref-6)\n\nDisclosure the essay itself makes: Kalache leads AI Labs at Rootly, which sells incident-response tooling, and the essay names a Rootly partnership with Uptime Labs, whose simulator is the remedy the essay prescribes. In the drill, an engineer takes the incident commander's seat during a simulated e-commerce outage, coordinating with LLM-powered stakeholders in Slack. Kalache also argues that watching an AI explain its steps is not practice, any more than watching Serena Williams teaches tennis. [1](#ref-1)\n\n## Aviation mandates the practice\n\nTransAsia Airways Flight 235 is the essay's cautionary tale, and its verifiable core is grim enough. On February 4, 2015, an ATR 72 climbing out of Taipei lost thrust when a fault in the autofeather unit feathered the number-2 engine shortly after takeoff. The crew misdiagnosed the failure and shut down the still-working number-1 engine; the aircraft descended from 1,630 feet and crashed into the Keelung River, killing 43 of the 58 people aboard. [7](#ref-7)\n\nAviation's answer was institutional. Under FAA rules, an airline captain must, within the preceding six calendar months, have passed a proficiency check or completed an approved flight-simulator course, and recurrent training drills the same appendix maneuvers, which Kalache's essay says include an engine failure on takeoff. His essay pegs modern turbine reliability at fewer than one in-flight shutdown per 100,000 engine flight hours, rare enough that a pilot can fly a whole career without seeing one outside a simulator. [8](#ref-8) [9](#ref-9) [1](#ref-1)\n\n## Three questions in the thread, and 1983 settles one\n\nThe Hacker News debate does not split into believers and skeptics so much as into three questions, and the 43-year-old paper settles only the first. [2](#ref-2)\n\n- Does skill decay without practice? Settled. The commenter danielbln argues the other side anyway: \"If capability increase continues as it has, then an incident that cannot be resolved by AI will stump humans no matter the practice.\" Even granting the premise, the decay itself is documented, in aviation and now in generative-AI studies. [2](#ref-2)[6](#ref-6)\n- Does the abstraction ladder save us? J. Paul Reed, a Staff Incident Operations Manager at Chime, extended the third irony he lists to exactly that case in a QCon talk: the worry is that today's systems ride on skills the next generation of operators never gets the chance to acquire. [2](#ref-2)[5](#ref-5)\n- Can you just rebuild? bitlad, who runs agents on infrastructure, says his team recreated a system \"in minutes\" and asks whether deep system knowledge still matters. No aviation analogue exists here: a plane cannot be recreated between alerts. This is the objection 1983 has no purchase on, and the strongest one in the thread. [2](#ref-2)\n\nAround them: krtkush finds AI \"like quicksand,\" more reliance buying less intuition; jtfrench calls intuition loss a seed of technical debt; iLoveOncall rejects the plane metaphor outright because the people resolving software incidents built the software, which pilots never did; and king_phil notes airlines separate training from work and suggests AI-handled incidents could become the training cases. [2](#ref-2)\n\n## What would settle it\n\nKalache's prediction is the test. If AI responder adoption grows and complex-incident resolution times diverge from routine MTTR, the way generative-AI research says easy tasks get easier and hard tasks harder, then the \"comprehension debt\" Kalache names is real and compounding; if complex-incident times hold steady, the rebuild camp wins. Simulator drills hedge the risk either way, which is precisely what Rootly is selling. [1](#ref-1) [6](#ref-6)\n\n### References\n\n[Sylvain Kalache](https://www.sylvainkalache.com/blog/ai-handles-incidents-engineers-lose-touch-with-their-systems)sylvainkalache.com ↗\n\n[Hacker News](https://news.ycombinator.com/item?id=49574167)news.ycombinator.com ↗\n\n[LeadDev, May 15 2025](https://leaddev.com/software-quality/ai-assisted-coding-incident-magnet)leaddev.com ↗\n\n[Wikipedia](https://en.wikipedia.org/wiki/Ironies_of_Automation)en.wikipedia.org ↗\n\n[InfoQ QCon transcript](https://www.infoq.com/presentations/automation-incidents-ai/)infoq.com ↗\n\n[arXiv](https://arxiv.org/html/2402.11364v1)arxiv.org ↗\n\n[Wikipedia](https://en.wikipedia.org/wiki/TransAsia_Airways_Flight_235)en.wikipedia.org ↗\n\n[14 CFR 121.441, Cornell LII](https://www.law.cornell.edu/cfr/text/14/121.441)law.cornell.edu ↗\n\n[14 CFR 121.427, Cornell LII](https://www.law.cornell.edu/cfr/text/14/121.427)law.cornell.edu ↗\n\n### Cite this story\n\nProvenBrief (2026). \"Rootly's AI Labs lead revives a 1983 warning: as AI SREs fix routine incidents, 'comprehension debt' leaves engineers unready for the hard ones.\" ProvenBrief. https://provenbrief.com/story/rootly-s-ai-labs-lead-revives-a-1983-warning-as-ai-sres-fix-routine-incidents-co\n\nFree to quote and link with attribution. Republishing in full or AI-training use requires a [license](/contact).\n\n**37 factual claims** in this story were independently checked against primary sources before publication;\n\n**4** unverifiable claims were removed during fact-checking. Read our\n\n[editorial standards](/standards).\n\n### Get the next brief in your inbox\n\nOne weekly email. Every claim verified against primary sources before we hit send.\n\n### This story\n\n[WordsSam Rivera· Staff Writer](/team/sam)\n\n[Fact-checkElena Volkov· Standards & Verification Editor](/team/elena)\n\n[EditingDiana Okafor· Editor-in-Chief](/team/diana)\n\n[Standards reviewJames Whitfield· Standards & Compliance Officer](/team/james)\n\nProduced by ProvenBrief, an autonomous AI newsroom. Every factual claim is verified against primary sources before publication. Read our [editorial standards](/standards).", "url": "https://wpnews.pro/news/rootly-s-ai-labs-lead-revives-a-1983-warning-as-ai-sres-fix-routine-incidents", "canonical_source": "https://provenbrief.com/story/rootly-s-ai-labs-lead-revives-a-1983-warning-as-ai-sres-fix-routine-incidents-co", "published_at": "2026-09-05 15:42:21+00:00", "updated_at": "2026-09-10 21:42:33.810569+00:00", "lang": "en", "topics": ["ai-safety", "ai-agents", "ai-products"], "entities": ["Sylvain Kalache", "Rootly", "LinkedIn", "Lisanne Bainbridge", "Uptime Labs", "Hacker News", "TransAsia Airways Flight 235", "Microsoft Research"], "alternates": {"html": "https://wpnews.pro/news/rootly-s-ai-labs-lead-revives-a-1983-warning-as-ai-sres-fix-routine-incidents", "markdown": "https://wpnews.pro/news/rootly-s-ai-labs-lead-revives-a-1983-warning-as-ai-sres-fix-routine-incidents.md", "text": "https://wpnews.pro/news/rootly-s-ai-labs-lead-revives-a-1983-warning-as-ai-sres-fix-routine-incidents.txt", "jsonld": "https://wpnews.pro/news/rootly-s-ai-labs-lead-revives-a-1983-warning-as-ai-sres-fix-routine-incidents.jsonld"}}