X World 2026
25-26 AUGUST • NAARM | MELBOURNE
Sammy Myers
Making Agents Useful for Apple Admins
An agent can give you a very convincing answer and still not have done the job. For Apple admins, getting useful results is less about finding a magic prompt and more about setting the work up properly: a clear finish line, access to the right live context, sensible permissions, checkpoints, and evidence you can actually inspect. This practical session looks at how to brief, supervise, redirect, and verify agents while tackling real administrative tasks. Personal-project experiments and clearly labelled workplace “what ifs” will cover Mac troubleshooting, OS update research, compliance-baseline assessment, change preparation, and safe maintenance. We’ll also break down what makes a reusable agent skill genuinely useful: when to use it, what it may access or change, how it handles failure, and what proof it must leave behind. This is for Apple administrators and support practitioners, not just developers. Attendees will leave with a six-question task brief, a simple inspect–act–check loop, and practical patterns for deciding when an agent should proceed, ask for help, or stop.
About the presenter
Sammy is a Melbourne-based Apple platform practitioner and independent sloperator with hands-on experience across macOS administration, Jamf, automation and recently agentic development and workflows. Their recent projects explore how AI agents can move beyond clever answers and become genuinely useful collaborators for real tasks. They are particularly interested in the unglamorous bits: clear boundaries, useful failures, human approval at the right moments, and evidence that the work actually finished. They also spend far too much time turning repeated workflows into tools, skills, and automations so they do not have to solve the same problem twice.
Interview
We asked Sammy a few questions about where Apple's heading and what they're working on.
What are you most excited about in Apple’s upcoming OS releases?
We’re finally getting a version of Siri that is actually going to be what Apple has been promising for years! We’re all wary of massive AI corporations having ownership of the data we’re handing over, and Apple are handing us our own personal AI tooling to do it the way we want (with the obvious caveats!).
Is there new hardware you’re hoping Apple ships?
With the RAMpocalypse not ending anytime soon, and cheap memory and storage now a distant memory (hah), I’m still waiting to see Apple ship hardware capable of running larger, more ChatGPT/Opus-level intelligence at home than we can currently.
The rise of AI: excited, scared, intrigued, or something else?
AI is probably the most overused marketing term of recent memory! If a rice cooker is “AI powered” then it’s a clear sign humanity has gone down a dark path. We sit at this crossroads of people trying to make current LLMs do too much or too little, and we need those of us who can find a balance.
Right now I’m leaning heavily toward being interested in what’s coming over the next 12 months. A lot of people are still hand-writing prompts instead of using dictation for work, and I think we should be moving to conversational AI for day-to-day interaction.
What’s the best thing you’ve built that nobody’s heard of?
RosterEase, my iOS/iPadOS app for shift workers and people who do a lot of field work (home visits, etc.). Over the past 18 months I’ve slowly been fleshing it out, and it’s my first app built entirely with an LLM: from prototypes in tldraw, to using ChatGPT imagegen for screen mockups, to realising it in a simulator.
It’s also my first real taste of how bad App Store Connect is.. though thankfully we have CLI tools like asc (https://asccli.sh/) to make it easier for our agentic friends to walk us through it, or even do it for us.
Finish this sentence: “You should come to my talk if…”
…you enjoy listening to someone with self-diagnosed AI brain-rot who sloperates problems away by spending far too much time arguing with an agent that hasn't actually aligned with the spec I gave it.. and want to understand how to get these things actually doing what they're supposed to instead of accepting "You're absolutely right!" as confirmation.
