CHAPTERS
Hilary’s new reality: mom + entrepreneur, and why time suddenly matters more
Claire welcomes repeat guest Hilary Gridley back to the show and sets the context: the AI tool landscape has evolved, and so has Hilary’s life. Hilary frames her current challenge as a sharp increase in the opportunity cost of time—paired with fragmented attention as a new mom and founder.
- •Shift since last episode: from product leader/manager to entrepreneur and new parent
- •Opportunity cost of time is higher; attention is more fractured
- •Goal isn’t “hyper-optimization,” but reclaiming time for meaningful work and life
- •Why AI workflows feel newly necessary under increased constraints
The “anti-system system”: avoiding heavy setup and maintenance
Hilary explains her core philosophy: productivity systems often fail because they demand upkeep. She wants minimal configuration, fast starts, and a workflow that improves itself over time—otherwise it defeats the purpose.
- •Many AI ‘systems’ require too much setup and organization
- •Hilary’s bar: don’t create a second job maintaining the system
- •Start simple; let complexity ‘earn its keep’ later
- •Positioning on the organization spectrum (Notion-heavy vs plain-text chaos)
A metaphor for messiness: the Al Gore desktop story
Hilary shares a memorable story: seeing Al Gore’s chaotic desktop made her feel validated that “having it together” doesn’t require perfect organization. This sets the tone for building workflows that work even for naturally messy inputs.
- •Chaotic inputs are normal—even for high-achievers
- •System should tolerate mess rather than demand pristine habits
- •Reframing productivity as practical support, not aesthetic organization
Zero-AI capture layer: iPhone Back Tap shortcut for instant task capture
Before any AI, Hilary builds the simplest possible capture mechanism: a lock-screen shortcut triggered by double-tapping the back of the phone. She dictates a task (e.g., “reschedule pediatrician”) and it lands in an inbox without friction.
- •Use iOS Shortcuts: dictate text → add to reminders/inbox
- •Enable Accessibility → Touch → Back Tap → Double Tap
- •Prioritize ultra-low-friction capture for leaky-memory moments
- •Separating capture from organization (capture first, sort later)
Demo: “Plan My Day” in Claude Code—terminal as a life OS
Hilary shows how she runs “Plan My Day” inside Claude Code from the terminal. Claude pulls reminders, a preferences file, and calendar commitments to propose an actionable schedule, then helps choose what to prioritize.
- •Claude Code setup is lightweight (install + run in Terminal)
- •Plan My Day command triggers a repeatable instruction set
- •Claude reads reminders and organizes them into a markdown structure
- •Claude incorporates calendar constraints to propose a realistic day plan
Claude-managed files: markdown as the backend and Obsidian as optional UI
Hilary emphasizes the simplicity of the underlying system: a folder of markdown files that Claude edits on her behalf. Obsidian is used sometimes for writing, but the core idea is “files only”—and the human doesn’t need to micromanage the agent’s internal process.
- •A single “Claude folder” holds preferences, logs, and structured notes
- •Markdown files as a universal, durable data format
- •Obsidian is optional; Claude can operate without the user opening files
- •Mindset: treat the agent like an employee—spot check, don’t hover
Auto-learning preferences and living with small error rates
Instead of manually defining productivity preferences, Hilary lets Claude observe behavior and update constraints over time (e.g., pumping schedule, weekend childcare coverage, avoiding stimulating work before bed). They discuss the tradeoff between trusting automation and accepting occasional misses.
- •Preferences inferred from actual behavior, not aspirational plans
- •Important for real constraints: childcare, pumping, energy patterns
- •Trust-but-verify depending on task criticality
- •Accepting a small error rate can still be massively net-positive
Beating procrastination: break big life-admin into the smallest first step
Hilary explains why tasks like a baby passport stall: they’re too big to fit into daily margins. Claude helps by translating “do the passport” into a concrete, tiny step (e.g., “book the post office appointment”) that can be scheduled into a short block.
- •Overwhelm often comes from trying to do the whole task at once
- •AI reframes into a 10–15 minute “first step”
- •Calendar time-blocking makes commitments real
- •Progress compounds when small steps are consistently scheduled
The “yappers API”: talk to Claude instead of building integrations
Hilary introduces her favorite low-tech pattern: narrate what’s happening rather than wiring every app together. Keeping Claude open beside her work lets her speak updates and context as she goes, dramatically reducing integration effort and anxiety about background access.
- •Avoid premature OAuth/API wiring; narrate what you’re doing instead
- •Claude stays open in a split-screen workflow all day
- •Terminal as a multipurpose interface to reduce context switching
- •“Complexity must earn its keep” before deeper integrations
Daily notes + observation loop: logging what you planned vs what happened
Claude generates a daily markdown note containing the planned schedule and a running log of observed actions. Hilary uses this to review patterns, understand gaps between intent and reality, and iteratively improve the workflow without heavy manual reflection.
- •Daily note includes schedule + behavioral log
- •Periodic check-ins: “what are you observing?”
- •Pattern detection: time-boxing failures, priorities crowding out writing
- •Continuous improvement: learn-by-doing beats upfront orchestration
Screenshots as a universal context primitive (and workplace workaround)
Claire and Hilary highlight screenshots as a powerful low-friction way to feed context to AI when integrations are blocked. Hilary notes this is especially useful at work: prove value with “janky” workflows (screenshots + narration) before requesting deeper access.
- •Screenshots transmit rich context without complex permissions
- •Great for constrained enterprise environments
- •Prototype with ‘screenshots and prayers’ to validate ROI
- •Use proof-of-value to justify secure access later
The 10x impact framework: deciding what to automate vs keep human
Hilary shares her core decision rule: if being 10x better at a task wouldn’t yield 10x impact, automate it. If it would, invest human effort—while also considering joy and meaning, not just productivity.
- •Question: “If I were 10x better, would it have 10x impact?”
- •Automate low-leverage busywork; protect high-leverage thinking/creative work
- •Apply beyond work: protect joy and life enrichment too
- •Decompose tasks (e.g., talks): keep narrative/ideas human, automate deck labor
Career-stage nuance: what’s ‘low leverage’ depends on learning curves
They add an important caveat for managers and individual contributors: leverage changes as skills mature. Work that’s low leverage for a senior person (e.g., slide polishing) can be valuable reps for someone earlier in their learning curve.
- •Framework depends on where you are on the learning curve
- •Early-career reps can be developmental even if ‘busywork’ later
- •Manager takeaway: tailor automation guidance to growth needs
- •Don’t remove skill-building tasks before mastery is established
Lightning demos: ‘Recording On’ anonymization + building a returns-tracking skill
Hilary shows two practical “skills” enabled by Claude Code: a ‘recording on/off’ mode that anonymizes sensitive info during demos, and a returns workflow built from a simple problem statement. The segment reinforces that you can start with plain language and iterate into automation.
- •‘Recording On’ toggles anonymization without maintaining a separate demo environment
- •Potential B2B use: demo production-like data safely via redaction
- •Returns automation built by conversing: questions → script → slash command
- •Principle: start with the problem statement; let Claude propose and implement
Building the habit: one daily use to develop ‘reach for AI’ muscle memory
Hilary closes with advice for beginners intimidated by the terminal: do one thing with Claude Code every day. Consistency rewires your reflexes, making AI assistance feel natural—and reveals the terminal’s power once the initial discomfort fades.
- •Daily practice builds the ‘alien in my computer can help’ reflex
- •Terminal fear is common; familiarity arrives quickly with repetition
- •Hilary’s on-ramp: used Claude Code when Cursor limits hit, then adopted it
- •Fun and leverage increase once muscle memory forms
Where to find Hilary + closing remarks
Hilary shares where people can follow her work—primarily her Substack—plus mentions her course for managers and events for women in AI. Claire wraps up with standard show sign-off and where to find the podcast.
- •Hilary’s newsletter: hills.substack.com
- •Course for managers on using AI; events for women in AI
- •Podcast availability + request to like/subscribe/review
- •Show website: howiaipod.com
