Knowledge work demands that we stay current. New frameworks, shifting best practices, emerging research - there’s always something we should be reading. The traditional approaches all have problems:

  • RSS feeds require curation and still demand you do the reading
  • Newsletters are written for broad audiences, not your specific interests
  • Social media optimises for engagement, not learning
  • Periodic Google/Google New searches even if scheduled are limited in scope

With tech moving so fast, finding a way of maintaining visibility of the whole picture became impossible. Focus had to shift on just those parts of most value to me.

Scheduled Prompts as a Solution

Tools like Perplexity, Claude Code and CoPilot offer scheduled searches - you write a prompt, set a frequency, and receive synthesised results automatically. It’s not a new concept (Google Alerts has existed for years), but the AI layer makes it genuinely useful.

The difference is synthesis. A Google Alert sends you links. A scheduled prompt sends you a summary with sources, already filtered for relevance and recency.

Here’s an example prompt I use for FinOps:

What are the most significant developments in cloud cost management and FinOps practices from the past week? Focus on new tooling, methodology changes, and case studies from enterprise organisations.

Every Monday morning, I get a digest. Some weeks it’s thin - nothing much happened. Other weeks there’s a new framework or tool worth investigating. Either way, I didn’t have to remember to look.

Designing Prompts for Learning

The prompt design matters more than you might expect. I’ve found a few patterns that work well:

Be specific about your level. A prompt asking for “developments in quantum computing” will return different results than one asking for “developments in quantum error correction relevant to someone with a physics background.” The AI calibrates its synthesis accordingly.

Ask for contrarian views. One of my prompts explicitly requests “perspectives that challenge mainstream thinking on [topic].” This surfaces minority opinions I’d never find in algorithm-curated feeds.

Request practical applications. “How are organisations actually implementing [concept]” yields more actionable information than “what is [concept].”

Set appropriate frequency. Fast-moving fields (AI, security) benefit from weekly prompts. Slower domains (management theory, hardware trends) work better monthly.

Building a Personal Curriculum

Where this gets interesting is when you treat scheduled prompts as a learning system rather than a news feed.

I’ve started thinking of my collection of prompts as a personal curriculum. Each one represents an area I’ve decided matters to my professional development. The prompts collectively define what I want to know about, and the scheduled delivery ensures I make progress without relying on motivation or memory.

Currently I have prompts covering:

  • Cloud cost optimisation (weekly)
  • Emacs ecosystem developments (weekly)
  • AI infrastructure and MLOps (weekly)
  • Management and leadership research (monthly)
  • Scientific computing trends (monthly)

The monthly prompts are particularly valuable. These are areas where I want awareness - I don’t need to track every development, but I want to notice when something significant shifts.

The Compound Effect

The real value emerges over time. Key is not the development of deep expertise (something that requires significant and sustained focus and action) - it’s informed awareness. But informed awareness is often exactly what knowledge workers need. You don’t need to be an expert in everything; you need to know enough to ask good questions and recognise when something is relevant to your work.

Limitations and Caveats

This approach has obvious limitations. The AI can miss important developments, hallucinate significance where there is none, or surface the same information repeatedly. I treat the digests as starting points rather than authoritative summaries.

There’s also a risk of false confidence. Reading a weekly digest about quantum computing doesn’t make you a quantum computing expert, but it can make you feel like one. I try to stay honest with myself about the difference between awareness and understanding.

Finally, this only works for domains with regular public discourse. Niche topics with limited online discussion won’t generate useful results.

Getting Started

If you want to try this approach:

  1. Pick 3-5 topics you want to track
  2. Write prompts that specify your level and what kind of information you want
  3. Set frequencies that match how quickly each field moves
  4. Review and refine prompts after a few weeks - you’ll learn what works

The setup takes perhaps an hour. The ongoing time investment is whatever you choose to spend reading the results. For me, that’s about 30 minutes on Monday mornings.

Closing Thoughts

I’m increasingly convinced that the value of AI tools lies in these small, unglamorous applications. Scheduled prompts won’t transform your career, but they might help you stay informed with less friction. In a world of information overload, that’s worth something.

The goal isn’t to read more - it’s to learn steadily without the cognitive overhead of remembering to learn. Let the robots do the searching. Save your attention for the thinking.