A Reading List and Two Habits: Staying Current in Ten Minutes a Week
The final briefing of the AI for Leaders course, and the closing post of the full AI Zero to Hero portfolio. Here is the reading list and the two habits that keep a senior leader current without drowning.
Bottom line. This is the last briefing of Course 4, and the closing post of the full AI Zero to Hero portfolio. After fifteen L-briefings giving you frameworks, decisions, and checklists for running an AI program, this one is about how to keep those frameworks current when the field shifts under them — without losing your weekends to AI news. The answer is a short curated reading list (four sources) and two specific habits (ten minutes a week). Everything else in the AI news ecosystem is optional. Read this briefing once; set up the habits; and you're done. Forward to your peers if it's useful.
This is the post you can bookmark. If the previous fourteen briefings did their job, you have the frameworks you need. This briefing is the maintenance guide.
The reading list
Four sources. That's the whole list. Resist the urge to expand it — the cost of adding a fifth is not zero, because attention is finite, and the marginal source is always lower-quality than the best four you already read.
1. The provider changelogs (10 minutes per week)
What: read the release notes and blog posts from Anthropic, OpenAI, and Google AI. Not every article — just the release notes and the "what changed" announcements. These are short, factual, and directly actionable for any company using frontier models.
Why it wins: these are the source-of-truth documents for capability shifts. When Anthropic ships a new model version, they publish a release note describing what changed. When OpenAI ships a new pricing tier, they publish a specific announcement. These are the fastest way to know about Shift 1, Shift 2, and Shift 3 from L4.1 the moment they happen.
What to skim past: the marketing-y customer success stories. Read the technical changelogs, not the press releases.
Expected output: 1-2 updates per week worth noting. Most weeks are quiet. When a frontier model version ships, the week is not quiet.
2. Jack Clark's Import AI newsletter (15 minutes per week)
What: a weekly email newsletter from Jack Clark (co-founder of Anthropic, previously at OpenAI, policy and strategy background). Summarises the week's most important AI research, policy developments, and strategic signals. Calibrated, specific, and written for readers who prioritise signal over noise.
Why it wins: one of very few AI newsletters that is (a) written by someone who is actually inside the field, (b) calibrated about what matters vs what's hype, (c) concise enough to read in a single sitting, and (d) consistently published on a predictable schedule. Free.
What to skim past: the "tech tales" section at the end (fictional vignettes — interesting but not decision-relevant).
Expected output: one or two items per week that update your mental model of the field. Over a year, this is more useful than reading 100 other sources.
3. Simon Willison's blog (20 minutes per week)
What: Simon Willison is a senior engineer who has been covering AI tooling and capabilities from a practitioner's perspective since 2022. His blog covers new tools, new capabilities, and honest assessments of what works. He's known for being one of the few people who actually uses the capabilities he writes about in real production work.
Why it wins: Simon has three rare properties as a source: (a) he ships real code with real AI tools, (b) he's calibrated — he'll tell you when something is genuinely useful vs when it's hype, and (c) he writes in plain language for a technical audience without being condescending or breathless. For a senior technical leader, this is the single best practitioner-voice source I know.
What to skim past: the deep-dive posts on specific technical details (unless you're directly affected). The strategy-level posts are the most valuable for leadership.
Expected output: 1-3 insights per week about what's actually shippable vs what's still rough. Over a year, this reshapes your sense of what's buildable now vs what's still speculative.
4. One trade publication specific to your industry (15 minutes per week)
What: whichever industry publication serves your sector and covers AI adoption in it. For SaaS, that might be something like Latent Space, SaaStr, or a specific vertical publication. For healthcare, STAT News or Health IT News. For finance, Finextra or American Banker. Your industry has one or two publications that cover AI specifically for your sector.
Why it wins: the other three sources give you horizontal AI news. This source gives you vertical context — what competitors in your industry are doing, what regulators in your sector are saying, what customers in your category are asking for. The horizontal news is useful for mental models; the vertical news is useful for competitive positioning.
How to pick: ask three peers in your industry which AI-related publication they find most useful. If two of them name the same source, subscribe to it.
Expected output: one or two items per week about your specific sector's AI dynamics.
Total weekly reading time across all four sources: about 60 minutes. That's one hour per week to stay current at the level a senior leader needs. Compared to the alternatives — reading AI Twitter, attending conferences, hiring consultants to brief you — this is the cheapest high-signal option available in 2026.
Four sources, one hour per week, decent coverage. Everything else is optional.
What NOT to read
Symmetric list, because what you avoid matters as much as what you read. These are sources that consume attention without producing decision-relevant signal for senior leaders.
- AI Twitter / X on a daily basis. High noise, low signal, emotionally draining, and built to maximise engagement rather than clarity. Read it weekly at most if you must; preferably, don't. The good content on Twitter eventually shows up in Jack Clark's newsletter or on Simon Willison's blog, filtered for quality.
- Hacker News front page. Similar problem. The AI discussions are dominated by a vocal minority with strong priors, and the signal-to-noise ratio is poor for decision-making. Useful for general tech-culture signal; not useful for AI strategy.
- "AI for executives" webinars. Most are thinly-disguised sales pitches for consulting or tooling. The useful content is available in the sources above. If a specific webinar is highly recommended by someone you trust, attend selectively; do not default to attending.
- Gartner, Forrester, and similar industry analyst reports on AI. They are usually 6-12 months behind the frontier, and their taxonomies confuse more than they clarify. Useful if your board specifically asks about a Gartner framework; otherwise skip.
- LinkedIn thought-leadership posts about AI strategy. The cost-to-value ratio is extremely poor. Even genuinely smart people write poor-quality content on LinkedIn because the format rewards breadth over depth. The 2% that's good isn't worth the 98% that isn't.
- "AGI is close" and "AI will destroy us" content. Neither helps you make decisions on a one- to three-year horizon. Both are optimised for engagement, not calibration.
- Every new AI newsletter that pops up. New AI newsletters are launched every week in 2026. Most are aggregators of the sources you already read. The cost of subscribing is low; the cost of reading is non-trivial. Be ruthless.
- Podcasts about AI strategy. Podcasts are enjoyable and inefficient. A 60-minute podcast has maybe 15 minutes of signal for a senior leader. Reading the transcript or a summary is 4x more efficient. If you enjoy podcasts as entertainment, that's fine; don't confuse them with information diet.
The principle: the AI news ecosystem is optimised to maximise your attention, not your clarity. Every source you add has a cost. Adding a 5th source to the list above is usually a net loss because the marginal signal is small and the marginal time is real. Keep the list at four.
Habit 1: the 30-minute weekly reading block
The habit: block 30 minutes once a week on your calendar — same time every week — specifically for AI reading. During that block, read the provider changelogs, skim your newsletter, check Simon's blog, and glance at your industry publication. When the 30 minutes are up, stop, regardless of whether you finished everything.
Why this works: unstructured reading expands to fill any available time. A 30-minute block trained against a closed list is dramatically more efficient than "I'll read AI content as things come up." The block also creates a ritual — something your brain expects and is prepared for — which makes the content actually sink in rather than skim past.
The specific timing that works: first thing Monday morning or last thing Friday afternoon. Monday morning is for "what changed last week that I need to know this week." Friday afternoon is for "what changed this week that I can reflect on over the weekend." Pick whichever matches your rhythm; don't try to do both.
The specific thing to avoid: reading AI content between other meetings as a filler activity. This produces the feeling of reading without the comprehension of reading, because your attention is divided. Block the time; honor it.
The secondary effect: other leaders on your team will notice the block and ask what it's for. When you explain, they'll either adopt the habit themselves (good) or mentally note that you're the person who thinks about AI strategically (also good for organisational positioning).
Habit 2: the quarterly "surprise me" review
The habit: once per quarter, set aside 60-90 minutes to review the AI-related things that surprised you in the previous quarter. A surprise is something that updated your mental model — a capability you didn't expect to see, a failure mode you didn't expect to hit, a competitor move you didn't predict, a regulatory development you missed. Write them down. Review them together.
Why this works: your mental model of the field is constantly being shaped by the specific things you notice week-by-week, but you rarely step back to ask "what did I get wrong, and what does that say about my current model?" The quarterly review is the forced checkpoint. It's the difference between a leader whose mental model keeps up with reality and one whose mental model hardens over time and drifts further and further from reality.
What a surprise looks like:
- "I thought Capability X was 2 years away; it shipped this quarter."
- "I expected Vendor Y to dominate this category; instead Vendor Z took over."
- "I assumed our internal adoption of AI tools would be faster than it was; it's actually slower and here's why."
- "I was surprised by the regulatory action in Jurisdiction W — it was more aggressive than I predicted."
- "I didn't expect our competitor to position around Axis V; they did and it's working."
Write 5-10 surprises per quarter. Over a year, you'll have 20-40 data points about where your mental model is weak. That's calibration.
The specific output: one paragraph at the end of the review that says "given these surprises, here's what I'm updating in my mental model of the field." This paragraph is the durable artifact. It's also the thing you'll refer back to the next quarter to see which of your updates held up.
The secondary effect: you become one of the very few leaders who is actively calibrating their understanding of AI. Most leaders have static mental models that were formed in 2024 or 2025 and are now out of date. You'll have up-to-date understanding, which gives you an enormous advantage in strategic discussions.
Two hours per quarter, eight hours per year. Cheap at the price.
A closing note on what the portfolio was for
This is the last briefing of the last course of the portfolio. 96 posts across four courses — AI Zero to Hero (concepts for anyone curious), AI for Builders (engineering-level how-to), AI for Product (PM-level frameworks), and AI for Leaders (exec-level briefings). Each course was written for a specific audience with a specific voice and a specific set of decisions the audience has to make. Each one stands alone; together they cover the full surface of AI literacy for anyone who wants to learn it.
The reason the portfolio exists is simple: in 2026, "I need to understand AI" is a statement with four different meanings depending on who's saying it. A curious high-schooler, a working developer, a product manager, and a C-level executive need fundamentally different things from an AI learning path, and most existing material serves one audience well and the others poorly. The portfolio is four books written together so that any user can find their entry point and get calibrated material for their specific decisions.
If you came in through AI Zero to Hero and wanted to understand how neural networks work, that course gave you the mental models. If you came in through AI for Builders and wanted to ship production LLM features, that course gave you the engineering handbook. If you came in through AI for Product and wanted to scope AI features, that course gave you the PM frameworks. If you came in here — AI for Leaders — and wanted to make investment decisions, these fifteen briefings gave you the tools.
What's consistent across all four is the voice: no hype, no doom, specific over vague, calibrated rather than certain. This is the voice I wish existed when I started learning about AI. The portfolio is my attempt to make it exist for everyone else.
What the portfolio is not: an attempt to predict the future. A specific vendor recommendation. A political statement about AI safety or AI utopia. The field moves too fast for any of those to survive long. What survives is the frameworks — the mental models, the decision trees, the checklists, the sorting tests — because those are built on the underlying structure of the problem, not on the specific technology state at any given moment. The frameworks will age better than the specifics.
What the portfolio is for: giving you a set of tools you can reach for in specific moments over the next several years. When your team asks for an AI strategy, you have L1.1's framework. When your competitor ships an AI feature, you have L4.2's matrix. When a vendor pitches you, you have L1.3's seven questions. When an incident hits, you have L3.4's six-hour playbook. These are durable tools. Use them, adapt them, throw the ones that don't fit your context, and build your own when you find gaps.
The final what-to-do-on-Monday
- Subscribe to the four sources. Do it right now: Anthropic release notes, OpenAI blog, Google AI blog, Import AI newsletter, Simon Willison's blog, your industry trade pub. Twenty minutes total.
- Unsubscribe from everything else. AI Twitter, LinkedIn thought leadership, every weekly roundup newsletter that aggregates the four sources you just subscribed to. Cut ruthlessly.
- Block 30 minutes on your calendar for weekly AI reading. Same time every week. Non-negotiable.
- Schedule the quarterly "surprise me" review on your calendar for 12 months out. Q3 2026, Q4 2026, Q1 2027, Q2 2027. 60-90 minutes each.
- Share this briefing with one peer if you found it useful. The maintenance habits compound when multiple leaders in an organisation practice them.
- Come back to this briefing in a year to check whether your habits are still in place. If they are, you're a meaningfully different leader than the one who first read this post. If they aren't, restart them — it's not too late.
And that's the end of Course 4 — AI for Leaders, and the end of the full AI Zero to Hero portfolio. Thank you for reading. Good luck out there.
The field will keep moving. Your frameworks will hold. Come back when something surprises you.
Course navigation
| ⬅️ Previous | 📍 You are here | Next ➡️ |
| ⬅️ Previous L4.2 · Decide, Defer, or Ignore | L4.3 of L4.3 | Course end 🏁 |
📚 AI for Leaders · Course Home — 15 briefings, four modules.
This is the closing post of the full AI Zero to Hero · Course Portfolio: 96 posts across four courses covering curious non-technical learners, working developers, product managers, and senior leaders. Thank you for reading. Good luck out there.
Cover photo via Unsplash. This post is the conclusion of the AI for Leaders series.
