Out-Thinking the Algorithm
Teaching the skills AI can’t.
In addition to this newsletter, I recommend some other great ones. All free. Check them out here.
Friends,
Overwhelmed with the response to Fifty for Fifty from last week. Thanks for your replies and good wishes. No, I won’t be changing the newsletter to focus on this, but I may have to do it again, given how popular it was! And, yes, I wrote most of these truths from two stints in a hospital bed. It’s funny how clarity can come when you’re at your lowest. You can find a shorter digestible version here that you can share - and please do.
On Tuesday last week, I caught myself auto-accepting Claude's first re-draft of a critical report I was penning. Without checking. Without critiquing. Exactly the opposite of what I'm about to tell you. I had to start again. It’s too easy to slip into the algorithm’s recommendations.
This week, I’d like to introduce you to Maya.
Maya, fifteen, finds a blind spot in ChatGPT's account of Indian history. Her teacher calls it her first victory over the machine.
Socrates admitted he learned love only secondhand from Diotima, yet this made him Athens' greatest teacher. Today's schools prepare students for a future they've never lived, but like Socrates, they can still teach wisdom.
Machines remix facts in milliseconds. Humans stay valuable when they can notice their own thinking, weigh consequences, and judge when to trust the model. These are the SuperSkills.
We need a principled hypothesis for education [in an AI age] based on three core frameworks.
Human-AI Complementarity: Education should explicitly map and develop capabilities where humans add distinct value alongside AI. The principle is to cultivate what machines cannot do well but can amplify when paired with human judgement.
Concentric Capability Development involves Building capabilities that reinforce each other in a nested system. Picture five nested rings: metacognition feeds ethics, ethics guides context, context sharpens synthesis, and synthesis sparks creativity.
Process Over Product: In a world where AI generates endless content, the human process of inquiry becomes more valuable than specific outputs. The principle is to assess the thinking journey rather than just the final result.
My SuperSkills Thesis
SuperSkills differ from soft skills: they are meta-capabilities built for human-AI teamwork. These are areas were need to consider [not a complete list]:
Metacognitive awareness: tracking how you think while AI processes information. Noticing your own leaps.
Ethical judgement: applying human values to algorithmic patterns. Challenging hidden bias.
Contextual intelligence: navigating cultural nuance where machines remain blind. Spotting missing perspectives.
Critical synthesis: questioning assumptions across disparate knowledge domains. Connecting unlikely patterns.
Creative recombination: forging conceptual leaps beyond spotting patterns. Imagining unseen possibilities.
As I mentioned two weeks ago:
Knowledge is common. Judgement is rare.
The World Economic Forum's Future of Jobs 2023 report, which looks ahead to 2027, puts analytical thinking, creative thinking and AI literacy at the top of its skills table. Ninety per cent of executives in LinkedIn's 2024 survey still say human-centred capabilities outrank pure technical know-how.
From Theory to Practice
For me this means educational institutions must become SuperSkills cultivation centres. There are some that are already leading the way:
In Finland, Saunalahti school uses personalised portfolios tracking thinking processes rather than test scores.
In India the American School of Bombay integrates extended-reality (XR) tools through paired prompting, teaching students to treat AI as a studio partner.
In the US, High Tech High conducts exhibition critiques where students receive live feedback from domain experts, rewarding reflection over shortcuts.
In South Korea, Sunae High now grades essays on "reflection logs" in a 2024 ministry pilot.
In Uruguay, Plan Ceibal's community-based support programmes close the digital competence gap for low-income families through collective learning.
Fifteen-year-old Maya keeps a "thinking log" alongside her history project. After receiving AI-generated research on post-colonial India, she notices something off. "The AI pulled mainly from Western sources," she writes. "When I asked about Gandhi's Salt March, it missed how women participated." She cross-references with local oral histories her grandmother shared, identifies the cultural blind spot in the data, and revises her project brief to incorporate multiple perspectives. Her teacher notes how she's mastering contextual intelligence: seeing what the AI cannot.
Sarah, a mid-career nurse in Manchester, demonstrates that SuperSkills span all ages. She uses an AI symptom-checker alongside her assessment. When the AI suggests a common diagnosis for a patient with chest pain, Sarah notices it's missed subtle indicators in the patient's history. She documents her reasoning, the AI's limitations and her decision pathway in her continuing professional development log, earning credit while strengthening metacognitive awareness that could save lives. A missed nuance could have sent the patient home.
I recently received a GPT-crafted message for my birthday. No effort was made to make it personal; I was just a transaction. Safe to say, my estimation of this friend has dropped….significantly. I had guilt in potentially accepting what an AI has written, but this friend had none.
Across the UK, the share of job adverts mentioning AI skills has risen roughly seven-fold since 2012, with finance and ICT in the lead. There’s more.
PwC's 2024 AI Jobs Barometer finds UK roles that blend AI fluency with problem-solving pay about 14 per cent more than comparable posts. In some markets the premium tops 25 per cent, while its economic impact study projects up to £232 billion in additional GDP by 2030. Delay could cost UK £55 bn a year in lost productivity by 2030.
Lightcast's latest dashboard shows finance now tops the league table for vacancies that pair data skills with human judgment.
Deloitte's study found recent graduates show a concerning shortfall in teamwork and communication skills. PwC's study similarly cited soft skills gaps, with both firms implementing additional training to address these gaps.
And as one big tech exec captured it perfectly in reply to my knowledge is no longer power piece:
"We don't need people who know things. Our systems know things. We need people who can question what our systems tell them and make meaning from uncertainty."
Boards now cite ethical AI oversight in annual reports. Early case studies suggest that adding a human bias review step reduces flagged incidents in internal model audits despite the small sample sizes. This dovetails with EU AI Act requirements, turning SuperSkills from a nice-to-have into a compliance asset. As AI handles computation, humans must master judgement. Hire for judgement, practise it daily, and reward it publicly.
A Common Sense Media survey reveals that only 26 per cent of parents have experimented with ChatGPT, while 50 per cent of teenagers already use it for schoolwork. The gap is even wider in the UK, where Ofcom reports that 80 per cent of teenagers aged 13-17 use generative AI tools.
If you have kids “co-prompt” with your child for fifteen minutes a week. Discuss what the model gets wrong. Simple interventions like this can begin closing the digital guidance gap while broader initiatives, such as Ceibal's community support programmes in Uruguay, scale the approach. You could also co-prompt with peers who are not yet up to speed, and share your best prompts for a specific work task, or use libraries that those generous enough to do so give away for free.
As I build out SuperSkills, I’m asking questions like:
Do you notice when AI gives you incomplete information? (Metacognitive awareness)
Can you identify when AI outputs reflect problematic values? (Ethical judgement)
Do you spot when AI misses cultural context important to a situation? (Contextual intelligence)
Are you able to question AI-generated content using knowledge from other sources?(Critical synthesis)
Have you combined AI suggestions with your own ideas to create something new?(Creative recombination)
The Path Forward
When Maya next checks, she'll log that difference. Education's new mandate acknowledges fundamental uncertainty while providing direction: pair intelligent tools with wiser people. SuperSkills make the pairing work.
You can start with simple experiments:
Teachers can run a five-minute "think-aloud" where pupils narrate how they query ChatGPT, then critique a bias they spot.
Employers can add a single "assumption check" prompt in interview case studies.
Parents can swap roles, let the child critique your prompt for flaws.
These approaches must extend globally. In Kenya, Eneza's mobile service now pilots a low-bandwidth AI tutor that nudges rural learners to justify each answer, building critical thinking even on basic handsets. Brazil's SENAI now requires an "AI-reflection stamp" on apprentice projects.
In schools that implement metacognitive instruction, reading comprehension for Year 8 pupils can improve by up to 16 percentile points, according to the Gregory & Cahill meta-analysis on explicit strategy teaching. Clear thinking lifts scores even when content stays the same.
SuperSkills matter at every age: in classrooms, boardrooms and households.
It’s time to teach SuperSkills, steer the code, shape the century.
Stay Curious - and don’t forget to be amazing,
Here are my recommendations for this week:
One of the best tools to provide excellent reading and articles for your week is Refind. It’s a great tool for keeping ahead with “brain food” relevant to you and providing serendipity for some excellent articles that you may have missed. You can dip in and sign up for weekly, daily or something in between - what’s guaranteed is that the algorithm sends you only the best articles in your chosen area. It’s also free. Highly recommended. Sign up.
Now
The frontier firm is already happening: Frontier leadership begins with a clear shift: humans provide direction, and agents deliver scale. The transformation lies not in applying AI to existing systems, but in redesigning the very structure of work. The most forward-thinking organisations build around intelligence from the ground up. This approach lifts people out of repetitive tasks and places them where judgment, creativity, and strategy create real value. Microsoft report
The 7 Healthiest Snack Foods You Can Find at the Supermarket, According to Registered Dietitians. Here’s your checklist:
Choose snacks with less than five grams of added sugar per serving
Opt for snack foods with zero grams of trans fat and no more than three grams of saturated fat
Keep the sodium content to a minimum by picking snacks with less than 200 milligrams of sodium per serving
Select snacks with as few mystery ingredients, like additives, preserves, and artificial ingredients, as possible
Look for fiber content of at least five grams per snack serving
Five grams or more of protein per serving often indicates a healthier snack pick
At least three grams of unsaturated fat is an added bonus
Try to include a fruit or vegetable in your snack (or find one containing them) for optimal nutritional balance
Also: 10 Anti-Inflammatory Foods You Should Be Eating, According to Registered DietitiansStop being ‘too nice’ at work, says psychologist - this is what successful people do to be more genuine, trustworthy - British psychologist Tessa West argues against excessive workplace niceness, showing how artificial positivity backfires by making people seem insincere. Instead, she advocates specific feedback focused on behaviours rather than general impressions, starting with neutral topics to build trust gradually and transform overly pleasant workplace cultures.
I Track My Health Data Nonstop — Is This A Problem? experts warn excessive tracking may foster obsessive behaviours. Though wearables provide autonomy, psychologists caution against rigid goal-setting and loss of flexibility.
There’s no such thing as ‘background music.’ Here’s how your playlist affects your brain - Music isn't background noise, according to this article. It engages attention, memory and emotion centres. Use "purposeful passive listening" for productivity: lo-fi for emails, ambient for brainstorming. Build musical habits by pairing specific songs with tasks. Your brain creates dopamine associations, turning music into mental shortcuts for focus, energy and concentration.
Next
AI Friends according to Mark Zuckerberg - Mark Zuckerberg thinks your next close friend may be synthetic and he isn’t worried. He split future AIs into two breeds: hard-nosed assistants that fetch facts, and “companion-focused” agents that joke, console and rehearse awkward talks with your boss or partner. Meta already sees users asking its model to script difficult conversations; he expects that loop to deepen as personal data trains ever more “spontaneous, funny” avatars. Why the optimism? Loneliness. Zuckerberg cites Meta’s long-run metric: the average person lists fewer than three real friends, yet would like roughly fifteen. If software can fill part of that twelve-friend gap, he argues, it meets a genuine social need. Designers who outlaw such use-cases typically “haven’t worked out the value” people see in them. His caveat is attention, not intimacy. Glasses and headsets must “get out of the way”, surfacing AI only when summoned. Cluttered holograms that reward-hack users are a non-starter. Future Codec Avatars will supply the missing half of human communication, gesture and micro-expression, that text chat lacks. Today’s virtual confidants are still animated JPEGs, but demand is clear. If companion AI can bridge the friendship deficit without hijacking our focus, society may soon count machine friendship as rational, even healthy. Today we are wondering whether something was written by AI - tomorrow we will worry whether our virtual friends are.
Project Ludicrous - Will nuclear energy power the AI boom? -Silicon Valley giants are betting a half-trillion dollars on AI data centres powered by nuclear energy. Helion, backed by Sam Altman, promises fusion reactors by 2028 but faces transparency concerns. Microsoft and others are reopening Three Mile Island and pursuing small modular reactors, though scientists warn scaling nuclear takes years.
World’s Best Edtech Companies - Long list of who is changing the education system, but not necessarily in the way that I see it [see above]
An Age of Extinction Is Coming. Here’s How to Survive. - Technology creates an extinction bottleneck threatening human culture, relationships and survival. Virtual substitutes, from dating apps to AI companions, distract us from real-world connections while nations depopulate. Only deliberate choices to preserve traditions, families and communities can ensure humanity survives this technological transformation. Choose intentional action over passive digital consumption.
Tooth Successfully Grown in Lab - Scientists have successfully grown a human tooth in the laboratory using bioengineered materials that mimic natural tooth development conditions. The King's College London breakthrough could offer patients biological alternatives to fillings and implants, using their own cells to regenerate teeth that naturally integrate with existing jaw structure.
Free your newsletters from the inbox: Meco is a distraction-free space for reading newsletters outside the inbox. The app has features designed to supercharge your learnings from your favourite writers. Become a more productive reader and cut out the noise with Meco - try the app today
If you enjoyed this edition of Box of Amazing, please forward this to a friend. If you quote me, or use anything from this newsletter, please link back to it.
If you are interested in advertising in this newsletter, please see this link