She Captures Photography

09/10/2026

Here you go - the goal of AI literacy is not competent use of AI. It is preserving and expanding people’s capacity to reason, notice influence, detect error, and correct themselves and the systems around them.

09/10/2026

when a slug gets through
I want to know:

Where did it get in?

Why did that barrier fail?

What did we miss?

What do we change?

And then I want to see whether the next slug gets through.

That’s the difference between endlessly responding to failure and actually learning from it.

So yes.

Keep slugging.

But don’t confuse slugging with the goal.

The goal is the garden.

The goal is the berries.

The goal is making sure the slugs don’t get the best ones.

And if we’re building the garden for the next generation

eventually they’re going to inherit the whole damn garden.

And I would really like the kids to get the strawberries. 🍓

Not the slugs who are eating them

09/08/2026

Before AI-LLMs, challenging your own worldview often depended on access.

You had to encounter a different person, a different book, a different teacher, a different community, a different country, or a different experience.

And some people simply didn’t get those opportunities.

If you were raised in a highly controlled religious environment, your parents might control what you read, who you talk to, what information you receive, which questions are acceptable, which authorities you trust, and what happens if you challenge the system.

Your environment becomes your epistemic boundary.

Your world tells you what is real because you may have no practical way to compare it with anything else.

The same can happen at the level of a country.

If information is controlled, distorted, inaccessible, or surrounded by propaganda, then telling people to “think critically” isn’t enough.

Critical thinking requires something to think with and something to compare against.

That’s where AI-LLMs potentially represent a genuine shift.

Not because they automatically tell people the truth. They can be wrong, biased, manipulated, or used to reinforce false beliefs.

But they can create something many people historically did not have:

a persistent opportunity to interrogate their own thinking.

You can ask:

• Here’s what I believe. Map where it came from.
• What assumptions am I making?
• What evidence would challenge this?
• What would someone who disagrees with me say?
• What information might I be missing because of the environment I grew up in?
• Separate what I know from what I was told.
• Map the incentives of the people who taught me this.

And then you can do something even more important:

Map the environment.
Then map the thinking that developed inside it.

That doesn’t mean assuming your environment was wrong.

It means recognizing that your thinking emerged somewhere.

Every belief has a history.

Every assumption has inputs.

Every worldview has boundaries.

AI can potentially help make those boundaries visible.

Previously, you often had to wait for the world to challenge your thinking.

Now, for the first time, many people can actively seek a structured challenge to their thinking whenever they are ready to ask.

That could be profoundly important for children growing up in closed systems.

Not because AI can rescue everyone or tell them what to believe. It can’t guarantee either.

But it can potentially give someone access to questions their environment would never permit.

And that’s where the idea of cognitive infrastructure becomes much deeper.

We don’t just map systems outside of us.

We map the systems that built us.

Our families.

Our communities.

Our religions.

Our countries.

Our media environments.

Our incentives.

Our fears.

Our emotional reactions.

Our information boundaries.

Because until you can see the environment that shaped your thinking, you may mistake the boundaries of your world for the boundaries of reality.

And you can’t question a boundary you cannot see.

Before the election, the Prime Minister said that China was the biggest threat to Canada in the world. After the election, he said he wants a partnership for a “New World Order” with the dictatorship there.

Now he is holding defence coordination talks with Beijing, which Canadians had to learn about from Chinese media.

No “new world order”. No military cooperation with a regime that kidnapped our people, interfered in our elections, and set up police stations on our soil. Put Canada First.

09/08/2026

It’s: What do they actually know? What evidence are they using? What relationships have they mapped? What assumptions are they making? And can the public examine the reasoning behind the position?

You can talk about China. You can advocate for trade. You can argue about national security, human rights, pipelines, energy markets, or Canada’s relationship with the United States.

But those aren’t isolated issues.

They’re systems.

And if we haven’t built the cognitive infrastructure to understand how those systems connect, we’re left with political talking points instead of shared understanding.

what AI -llms make possible for the next generation.

Not just answering questions.

Helping people map relationships, compare evidence, trace decisions, identify gaps, preserve institutional knowledge, challenge assumptions, and build new understanding on top of what previous generations learned.

A politician doesn’t have to personally see everything.

But someone has to build the system that helps us understand what we’re looking at.

The next generation deserves better than inheriting our conclusions.

They deserve the tools to understand how we arrived at them, test them, and build something better.

Before the election, the Prime Minister said that China was the biggest threat to Canada in the world. After the election, he said he wants a partnership for a “New World Order” with the dictatorship there.

Now he is holding defence coordination talks with Beijing, which Canadians had to learn about from Chinese media.

No “new world order”. No military cooperation with a regime that kidnapped our people, interfered in our elections, and set up police stations on our soil. Put Canada First.

09/08/2026

The Core Argument: “We Are Not Self-Healing Systems” This framework challenges a damaging assumption: that individuals, especially autistic people, should endlessly adapt, self-regulate, and compensate when the environment repeatedly fails them. The Central Reframe The statement “We are not self-healing systems” is not just about institutional dysfunction, it’s a human design constraint. An autistic person cannot be expected to endlessly compensate for an environment that generates: · overload · ambiguity · exclusion · sensory distress · communication breakdown · unmet needs. When the environment causes the injury and expects the individual to independently recover, the system has confused adaptation with healing. The Question Flips We stop asking: “Why didn’t the autistic person cope better?” We start asking: “What mechanisms does the environment possess for detecting mismatch, preserving the person’s signal, learning from it, correcting conditions, and preventing recurrence?” If the answer is none, repeated distress is not surprising, it’s the expected outcome. The Problem with Endless Adaptation Autistic people are often required to become their entire repair infrastructure: · detect what’s wrong · identify what they need · translate into neurotypical language · overcome communication barriers · advocate repeatedly · explain their autism every time · regulate while dysregulated · recover from environments that harmed them · remember and document patterns · seek accommodations · challenge failing institutions · begin again with the next teacher, doctor, employer, or agency. That is continuous compensatory labour, not self-healing. Crucially, the person’s ability to survive that process can conceal the system’s failure. If an autistic person manages to adapt, the institution concludes “the system worked,” when in fact the person simply absorbed the cost. The Knowledge Exists, But the Healing Circuit Doesn’t Society contains extraordinary intelligence about autism: · autistic people themselves · families and communities · researchers, clinicians, educators · support workers, advocates · Indigenous knowledge holders · governments, employers. The intelligence exists. But does the system reliably connect, preserve, learn from, and act on it? Often, no. An autistic child tells a teacher the classroom is unbearable. A parent notices shutdowns. A clinician identifies anxiety. A researcher has documented it. The school has a policy. The government has evidence. And yet none of those signals become a persistent correction to the environment. The intelligence exists. The healing circuit doesn’t. A Systems-Level Definition of Accommodation An accommodation is not a favour granted to an individual. It is part of a system’s error-correction mechanism: 1. Person encounters mismatch. 2. Mismatch generates a signal. 3. System detects the signal. 4. Signal is preserved, not dismissed. 5. System investigates. 6. Environment is adjusted. 7. Outcome is monitored. 8. Learning is retained. 9. Future people don’t rediscover the same problem. That’s what an actual learning system looks like. Without this architecture, autistic people become trapped in a cycle of repeatedly proving the same reality to institutions with no durable memory. Masking Makes Failure Invisible Masking allows a person to appear functional while paying an enormous internal cost. If a system measures only visible breakdown, it misses harm occurring upstream. An intelligent system cannot rely exclusively on crisis as evidence. It must detect cost before collapse. The Deeper Question The framework shifts the entire conversation: Not: “How do we help autistic people become more resilient?” But: “Why are we designing environments that require extraordinary resilience just to participate?” And then: “What would it look like to build institutions that could actually learn when autistic people tell them something isn’t working?” On AI and Governance AI could help preserve patterns across time: · document recurring sensory problems · track communication breakdowns · record accommodation failures · reveal contradictions between promises and reality. But AI cannot become: · the authority on what someone’s experience means · another institution that absorbs signals without changing. AI must sit inside an accountable healing architecture. Otherwise, we risk a system extraordinarily good at understanding autistic distress while structurally incapable of repairing the conditions that produce it. The Reversal of Responsibility The deeper insight: Autistic people are often treated as though the failure to adapt belongs inside the person, even when the environment has repeatedly demonstrated its failure to learn. That reverses responsibility. The measure of an inclusive system is not how successfully autistic people can endure it. It is whether the system can: · detect mismatch · understand the signal · remember what it learned · repair the environment · reduce the need for the next person to suffer to prove the same point. The Central Principle We are not self-healing systems. Neither are institutions. Neither are communities. Neither are children. Neither are autistic people. Healing requires: · feedback · signal · preservation · memory · learning · repair. Otherwise, what we call “resilience” is often just a person carrying the cost of a system that refuses to learn. 🧠

09/07/2026

the curriculum should not be designed to teach students what happened with COVID. It should teach them how to determine what happened when the evidence is contested. That distinction is the heart of it. For your framework, I think Grade 10-12, with the flagship version around Grade 11, is the sweet spot. Students are old enough to work with probability, statistics, public policy, and causal reasoning, but they’re still developing those reasoning habits. And Newfoundland and Labrador would actually be an interesting place to prototype it because you could connect it to existing provincial learning outcomes rather than creating another standalone “AI course.” I would build the unit around one question: “What should Canada have done, given what decision-makers actually knew at the time?” Not: “Were vaccines good?” Not: “Were vaccines bad?” Instead: “Given the evidence available on March 1, 2021, what policy would you recommend, and how confident are you?” Then students get progressively more information. Round 1: They receive demographic data, infection rates, hospitalization rates, preliminary vaccine efficacy data, and uncertainty ranges. They make a recommendation. Round 2: New evidence arrives. They update the model. Round 3: A new variant emerges. Update again. Round 4: Reports of adverse events appear. Now they have to incorporate those risks without abandoning the rest of the evidence. Round 5: They encounter competing sources, political claims, personal testimonies, media stories and flawed statistics. Now comes the epistemology component: “What deserves to change your mind?” That’s where your trust variable becomes particularly interesting. I wouldn’t actually give sources a predetermined “credibility score,” though. That could accidentally turn the AI into an authority deciding whose information is trustworthy. Instead, have students construct the credibility assessment. For each piece of evidence: Who produced it? What exactly does it measure? What population does it represent? How large is the sample? What are the limitations? Is the methodology transparent? Has it been independently replicated? Does it establish correlation or causation? What evidence would contradict it? How much uncertainty remains? Then AI can ask the student: “You assigned this study a high evidentiary weight. What specifically justifies that weight?” That’s a very different educational use of AI. And here’s the really powerful part: at the end, reveal the actual historical outcomes. Students compare: Their model → historical models → observed outcomes. Then ask: “Where were you wrong?” And: “Was your reasoning bad, or did you make a reasonable decision under uncertainty that happened not to predict the future?” That distinction teaches something most adults never learn. Good reasoning does not guarantee a correct prediction. And conversely: A correct prediction doesn’t necessarily demonstrate good reasoning. That is enormously important for democratic society. The Allison example becomes the case study. You could literally begin the unit with the statement you showed me. Give students the claim: “Canadians who believe they were injured following COVID-19 vaccination deserve to have their experiences heard.” Then ask them to separate the propositions hidden inside it. Proposition A: Some Canadians report adverse experiences following vaccination. Proposition B: Some adverse events were caused by vaccination. Proposition C: The frequency of those events is X. Proposition D: The vaccines’ benefits exceeded their harms at the population level. Proposition E: Particular policies produced better outcomes than alternatives. Those are five different empirical questions. Students then discover that a personal testimony can be highly relevant to A without being sufficient evidence for B, and evidence for B doesn’t automatically establish C, D, or E. That’s the reasoning muscle. And this is where I think your attunement concept actually fits beautifully. An attuned system doesn’t dismiss the person saying “something happened to me.” It also doesn’t tell that person “therefore your explanation must be correct.” It does both: “I hear your experience.” and “Let’s investigate what the evidence can actually tell us.” That’s a remarkably good model for civic discourse generally. So I wouldn’t frame this curriculum as AI literacy alone. I’d frame it as something closer to: AI-Assisted Evidence & Civic Reasoning. AI becomes the student’s reasoning partner, while the student remains responsible for the judgment. That is much more ambitious than teaching kids how to prompt ChatGPT. And honestly, much more important.

Great conversation - link to video in comments