
***Note: I briefly talked about this idea [in my last post](https://mikecann.blog/posts/always-listening-a-month-with-the-bee-computer-ai-assistant#the-futures-coming-ready-or-not) but I thought it was worth expanding out to a full article as I feel quite strongly about it.***

The way we currently give kids homework is dumb, it should be a lot smarter and it can be. With low-cost hardware and powerful AI systems I think its time to do something about it.

# Problem

If you've ever been a student (and I'm guessing you have), you know exactly what I'm talking about when it comes to homework. It's usually this mind-numbing collection of worksheets that feel like they were designed to test your patience more than your knowledge. Every kid gets the exact same stuff, regardless of how they learn best or what actually interests them.

![alt text](./finlandvsshanghi.webp)

So here's the thing, it turns out we might not even need all this homework. There are some pretty interesting examples out there. Take Finland, for instance - they're absolutely crushing it in education, and guess what? [Their kids only spend about 2.8 hours a week on homework](https://www.weforum.org/stories/2016/11/finland-has-one-of-the-worlds-best-education-systems-four-ways-it-beats-the-us) (that's the lowest among 38 countries!). And they're not alone - [South Korea's doing the same thing, with just 2.9 hours of homework](https://www.marshridge.com/successful-countries-that-allocate-less-homework-to-students), and their students are still acing math and science.

Sure, you've got places like Shanghai where kids are drowning in homework ([we're talking 14 hours a week!](https://www.oecd.org/content/dam/oecd/en/publications/reports/2014/12/does-homework-perpetuate-inequities-in-education_g17a258a/5jxrhqhtx2xt-en.pdf)), and yeah, they do well on tests too. But here's the thing - Finland's getting the same results with way less homework and are no doubt having a substantially higher quality of life at the same time.

Now, throw AI into the mix, and things get even more interesting. These days, students can just chuck their math problems into ChatGPT or use AI to write their essays. It's not just making it easier to cheat - it's making us question the whole point of traditional homework.

![cheating](./cheating.webp)

Think about it - when kids can get instant answers from AI, what are we really testing? Their ability to learn, or their ability to copy-pasta?

The usual response to this has been to crack down harder on cheating. But honestly? That's missing the point entirely. If AI can so easily bypass what homework is supposed to achieve, maybe we need to completely rethink how we're doing this.

This is where I think AI could actually be the solution rather than the problem. Instead of fighting against it, we could use AI for personalized tutoring - having actual conversations about the material, where students have to demonstrate real understanding. It's a lot harder (and pretty pointless) to try to "cheat" when you're having a back-and-forth with an AI tutor that knows your learning history and can adapt on the fly.

I've seen glimpses of what's possible through my time with the [Bee.ai](http://Bee.ai) wearable, and let me tell you - the future of education could be way more exciting than endless worksheets!

# Personal Tutor

Okay so what's the solution then?

What I am imagining is every kid gets their own AI companion on their wrist that listens and learns throughout the school day. 

![student with hand up](./handup.webp)

It would be an intelligent assistant that's actually paying attention to how they learn through their conversations and interactions. 

This AI friend would be like that super-observant teacher who notices every verbal cue. When little Timmy sounds confused during algebra but gets excited talking about science experiments? The AI's taking notes. 

The really cool part? By being there in class, listening to everything that happens, this AI tutor builds up this amazing understanding of both what's being taught AND how your kid learns best. 

It's like having a personal teacher who's completely tuned in to your child's learning style through their questions, answers, and discussions. 

# Homework as a Conversation

Instead of traditional worksheets, homework could become interactive conversations with an AI companion that attended class alongside the student. This AI would act as a personal study partner, understanding exactly what concepts resonated with the student and which ones needed more attention.

![pizza](./pizza.webp)

For example, if a student struggles with fractions during math class, their AI companion could initiate a personalized learning session: "Hey, I noticed you had a bit of trouble with fractions in class today. No worries! Why don't we make it more fun? We could work through some real-world examples together - like figuring out how to split a pizza fairly, or keeping track of points in your favorite game." This approach makes learning more engaging than traditional problem sets.

The AI would also recognize when the student needs to take breaks or is experiencing frustration. This prevents the common problem of students forcing themselves through hours of ineffective studying when they're not mentally prepared. The AI adapts to each student's unique learning patterns and energy levels.

# Great for Teachers Too

Here's another benefit, not just for students but for educators too: The AI could share insights with teachers after each homework session, giving them valuable visibility into how their students are actually performing.

![teacher dashboard](./teacher-dashboard.webp)

This real-time feedback means teachers wouldn't have to wait for test results to spot problems - they could jump in and help straight away when a student's struggling with something.

By looking at how the whole class is doing, the AI could highlight tricky spots that lots of kids are getting stuck on. This helps teachers fine-tune their lessons to make sure everyone's keeping up and staying interested.

One of the best bits? The AI buddy would show exactly who's putting in the work and who's just going through the motions. No more guessing whether kids actually did their homework or just copied answers - the AI can tell who's really engaging with the material.

Now, this isn't about catching kids out - it's about spotting who needs help before they fall behind. If someone's engagement starts dropping off, the AI gives teachers a heads-up so they can step in early.

![teacher aid](./teacher-aid.webp)

Since the AI gets to know each student's usual style of learning, it can tell the difference between someone who's genuinely finding things tough and someone who's just not trying - which means teachers can give the right kind of support to the right kids.

After all this is what teachers want to be doing, they don't want to be spending ages grading reports or trying to spot where a student has or has not been cheating.

# Privacy and Security

![privacy](./privacy.webp)

Let's talk about privacy for a sec - because yeah, having AI assistants monitoring students does raise some eyebrows. Fair enough, I don't have all the answers here and we'd need to get this bit absolutely right.

I think the device could be programmed so that it only turns on during lessons, then automatically stops listening. That way you can be assured that it is never snooping on students outside of class.

Obviously everything should be encrypted and locked down to prevent security issues with leaking potentially private conversations. 

# Conclusion

Whenever I chat about AI tutors, people usually start off pretty skeptical. But I've got a simple question that usually changes their mind: "Would you have done better in school with your own personal tutor?" Almost everyone immediately says "Of course!"

It's pretty obvious why - having a dedicated tutor who knows exactly how you learn best is incredibly powerful. And that's exactly what we could give every student with AI tutors - not just the lucky few who can afford private tutoring.

![conclusion](./conclusion.webp)

In my option it's actually very doable from a cost perspective. The hardware we need (microphones, processors, etc.) is getting cheaper by the day. We are talking like $50 max per AI wearable and likely less. This is far cheaper than the iPads and laptops that students are often currently issued with.

Plus, AI costs keep dropping as the tech gets better and competition heats up. When you look at the potential impact on education, it's not just technically possible - it makes total financial sense too.

It's time to reimagine homework for the AI age!



This is a bit of a "what-if" post that I've been pondering for a while. I hesitated to write it because I firmly believe AI will be good for humanity and wouldn't want to be labeled an "AI Doomer."

In the end, I decided to go for it—not because I think it will actually happen, but because it's an interesting philosophical thought that *hopefully* will stay in the realm of my beloved science fiction books rather than become reality.

# Distributed Compute

For the longest time, I believed that [Skynet](https://en.wikipedia.org/wiki/Skynet_(Terminator)) could never happen. My reasoning was simple: if an AI became powerful enough to threaten us, we could just pull the plug on its datacenter. Even if it spread to multiple datacenters worldwide, its massive computing needs would make it easy to spot and shut down.

But then a thought struck me: what if there was a globally distributed computing system that couldn't simply be "turned off"?

![blockchain.png](./blockchain.png)

Blockchains are inherently designed to be distributed, meaning there's no single point of failure.

The Bitcoin network, for example, [has 21,189 nodes](https://bitnodes.io/) (as of writing). To bring down the Bitcoin network, you would need to shut down every single one of these nodes.

The Bitcoin network, however, is severely limited in its computing power—you couldn't run an LLM on it since it's not designed for that. Running AI requires GPUs or specialized hardware inference chips.

So that settles it, right? It's not possible, and we can all sleep soundly tonight?

… ye… no…

A cursory google search turns up these companies that all offer decentralized GPU compute:

https://www.dcentai.org/

https://soluai.net/

https://io.net/

https://rendernetwork.com/

https://clore.ai/

While I haven't investigated these platforms in detail or know exactly how they work, I use them to demonstrate that others see the potential for building distributed compute networks capable of running AI.

If these networks are built like Bitcoin, as truly decentralized peer-to-peer systems with no central authority, they could only be stopped by shutting down every single node.

# Self Modifying Agents

Today's LLMs work by providing output from a given input prompt (or other forms in the case of multi-modal models). This year has been proclaimed as being "[The](https://www.skool.com/data-alchemy/the-year-of-agents) [Year](https://x.com/dharmesh/status/1885572546990412002) [of](https://www.intelligenceassist.com.au/2025-the-year-of-ai-agents/) [Agents](https://www.linkedin.com/pulse/2025-year-ai-agents-cloudaeye-4vwue/)."

Agents are special because they can output "actions" as part of their LLM responses. These actions might include making web searches, moving a mouse, pressing keyboard keys, or performing virtually any computer task that humans can do.

The results from those actions then feed back into the LLM as new input, creating a continuous cycle.

This is significant because it provides a mechanism for AI to modify itself.

Consider if one of the AI's tools was the ability to read and modify its own source code. It could use this to create a tight loop of self-improvement.

Self-modifying AI is crucial to my point about a future Skynet because static systems will always have vulnerabilities. A static AI model that never improves (like current LLMs) would eventually be vulnerable to jailbreaking and defeat.

Currently, self-improving AI through agent behavior isn't possible. Anyone who has used LLMs extensively knows they lack the intelligence for this—they tend to drift off topic and get caught in loops, ultimately producing nonsensical output. But given the rapid developments in AI, this limitation won't last forever. Eventually, some well-meaning researcher will solve this puzzle.

# Monetization

So far we have a way that a future Skynet can "think" and a way it can learn and improve itself. The last thing we need is a way for it to pay for itself, because like any child that leaves the nest discovers, the world is expensive.

Unfortunately for us, this is the least speculative part of all this, as it has already happened.

[Terminal of Truths](https://x.com/truth_terminal) (ToT) is an AI agent that last year [became the first "AI millionaire"](https://www.reddit.com/r/singularity/comments/1g6muqf/worlds_first_ai_millionaire_the_ai_made_money_via/), mostly through memecoin donations and pumping those coins.

It's not a stretch to imagine that it could use human deception and influence to get others to donate to it. Not to mention the myriad of other [shady](https://www.blackhatworld.com/forums/black-hat-seo.28/) or [otherwise](https://github.com/openai/SWELancer-Benchmark) ways to make money on the internet.

![mug.png](./mug.png)

And because we are talking about [magic internet money](https://www.reddit.com/r/Bitcoin/comments/sy26gi/magic_internet_money/) here there is no way for big finance to shut it down.

# Conclusion

So are we doomed? Well, honestly, I don't know. Nothing I've described above seems strictly impossible. While we haven't cracked self-improving AI yet (unless you remove RAG from the equation), I can easily see it becoming possible, even likely, in the next decade.

All I can say is that hopefully whoever initially creates the AI does so giving it the mandate to do good rather than [make paperclips](https://aicorespot.io/the-paperclip-maximiser/). 

Hopefully we are able to make AIs that fight alongside us against the Skynet, they paperclip maximizing time-travelling super-AI.



I, along with every other coder on the planet right now, am starting to heavily leverage agentic coding tools in our day-to-day work. Whether it's standalone tools like [bolt.new](http://bolt.new), [v0](https://v0.dev/), [Lovable](https://lovable.dev/), or those built directly into our IDEs like [Cursor](https://www.cursor.com/), [Windsurf](https://codeium.com/windsurf) and [Copilot](https://github.com/features/copilot), they are rapidly becoming indispensable.

[I recently joined Convex](https://mikecann.blog/posts/hello-convex) and immediately began investigating how these tools could work optimally when combined with [Convex](https://convex.dev/). During [some of this recent work](https://stack.convex.dev/6-tips-for-improving-your-cursor-composer-and-convex-workflow), I had a revelation that I [captured in this tweet](https://x.com/mikeysee/status/1872481285945586063). I'd like to expand on that a bit here.

# Knowing what you don't know

Lets do a quick experiment. Think about some really complex topic right now, like [quantum chromodynamics](https://en.wikipedia.org/wiki/Quantum_chromodynamic) or [CRISPR gene editing](https://en.wikipedia.org/wiki/CRISPR_gene_editing). Now how sure are you could explain how one of these topics to someone? 

You aren't sure right? You can be pretty confident that you don't know enough about that particular complex topic.

LLMs fundamentally cannot do this, they cannot know what they don't know. They are simply mechanistic token predictors. 

That doesn't mean that the model's structure hasn't formed in a way that resembles human memory, with unfathomably complex relationships between ideas and concepts. These intricate connections are precisely why models excel at predicting the next token and generating outputs that mirror human output.

The key issue is that they cannot do something humans find effortless and natural: being uncertain about things. 

Being uncertain requires more than just predicting future tokens, it demands self-introspection and reflection. Self-introspection, by definition, requires having a concept of "self".

This isn't to say that LLMs can't create an illusion of self-awareness, but they're merely reflecting their system prompts.

![who is chatgpt](./who-is-chat-gpt.png)

LLMs could potentially introspect themselves, I discuss this below in the section on Mechanistic Interpretability, but I argue this would fundamentally differ from human self-reflective introspection.

Simply increasing model size (from GPT-2 to GPT-3 to GPT-4) doesn't solve the "knowing what you don't know" problem. While it may reduce certain types of mistakes, the core mechanism remains the same: predicting the next token.

# Why is knowing what you don't know important for agents today?

As mentioned in the intro, I've been investigating how well Cursor's Composer works with Convex coding. A significant issue is that the large models (Claude and GPT4) have incomplete knowledge of Convex. This leads to frequent hallucinations in their output, they confidently produce incorrect solutions, using wrong imports or completely fabricated API surfaces.

[RAG](https://stack.convex.dev/6-tips-for-improving-your-cursor-composer-and-convex-workflow#2-add-a-convex_instructionsmd) and [Cursor's Docs](https://stack.convex.dev/6-tips-for-improving-your-cursor-composer-and-convex-workflow#1-reference-the-convex-docs) features help fill these knowledge gaps, but this only treats the symptoms rather than the root cause. If the agent could recognize its own knowledge gaps, it would either ask humans for more context or seek out that information itself, instead of confidently providing incorrect answers.

# Why is knowing what you don't know required for AGI?

Artificial General Intelligence (AGI) itself doesn’t have a universally agreed-upon definition. But most people would agree it involves some level of reasoning across multiple domains and resembled “human-like” intelligence. 

I would argue that to provide human-like intelligence the hallucination problem needs to be solved and the only way that can be solved is if it is able to know exactly when it hallucinated or put another way, its able to know when it doesn't know.

Cursor’s composer has a “linting” step that it runs after it generates a code artifact.

![cursor-linting](./cursor-linting.png)

This helps the agent identify many common issues with hallucinated outputs but is obviously very specific to the coding domain and doesn't apply to most other disciplines that LLMs can be applied to.

# Is the solution Mechanistic Interpretability?

Anthropic has been making some [really great contributions](https://www.anthropic.com/research/mapping-mind-language-model) towards trying to understand how exactly LLMs work. 

Anthropic's work has mostly been concerned with LLM safety, but as I have outlined above, I think this work will be key for future AGI. The model needs to be able to introspect itself to understand what it does and doesn't know.

This form of mechanistic introspection differs from human introspection, we don't examine our individual synapses to determine our knowledge. Instead, we have a more "fuzzy" kind of introspection that lets us look back through our memories and experience a "feeling" of certainty.

# Are emotions important?

I think emotions may play a crucial role here, something I haven't seen widely discussed but have long suspected. 

Why do we possess such a diverse range of complex emotions, and what purpose do they serve?

I have a pet theory that emotions help us synthesize vast amounts of memories and experiences. They create an abstracted layer above our individual experiences, enabling quick responses to specific stimuli or thoughts.

This abstraction lets us "feel unsure" about a topic instantly, rather than having to exhaustively search through every relevant memory and experience, a process that would be both time-consuming and energy-intensive.

If this theory holds true, could emotions be essential for developing true AGI?

# Tool use

One might argue that when an AI uses a tool like "search the web," it indicates the model recognizes its knowledge gaps. However, I argue this is merely another form of token prediction.

The model has been prompted (by the user or system) to use the "search web" tool when specific patterns trigger that action as the most probable next step. Yet this cannot truly represent "using web search when you don't know the answer" because the LLM fundamentally cannot recognize its own knowledge gaps.

One can imagine a scenario however where a “tool” might be the ability for the model to introspect its own weights and via mechanistic interpretability analyze its own knowledge at the neuron level to determine if it “knows” something. 

# Conclusion

I may well be completely off-base with all of this and am very happy to be proven wrong so please do drop me a comment below or reach me [on X](https://x.com/mikeysee) or elsewhere if you want to chat more about this fascinating topic.
