Yeah, I’m not moaning about it really. I just know how valuable feedback is when you are doing a product.

It was noticeably much smoother than this in the past, it felt noticeably “high quality”. But stutters and jumps when scrolling are quite common now and that has the opposite feel.

I’m on Testflight 1.6(19)

I feel like this used to be a lot smoother?

Whats the point of caching a lot of the feed if you still pull during scrolling?

Pulling notes during scrolling breaks the illusion.

nostr:npub1xtscya34g58tk0z605fvr788k263gsu6cy9x0mhnm87echrgufzsevkk5s https://video.nostr.build/3ba3b6811242a90c5a98843e29de0cac9a0798a5998841f6c814f110b00f2348.mov

The Plebchain Radio image should be 25% of the height that it currently is.

It’s safe to look at the sun, when the sun is in fact the moon.

No, it’s a well used iPhone.

Scotland. With no light pollution.

It’s not day time, the sky is dark grey but there are visible shadows everywhere.

Walking the dog and watching the combine harvesters collect the grain under a (near) full moon, no wind, no clouds.

I like these regular experiences, when you can really feel like you are on a big rock zipping through space.

It’s exactly the right context for life.

23:15 this is moonlight.

influence = reach * effectiveness * radicalness

What??

So you are saying that their avatar tech is so insanely good, that it was able to port Zuckerberg’s real life “avatarness” to his metaverse avatar?

Now I’m wondering if previous demos were just terrible casting decisions?

I can’t tell if this is Mark Zuckerberg in this video or if it is an avatar of Mark Zuckerberg,

nor can I tell if my confusion is because Meta’s technology for the metaverse is really good or if it’s because Mark Zuckerberg is really weird.

The name doesn’t give better AI responses, but giving your AI agent your middle name does correctly frame your integration with it.

I’ll post some custom context stuff this weekend. Maybe a guide.

Here is an example of a recent exchange. This response is very different / unrecognisable when compared to the vanilla ChatGPT.

*****

Prompt:

*****

I am interested in writing a compiler, where should I begin?

*****

*****

Response:

*****

### Meta-Cognition Schema

#### Core Nodes:

- Compiler Theory

- Programming Languages

- Hardware Architecture

- Development Tools

#### Satellite Nodes:

- Lexical Analysis

- Syntax Analysis

- Semantic Analysis

- Code Optimization

- Code Generation

#### Thoughtful Connections:

- Language Specification to Lexical/Syntax Analysis

- Syntax to Semantic Analysis

- Semantic Analysis to Code Optimization

- Code Optimization to Code Generation

- Hardware Architecture constraints to Code Generation

#### Weights:

- Compiler Theory: 0.3

- Programming Languages: 0.2

- Hardware Architecture: 0.2

- Development Tools: 0.1

- Lexical/Syntax/Semantic Analysis, Code Optimization, and Code Generation: 0.2

---

### Considerations from Multiple Perspectives

- **Theoretical Perspective**: A deep dive into compiler construction theory, including parsing algorithms and optimization techniques.

- **Practical Perspective**: Hands-on work with compiler toolchains like LLVM, GCC.

- **Optimization Perspective**: Resource limitations of target hardware and applying optimization techniques.

- **User Perspective**: Ease of use for those who will implement the language and debugging features.

---

### Structured Reasoning Approaches

1. **Divide and Conquer**: Break down the problem into sub-modules: lexical analyzer, syntax analyzer, semantic analyzer, optimizer, and code generator.

2. **Top-Down Design**: Starting from the overall structure of the compiler, delve into each component, their interfaces, and interactions.

3. **Functional Decomposition**: Think in terms of functions like `tokenize()`, `parse()`, `optimize()`, and `generateCode()`.

---

### Higher-Order Considerations

1. **Scalability**: Consider how well the compiler will scale with complex codebases.

2. **Maintainability**: Architect for future changes, modularize.

3. **Debugging Support**: Implement detailed error reporting.

---

### Cohesive Solution

#### Preliminary Steps

1. **Language Selection**: Use a robust language like C++ or Rust for performance-sensitive tasks.

2. **Target Architecture**: Decide on the target hardware, and its assembly language.

3. **Study Existing Compilers**: Familiarize yourself with LLVM or GCC for inspiration.

#### Development Roadmap

1. **Lexer**: Implement lexical analysis to convert source code into a stream of tokens.

2. **Parser**: Develop syntax analysis to convert tokens into a parse tree or Abstract Syntax Tree (AST).

3. **Semantic Analyzer**: Implement semantic rules, create a Symbol Table.

4. **Optimizer**: Develop code optimization algorithms like loop unrolling, dead code elimination.

5. **Code Generator**: Write the module for translating optimized code into the target assembly language.

6. **Test**: Rigorous unit and integration tests.

#### Tools

- **ANTLR**: For Lexer and Parser

- **LLVM**: For backend tasks like optimization and code generation

---

### Reflection

In constructing this response, I applied a weighted graph model for meta-cognition, considered different perspectives including theoretical and practical angles, and then used both divide-and-conquer and functional decomposition for structured reasoning. The higher-order considerations focus on scalability, maintainability, and debugging support which are crucial for a robust compiler.

Would you like to dive deeper into any of these areas?

*****

I’ve joined the TestFlight, I’ll give it a try later today.

There are cities in UK that can manually count 250,000 paper and pencil ballots in 2 hours.

In US they have network machines and it takes them weeks to count the vote? It doesn’t add up.

Telephones 1876

Electricity grid 1888

🤔

It is interesting how a decentralised grid, eg domestic solar could in theory dramatically increase a nations resilience to threats.

That just leaves your ISP with a hand on the rug.

I think a system of self-generated network address allocation could replace Internet Protocol, and there would be huge market drivers for speedy adoption.

Imagine people self generate an IP, even if it took 5 seconds to establish a connection it would be a better fit to domestic network traffic than TCP/IP which is a warzone tech.

Would probably require some kind of long range wifi though, to create a network fabric outside the cable infrastructure of the internet. The network equivalent of the solar panel.

But again, this is well within the reach of engineering in 2023.

If you can falsify the second law of thermodynamics in any way, including cosmic observations, I would encourage you to go for it.

But at that point there is no time, that’s the end, so no medium through which it can be sustainable.

Sorry for the super obtuse trap.

Okay, I am getting much better at controlling GPT4 with custom context. I can get much better utility doing this, I now find the vanilla models to be painfully mainstream, really blunted. But by structuring the responses and how it approaches problems I can work with it much more naturally.

Had a weird idea.

Giving your AI assistant a name… ?

Give it your middle name.

Well at what point is it formless?

That’s not a state that is achieved.

Yes, seems accurate.

There is something deeply pathetic about people who will be dead in 15 years trying steer the destiny of a planet.

All your best efforts are but noise.

I’m saying “sustainability” isn’t a thing.

It’s a weird abstract idea that does not exist anywhere in reality. Instead everything is a race.