I used OpenAI’s o3-mini reasoning engine to construct a thesis for the Russia-Ukraine conflict using the positions of Jeffery Sachs and John Mearsheimer. I then asked it to create a counterargument to those positions - the antithesis. And I finally asked for a synthesis that concludes in lasting peace in the region. In all three cases, I prompted it to only use axiomatic deductive reasoning and first principles thinking, emphasizing facts and ignoring the opinions and assertions of so-called experts. Tl;dr peace is achievable 👀 nostr:note1c8xnnu4qtp0fcn9j65djsexhqnfty4m9t0f2dg4l34h50q9qugxsa3yc0e
Discussion
Axiomatic peace? Did it ask about Bitcoin? 😂
Seems a generally well-constructed peace deal arrangement.
However, what’s the reasoning behind Zelensky (apparently) not wanting at least a cease fire? I can only assume:
- The terms of the cease were too ludicrous to be able to accept
- A cease fire makes a peace deal more likely, and for whatever reasons he doesn’t want that (e.g. corruption).
I suspect it’s the latter
Unfortunately I think so. And the UK/EU (in particular UK and France) seem eager to fill the void of the US.
We should just get done with it and elect AI instead of these losers
It’s all about the prompt and the training though. How could we ensure it reflects our principles?
Everything would have to be open source and all prompts publicly viewable. Don’t think that solves everything, but it will go a long way in building trust through radical transparency
We could open-source the “system prompt” which is what I started drafting up here (in particular the Charter of Rights and Responsibilities):
Reading now, love it! Might be a good topic to talk about on your return episode on Plebchain Radio. Are you planning on adding more to Pactum?
Perhaps elections would be voting on pull requests ;-)
The idea is to have small government that is more like a service provider. Different jurisdictions would have different forks of it. E.g. some may include education or healthcare as a service of the government (with higher tax) whilst others would exclude that.
Been toying with the idea that you would subscribe to a different “government” regardless of geography. Not thought it through yet … I think that may have been mentioned in the sovereign individual book but been a while since I read it.
I wrote this 10 years ago
The-Null-Party
- The Null Party is an apolitical organization.
- The Null Party is not a protest vote, it supports all forms of democratic government, parties and systems.
- The Null Party represents its voters and non voters everywhyere.
Goals
The Null party has one, and only one goal:
- To promote financial transparency leading to financial democracy.
Advocacy
- The Null Party proposes that all tax payers get a reciept for paying taxes broken down into understandable sections to show how the money is spent.
- The Null Party aims to provide financial transparency for tax payers, in order to help identify non essential and wasteful services.
- The Null Party campagns to allow tax payers to opt out of funding such programs while continuing to fund all the programs that are deemed necessary to realize a functioning democracy.
https://github.com/TheNullParty/The-Null-Party/blob/master/README.md
Nice, I was thinking the other day that people’s pay slips should list how the tax is being divided between departments. Would start off a very long payslip! But would help drive down the number of departments of government with time.
i think australia does something like advocacy(1) - providing a breakdown of what % of your taxes go to which spending category
Zelensky, his family, and the existing power structure will all - literally - have their heads on pikes if they lose power. Their power is seated in this conflict. The conflict continues because of the government, not because of the people. The Russia war machine knows this, the US war machine knows this. They know Ukraine will fight forever because of it. It is not always the people in government calling the shots, they are dealt the hand they have. Trump, and this administration, is a royal flush for the people. The problem is, we are dealt a new hand every election.
Wait till you find out about the Unnecesary War...
BTW Mearscheimer's thesis that all Great Powers act the same, is flawed, and lacks evidence.
That’s why I framed the analysis the way I did. Neither his, nor Sachs’ position was accepted as an absolute truth
I dont think even Mearschiemer accepts his position as an absolute truth. Just a theory for prediction, with about 80% accuracy, as he once said.
But there is an obvious thing that different cultures have different behavious when it comes to war, or great power politics.
I like your creative technique in using LLMs. this way. However the training material will have systematic bias, esp. sourced from common crawl etc. As they say, if news is fake, what is history?
Here’s one of the prompts I used. In theory, o3 being a CoT reasoning engine, should not be too contaminated by the “garbage-in” problem, especially with this type of prompt. Not perfect by any means, but about as good as it’s going to get today:
“using axiomatic deductive reasoning alone, reason through both their positions from first principles. Ignore the opinions and assertions from mainstream experts and just focus on facts and logic”
A perfect combination of human and machine working together
It’s only going to get better, IMO. We’re at the point now where TPUs are being used to design better TPUs that human’s can’t even conceive of. Step 0 of the singularity 🙂
Yes, I now write 2k lines of quality code an hour using prompt driven development. Next year it will be 10k lines. Then when we get agentic nostr going they will start compounding the output. Stretch goal is to get to 1 million lines of code a day.
‘agentic nostr’ gonna steal that.
By the way I think techniques like this will become invaluable for projects like bitcoin. Especially if we can learn to subtract biases introduced by the group think. If bitcoin development could return to its roots of logical arguments to optimize the system, rather than, advocacy and politics. We could pave a cow-path towards ossification, while mitigating major risks, such as a chain-split.
The influencers need to be culled wholesale first, but yes, I agree that’s what is needed
On a slightly tangential note, I wrote this using a similar approach a couple of weeks back. AI gave me the bullet points using deductive reasoning and I then rewrote it in my style: https://primal.net/avi/1Tb8pIKHh7y3ZL0Ctae4J
Nice work. 3 weeks is like a year in Agentic AI right now. Already things have evolved. I think wrt bitcoin it helps to align to the bitcoin epochs (ie halvings). As each epoch essentially has a new security model. Nostr was in part created to facilitate this agentic future. We have a few people working on these types of things, outside of the mainstream. I wanted to get autonomous agents into nostr basically from the start. That was a political battle that I lost. It will happen anyway. Lightning should work in combination with on-chain, one being always on, which is resource heavy, the other being more reliable, but also needing alternative infrastructure. These things are happening in different places, and I think bitcoin, nostr and lightning will thrive in multi agent scenarious with high agency. Taproot is the key to this, and it's happening slowly, right now. But hopefully people will realize the opporutnity there soon. Great piece, in any case!
Do you have any examples of this agentic nostr framework? GitHub link would work as well (I’ll just give it to AI and ask it to explain 😂)
It's evolving rapidly. And we are basically stitching together a number of different technologies, of which nostr is one useful part.
Some pointers to things we like so far:
- Nostr, various NIPs, NIP-05, NIP-05, NIP-98 ... but NOT DVMs (too complex)
- Ditto: https://soapbox.pub/ditto/
- MCP : https://modelcontextprotocol.io/
- Claude Code: https://docs.anthropic.com/en/docs/agents-and-tools/claude-code/overview
- Open Deep Research: https://huggingface.co/blog/open-deep-research
- Bitcoin and layer 2 technologies: lightning, liquid, single-use seals, taproot assets, utxo management systems, tendermint, Elements, testnet4
- Parts of the W3C stack, such as Linked Data, and wwwarch for scalling
- All the latest advances in the Agentic AI stack, models, frameworks, routers, orchestration, multi agent comms (where nostr helps).
- Agetntic Identity, including using nostr/taproot and extensions
We're not prioritizing writing and documenting a whole agentic framework and putting it on github, right now, though some things will go be released. Ultimately the agents will write the future code, and we will direct, as needed. Is just a fraction of what we're doing. Sorry, a bit of a nebulous answer, but might feed into a prompt, still :)
Awesome, will check these out!
I, on the other hand, shared a meme. 😉
Gm, Avi!