[ANN} An alternative foundationfor AI - not controlled by any single government

[bogdyak]

POH — An Alternative Foundation for AI
Not controlled by any single government, company, or server

════════════════════════════════════════════════════

What happens when a handful of companies control all the AI?

Right now, almost every AI system in the world runs on servers owned by a small number of corporations — most of them in one country. They decide what the AI can say, what it can see, and who gets access. Governments pressure them. They comply. You never know.

This is a problem that gets worse as AI gets more important.

════════════════════════════════════════════════════

What is Proof of Human?

POH is an AI network that runs on the computers of the people using it — not on any company's servers.

Think of it like Bitcoin, but instead of processing financial transactions, it processes AI jobs. The "miners" are regular people running a small app on their laptop or desktop. Each one runs an AI model locally. Jobs come in, get processed, results go out. No central point. No single country. No single company that can be pressured into shutting it down or changing how it behaves.

The AI is as decentralized as the network running it.

════════════════════════════════════════════════════

How it works

1. Download the miner app. Pick an AI model.
2. Your computer joins the network.
3. AI jobs arrive. You process them locally.
4. Results are verified by the network.
5. You earn POH tokens for every job.



The current live use case is the Address Scanner: submit a crypto wallet address, get back an AI-powered assessment of whether it belongs to a real person. It works today, across three production nodes in different locations.

════════════════════════════════════════════════════

Skills — Teaching the Network New Things

This is where it gets interesting.

The network is not limited to one use case. Skills are small programs that run on every computer in the network and teach the AI what to look for. Anyone can write one and propose it.

Examples:

• Detect suspicious patterns in smart contract code
• Score a transaction for wash trading
• Classify on-chain behavior for compliance checks
• Any other "look at this, tell me what you see" task

How a skill goes live:

Code:

  PROPOSE ──→ STAKING ──→ GRADUATED ──→ ACTIVE ──→ EARN
  1,000 POH   community   10,000 POH    all nodes   per job
              stakes

• You write the skill and submit it with 1,000 POH as a deposit
• The community stakes behind it toward a 10,000 POH threshold
• Once it reaches 10,000 POH, it goes live automatically — every computer on the network picks it up within seconds
• Every time your skill runs on a real job, you earn a share of the fee — forever

════════════════════════════════════════════════════

Why This Matters

Governments can pressure a company. They cannot pressure a protocol.

The problem with centralized AI is not just privacy or cost. It is that the entity controlling the AI controls what questions can be asked, what answers can be given, and who gets cut off. We have already seen this play out with content moderation, financial access, and information filtering.

Code:

  Centralized AI                 POH
  ───────────────                ───────────────────────────
  One company owns it            Network owned by its users
  One country can ban it         No single country to ban
  One server to subpoena         Thousands of private machines
  Model changes when pressured   Protocol changes by consensus
  You pay per API call           You earn by contributing

A network of thousands of computers, each running AI locally, each owned by a different person in a different country, does not have a throat to choke.

No single country can ban it. No company can be acquired and pivoted. No government can send a letter demanding changes to the model.

The AI belongs to the network. The network belongs to whoever runs a node.

════════════════════════════════════════════════════

Current Status

Code:

  ✓  Three production nodes live (independently operated)
  ✓  Address Scanner live and processing real jobs
  ✓  Skills Directory open for submissions
  ✓  Miner app: GUI installer, one download, Windows / Mac / Linux
  ✓  Open protocol — run a node, build a skill, or fork the code

════════════════════════════════════════════════════

Get Involved

Run a node: https://proofofhuman.ge
X: @poh_network
Chat: @poh_network_group
Chat: Pitch Deck

════════════════════════════════════════════════════

Questions welcome. Especially interested in hearing from people who have thought seriously about AI governance and what happens when the infrastructure for intelligence is owned by the few.

[Exitoral]

Wow, this is interesting. Very interesting. Same as Bitcoin protocol. I love this. There are some questions that I would love to ask, but this AI we have right now are almost dodging the questions. If I may ask, do you mean that we can ask it any questions and it will give us? Any. I know why I am emphasing on the word any. Current AI holds back a lot of things. Using AI to learn a skill is good but it holds out information. As you know this in our current world that information runs most of the shows. Any question, that would be nice. Freedom in the palms of my hand. Now, it will be up to us to use it for good. I know people will use it for bad als, but it's best good people get in on this fast too.  Will love to see how this later turns out. Freedom as last.



I tried running your website and this is what I got. You should look into that.

[bogdyak]

Firstly, regarding the website. It might be that you've opened it while I was publishing updates. Can you check it again now?

Secondly, my idea is to deliver the infrastructure for AI-compute, but you are free to ship any skills 

[bogdyak]

Anthropic is limiting access to Fable and Mythos.

That's what I am talking about...

AI has it's "owner", but with https://proofofhuman.ge ownership is shared with community

[bogdyak]

New Video

Yesterday, was the largest IPO in history. AI-compute is one of the most valuable resources of this decade. Infrastructure spend is reaching trillions. Orbital AI-compute by SpaceX is a great example.

The efficient path? Leverage millions of idle laptops via Proof of Human.

Linkedin: https://www.linkedin.com/feed/update/urn:li:activity:7471473999311589376/

X: https://x.com/bogidotcom/status/2065432139479343526

[rbynxx]

I think you can move this one here: https://bitcointalk.org/index.php?board=159.0

Is there any token economics for this? How does that work? Or there's an unlimited supply (POH) on it as long as your miner does a job? It's not that I'm not pro to making this decentralized but I think centralization has its own advantages as well. Yeah just yesterday Fable was disabled to non-US residents but if it could undermine national security that's another level of threat.

I think this is a novel idea and it probably will scale but I'm afraid with the way how the miner works because we know token using a model is expensive especially if you use those models that includes higher usage (e.g. Fable, Opus 4.8, GPT-5.5). Does mining have to rely on models too?

[bogdyak]

I think you can move this one here: https://bitcointalk.org/index.php?board=159.0

Let me figure out how to do it... 

Is there any token economics for this? How does that work? Or there's an unlimited supply (POH) on it as long as your miner does a job?

Thank you for the questions! Still defining tokenomics, Currently, the supply is unlimited, correct. As long as miner does a job. The way it work is miners have to simply launch an app on linux/mac/windows/server. When new AI job or task is broadcasted to the network - miners pick it up and compete to solve it first and get the reward.

It's not that I'm not pro to making this decentralized but I think centralization has its own advantages as well. Yeah just yesterday Fable was disabled to non-US residents but if it could undermine national security that's another level of threat.

Indeed, both decentralized and centralized have it's own pros and cons, but as of now I see more cons of centralization. For example, there can be some sort of sanctions that would limit AI in one or another country (just for the sake of AI race advantage). Secondly, centralized AI can be biased or prioritize truth, responses in a way that is more applicable to the stakeholders. Additionally, as you know AI wipes out a lot of jobs and again, all value from replacement lands in USA Nothing personal against USA, but they wiping out jobs not only in USA.

I think this is a novel idea and it probably will scale but I'm afraid with the way how the miner works because we know token using a model is expensive especially if you use those models that includes higher usage (e.g. Fable, Opus 4.8, GPT-5.5). Does mining have to rely on models too?

That is the best part of Proof of Human, you don't pay any subscriptions, instead every miner has local inference that is used to execute jobs. LLMs that are used are open source and can be customized. For example, I am testing mixtral, qwen, deepseek, llama

[rbynxx]

Let me figure out how to do it... 

In the bottom left corner click move and choose the right board or paste that link I just gave.

That is the best part of Proof of Human, you don't pay any subscriptions, instead every miner has local inference that is used to execute jobs. LLMs that are used are open source and can be customized. For example, I am testing mixtral, qwen, deepseek, llama

I see and thanks for answering my questions, you really know your craft. By the way, give you a merit so you can post images to make this thread more interactive and has life but I think you're good at simplicity and that's fine as well, just want to give you an option since I saw your pitch deck and it has graphics in it.

[bogdyak]

In the bottom left corner click move and choose the right board or paste that link I just gave.

Found it, moving now.

I see and thanks for answering my questions, you really know your craft. By the way, give you a merit so you can post images to make this thread more interactive and has life but I think you're good at simplicity and that's fine as well, just want to give you an option since I saw your pitch deck and it has graphics in it.

Thank you, appreciate it! Stay tuned, updating node soon.

[bogdyak]

Introducing Skills: Teach the Network Anything

Proof of Human started with one question: is this wallet a real person or a bot?

Skills let you teach the network new capabilities and extend what the AI can do.


What Is a Skill?

A skill is a small program that runs on computers across the Proof of Human network. You give it an input (text, data, a question, an image, a file, etc.), and it uses the AI model running on that machine to process it and return a result.

Skills turn the base AI into something much more powerful. Instead of being limited to what the core model knows, anyone can add new abilities on top of it.

You write the logic. The network runs it.


What Can a Skill Do?

A skill can add almost any capability you can describe to the AI. It lets developers create custom functionality for the LLM and solve specific problems.

Examples include:

• Summarize long documents or conversations
• Answer questions using private or custom knowledge
• Analyze text for sentiment, intent, or specific patterns
• Classify content into categories
• Extract structured data from unstructured text
• Reason step-by-step through complex problems
• Generate reports, code, or creative content
• Perform specialized analysis based on your own rules or criteria

If you can clearly describe what you want the AI to do, you can turn it into a skill.


How It Works

Skills run directly on each computer in the network — not on a central server. When a job is submitted, the network sends it to a machine that has the skill active. The AI processes the input locally and returns the result.

Every skill runs inside a secure container. It cannot access files on the computer, make unauthorized connections, or interfere with other processes. The output is always clean and structured.

This design keeps everything decentralized while giving developers full control over the logic.


Launching a Skill

1. Write it
A skill is a JavaScript module that exports a run(input, config) function. It receives the job input and returns a result.

2. Propose it
Submit your skill to the network with a short description and a 1,000 POH deposit. This helps keep the system serious and high-quality.

3. Get community support
Other POH holders can stake behind your skill. Once the total staked reaches 10,000 POH, the skill goes live automatically.

4. It spreads automatically
Once approved, the skill appears on every computer in the network within seconds. No manual installation or coordination is needed.

5. Earn from every use
Every time your skill processes a real job, you earn a share of the fee. The more useful and widely used your skill is, the more it earns.


The Skills Directory

The Skills Directory is the place to discover active skills, stake behind skills you believe in, and submit your own.

Each skill shows:

• What it does
• How much POH is staked behind it
• Its current status (proposed, staking, or active)
• Earnings history once it's live

Skills that lose support can be deprecated. New versions or improvements can be proposed at any time.


For Developers

Build once. Earn forever.

Create a skill that adds new functionality to the AI, propose it, and get it staked to 10,000 POH. Once live, it runs across the entire network and you earn a share of every job it handles — indefinitely.

You don't need to run servers or maintain infrastructure. You write the logic once, and the network takes care of everything else.

Browse the Skills Directory: https://proofofhuman.ge

For Businesses

Add custom AI logic across many computers — without building infrastructure.

If you need specialized AI behavior (classification, analysis, reasoning, data extraction, etc.), you can create a private skill and run it on the computers you control. Your custom logic runs locally, results come back structured, and you pay per job.

No model hosting. No API maintenance. No data sent to third-party servers.

Get in touch: https://www.linkedin.com/in/bogdan-sizov-00b736386