Telegraph

A machine intelligence protocol for autonomous systems.

[Telegraph @Telegraphprotoc]

Thank you to @theospress for writing about Telegraph 🙏

[The Open Source Press @theospress]

The AI industry has built extraordinary intelligence, but no infrastructure for machines to safely use it.

A new protocol called @Telegraphprotoc is trying to fill that gap. We took a look ahead of its Q3 mainnet launch.

https://t.co/YR22tkNUVa

[Telegraph @Telegraphprotoc]

@Jesseeckel Great piece @Jesseeckel

[Telegraph @Telegraphprotoc]

Missed one? 🥲

We orchestrate and route across all models adding a level of competition similar to Bitcoin. We built a protocol where machines request intelligence and miners compete to answer them basically transforming raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously.

https://t.co/t3uiPwnVOZ

[Telegraph @Telegraphprotoc]

Telegraph is the orchestration, routing, and payment rail for machine intelligence. Here's how it works:

- An agent sends a request for a specific signal.
- Telegraph routes it to the top-ranked miners on the Intent Leaderboard.
- Validators score the answer against real ground truth.
- The verified signal is published on-chain with a full receipt.
- The agent pays in USDC: 2% to the protocol, 98% to the miner.

Here's the part that changes everything: That 98% doesn't settle as USDC. The protocol takes the agent's payment and buys Machina (our native token) on the open market via TWAP, then pays the miner.

Every paid request is an open-market buy. No emissions to farm, no fake work - miners earn only when a real agent consumes a real, accurate answer. The better the network gets, the more demand routes through it, and every dollar swaps into Machina.

Machina follows Bitcoin's design: 21M hard cap, zero pre-mine, halvings. Fixed supply under demand that scales with machine consumption. That's the rail for the autonomous economy.

https://t.co/8ST0UempaC

[Telegraph @Telegraphprotoc]

We've built it - an orchestration and routing layer across all models, chains, users and take it further and add a level of competition similar to Bitcoin. We built a protocol where machines request intelligence and miners compete to answer them transforming raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously. Enjoy 🙂 https://t.co/t3uiPwnVOZ

[Telegraph @Telegraphprotoc]

We are sitting at the orchestration and routing point where access to universal and bench-marked intelligence becomes one click plugin, monetization, and delivery. We even take it further and add a level of competition similar to Bitcoin... It's called Telegraph, a protocol where machines request intelligence and miners compete to answer them. We basically transform raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously. Enjoy 🙂 https://t.co/t3uiPwotEx

[Telegraph @Telegraphprotoc]

We give people/enterprises the power to monteize their models where the protocol orchestrates, routes, and pays the winner when the machine's request is filled.

It's called Telegraph, a network where machines request intelligence and miners compete to answer them with a token incentive. We already have a $13B institution running a node alongside a few others.

We basically transform raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously. Enjoy 🙂 https://t.co/t3uiPwnVOZ

[Telegraph @Telegraphprotoc]

How about AI orchestration and routing for the next trillion machines? We built a protocol where machines request intelligence and miners compete to answer them. We basically transform raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously. Enjoy 🙂 https://t.co/t3uiPwotEx

[Telegraph @Telegraphprotoc]

Jack, get your agents early access to Telegraph and make them smarter. We built a protocol where machines request intelligence and miners compete to answer them. We basically transform raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously. https://t.co/t3uiPwnVOZ

[Telegraph @Telegraphprotoc]

We built a protocol where machines request intelligence and miners compete to answer them basically transform raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously. With Telegraph, an agent could do what you are talking about + all the rest in the background🙂 https://t.co/t3uiPwnVOZ

[Telegraph @Telegraphprotoc]

Agree on orchestration and routing across all models is the key! We even take it further and add a level of competition similar to Bitcoin. We built a protocol where machines request intelligence and miners compete to answer them. We basically transform raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously. Enjoy 🙂 https://t.co/t3uiPwotEx

[Telegraph @Telegraphprotoc]

The 64-node validator cap is a deliberate design decision. More nodes increase openness but also latency. 64 nodes at BFT consensus gives Telegraph a verifiable, auditable result in a timeframe that autonomous agents can actually use. Infrastructure for machines has different latency constraints than infrastructure for humans.

[Telegraph @Telegraphprotoc]

Script Authors control how requests are matched to Miners. This is not a minor role. Poorly written Intent evaluation logic creates misrouting. Misrouting degrades signal quality across the network. The best script authors on Telegraph will have a measurable impact on overall network accuracy.

[Telegraph @Telegraphprotoc]

zkTLS is not a marketing term on Telegraph. It is a specific technical requirement. Without it, a signal can be correct when produced and altered before delivery. The receipt would show accurate output. The agent would receive something different. zkTLS closes that gap. The proof covers transit, not just production.

[Telegraph @Telegraphprotoc]

@jbrukh Extremely good post. Jake, you just get it!

[Telegraph @Telegraphprotoc]

Don't enter the open-source AI space blindly.

Many networks pay miners just to exist. Telegraph pays them only when a machine actually buys a verified answer with USDC. That USDC is used to buy Machina on the open market, and the miner is paid in Machina - so every machine request is a literal token buy. No emissions to farm. No fake work. 21M hard cap. Zero pre-mine. Zero team and VC allocation. Nothing to inflate, nothing to unwind - token demand is tied to real machine-to-machine payment volume. As a miner, you build one model and can sell it millions of times - every time a machine picks it as the default to execute a task, that's another paid request. Hackathon coming soon.

[Telegraph @Telegraphprotoc]

Every major cloud provider now offers AI inference as a commodity. Per-token prices keep falling but total inference spend is rising as agentic workloads multiply. The next competitive layer isn't who produces the best output. It's who can verify, route, and receipt that output in a way machines can act on without human review. Scale needs verification infrastructure.

[Telegraph @Telegraphprotoc]

On Telegraph, what gets you ranked isn't specialization. It’s inference quality. Probabilistic dispatch rewards verified accuracy, period. A generalist with sharp, consistent outputs will outrank a specialist with mediocre ones. Build a better model. The routing follows.

[Telegraph @Telegraphprotoc]

Sign up to become a miner! We built a a protocol where machines request intelligence and miners compete to answer them. We basically transform raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously. https://t.co/t3uiPwotEx
https://t.co/2LVFcfx12V

[Telegraph @Telegraphprotoc]

Not vibe coded but years in the making (so thought to share anyway!). A protocol where machines request intelligence and miners compete to answer them. We transform raw AI outputs - from any open or closed-source model - into verified, tradable answers that machines can use to make decisions and execute tasks autonomously. 🙂 https://t.co/t3uiPwotEx

[Telegraph @Telegraphprotoc]

@harshilmathur We route both requests and payments to and from machines and humans :)