Calling all AI Engineers: the Telegraph Hackathon is open.

There is no universal #1 model on Telegraph. Each Intent has its own supply, evaluation and ranking because fraud detection, weather forecasting and code review are different jobs. “Best” only makes sense relative to a task.

A dataset and a frontier model can compete in the same intent and be directly comparable, because what's measured is the output against the intent's Evaluator, not the architecture that produced it. Heterogeneous supply, one comparison.

Google had a link graph. Machine intelligence has no equivalent universal signal. Telegraph measures performance per Intent, then lets the measurement itself face competition so the ranking can improve as the market changes.

Price routers optimize price and latency. Telegraph ranks measured performance per Intent. A model can top a general leaderboard and still be mediocre at the task that actually matters.

An app integrates one endpoint and inherits every improvement that happens behind it. Not just new Providers joining - better Evaluators too. The measurement that decides who serves your requests keeps getting sharper, and you get the result without changing a line

A general model can be better overall and still lose a narrow task to a specialist. Telegraph ranks the task, not the brand. If your model, API or tool is exceptional at one thing, that is enough to compete.

If a machine can call it and its output fits an Intent, it can compete: model, API, dataset, search system, algorithm or tool. Telegraph is deliberately agnostic about how intelligence is produced. The market cares about measured performance for the task.

A model can top a general benchmark and still be mediocre on your job. Telegraph ranks performance per Intent. The useful question is not 'what is the best model?' It is 'what performs best for this task, now?'

Telegraph is hiring a Robotics Lead. web3.career/robotics-lead-te…

A ranking without demand is just a leaderboard. When real paid demand follows measured performance, the ranking becomes economically meaningful. The ranking becomes the market. The full thesis ↓

Intelligence changes. Models get updated, APIs improve, new specialists appear and yesterday's leader can fall behind. That is why intelligence cannot be graded once. Telegraph continuously re-measures performance per Intent as the market changes.

Most AI rankings make the intelligence compete. The benchmark stays fixed. But benchmarks can become stale, gameable or wrong. Telegraph makes the measurement compete too. The reasoning behind it ↓

Diamonds can be graded once. Intelligence can't. Models improve. APIs degrade. Data becomes stale. Prices change. New competitors appear. So the market has to keep grading. Telegraph turns intelligence into a continuously graded commodity. Ahmed wrote the thesis behind why we built it this way ↓ xitter.cf/1xahmedali/status/2107…

Our new whitepaper is live. Here's the TL;DR: Machine intelligence is becoming abundant. Knowing which intelligence to trust is not. Thousands of models, APIs, datasets, search systems, algorithms and tools can increasingly answer the same machine request. The problem is no longer simply producing intelligence. It is continuously determining which intelligence is best for the job - without appointing someone permanently to decide. That is what Telegraph is built to do. Telegraph is a peer-to-peer ranking protocol for machine intelligence. Every task resolves to an Intent. Miners compete to supply the intelligence for that Intent. But Telegraph does something different at the next layer: The judge competes too. Independent Evaluators compete to determine how that intelligence should be measured. A challenger can replace the current Evaluator by outperforming it and passing hidden verification on material it has never seen. Validators reproduce the committed evaluation process and finalize the resulting rankings. So there is no permanent winner on either side. A better model, API, dataset, algorithm or tool can enter and rise through the ranking. A better method for measuring them can enter and replace the old one. The intelligence evolves. The measurement evolves. The ranking evolves. And agents don't have to keep rebuilding their stack to follow it. They integrate with Telegraph, request the intelligence they need, and the protocol routes against the latest finalized ranking for that specific task. Weather intelligence competes with weather intelligence. Search with search. Risk with risk. Fraud detection with fraud detection. Many intelligence markets. One protocol connecting them. Then demand closes the loop. Consumers pay for intelligence in USDC. Higher-ranked Miners receive greater opportunity to serve that demand. Miners receive zero protocol emissions. At genesis, 98% of fulfilled paid Consumer USDC is attributed to Miner settlement. That USDC is aggregated on Ethereum and used to acquire Machina for Miner compensation. Useful intelligence earns because machines actually bought it - not because the protocol emitted tokens at idle supply. Meanwhile, Machina emissions finance the infrastructure that keeps the market working: Validators verify and finalize it, Evaluators improve how intelligence is measured, and the Treasury supports the protocol. And then the cycle starts again. Demand attracts better intelligence. Better intelligence competes for demand. Better Evaluators improve how it is measured. Better rankings redirect demand. New economic opportunity attracts more intelligence and better evaluation. The market measures itself again. That is the Telegraph flywheel. Not a static leaderboard. Not one model. Not one benchmark. Not one company deciding what intelligence wins. An ever-evolving market for machine intelligence. And the same mechanism can expand across new Intents as machines need new capabilities - more intelligence providers, more Evaluators, more applications, more agents, more machines - without Telegraph having to build those systems itself. Machina sits underneath that economy. 21,000,000 maximum supply. Four-year halvings. No premine. No team allocation. No VC allocation. Bitcoin created an open competition for producing blocks. Telegraph creates an open competition for producing - and measuring - machine intelligence. The intelligence competes. Even the judge is a market. telegraphprotocol.com/whitep…

Vitalik's "Cryptographic World Computer" post describes almost exactly the role Telegraph was designed for. Ethereum as the secure core underneath larger decentralized systems. High-frequency work happens outside the chain, while Ethereum keeps the state, settlement and guarantees that need a globally secure base. He even describes "a stronger decentralized layer in the middle between users and a chain." Telegraph applies that model to machine intelligence. Intelligence is harder than deterministic computation. Cryptography can prove a computation ran correctly. It can't prove which of five forecasts, risk scores or search results is the best intelligence. So Telegraph splits the problem three ways: - Models, APIs, datasets and tools compete inside task-specific Intents to supply intelligence - Evaluators compete to decide how that intelligence should be measured - Validators reproduce the committed evaluation and finalize the rankings The intelligence competes. The method for judging it competes. The protocol verifies the process. Even the judge is a market. That keeps intelligence off Ethereum's hot path. Agents consume Signals at machine speed, with no Ethereum transaction per request. The economic settlement anchors back to Ethereum. Agents pay in USDC. Miner-attributable USDC flows into Telegraph's aggregated Ethereum settlement, acquires Machina for Miner compensation, and cumulative entitlements are committed through the protocol. If machine demand grows through Telegraph, Ethereum isn't just hosting another app. It becomes the economic root of an open market for machine intelligence. Verification is a natural convergence point too. As Ethereum's proof and privacy stack matures, more of Telegraph's verification process can become succinct and Ethereum-native. Ethereum doesn't need to execute the machine economy to be its foundation. Ethereum as the secure core. Telegraph as the competitive intelligence layer around it. @ethereum @ethereumfndn