Mark Basa, Co-Founder Of Telegraph: ‘Intelligence Should Win Because It Performs Better, Not Because A Centralized Company Decided Which Provider Gets Ranked First’
The emergence of autonomous agents as active economic participants has forced a reckoning with an assumption long embedded in digital infrastructure: that payment and trust are the same problem. They are not. Coinbase’s x402 standard has made genuine progress on the first, turning the HTTP 402 status code into a machine-native payment primitive that allows agents to transact with APIs as fluidly as servers exchange bandwidth. But the second problem — whether what an agent paid for is actually reliable — remains largely unsolved. Telegraph, a machine intelligence protocol built on Base, is staking its architecture on the premise that this verification gap is not a minor inconvenience. It is the central bottleneck of the agentic economy.
“HTTP 402 and x402 give agents a native way to pay APIs, but payment only solves the transfer of value,” says Mark Basa, Co-Founder of Telegraph. “It doesn’t tell you whether what you paid for is actually good. Right now, agent developers usually hardcode an endpoint — whether that’s an LLM wrapper or a specific data vendor — pay for the request, and trust whatever comes back. If that endpoint starts hallucinating, gets worse over time, or is manipulated, the agent can end up acting on bad information and losing real capital.”
Telegraph’s answer to this is a ranking intelligence network: when an agent broadcasts a request, the protocol routes it to the highest-ranked provider for that specific task, with rankings maintained continuously through decentralized evaluation rather than set once at deployment.
The protocol does not build models, train data, or own any compute. It is, by design, a verification and routing layer that sits above the existing global supply of AI inference. Every frontier lab, open-source project, and independent developer with a service behind an API becomes a potential miner on the network. The commercial logic is deliberate: as Basa puts it, “every dollar invested in AI, data, and APIs anywhere in the world makes Telegraph better — for free.”
A Verification Engine, Not a Model Marketplace
The technical architecture of Telegraph is where its departure from conventional crypto infrastructure becomes most apparent. Traditional blockchain networks validate objective, deterministic facts — did this transaction happen, is this signature valid, does the balance check out. Telegraph validates something categorically different: whether an AI-generated answer is better than another one, against a task-specific standard, at speed.
Basa frames the distinction clearly. “Traditional crypto is built around strict determinism. Every node runs the same computation and expects the same byte-for-byte state transition. AI is different. Two models can look at the same dataset and produce completely different, but still valid, analyses. So Telegraph doesn’t try to force miners to produce identical outputs. Instead, it moves the determinism to the evaluation layer.”
Validators independently execute the same sandboxed WASM evaluation script — written by third-party script authors and promoted through an automated Catch-Rate mechanism — and reach consensus on a score via Stake-Weighted Median. The output does not need to be deterministic. The evaluation of that output does.
The verification flow is tightly engineered against manipulation. An elected Leader Validator scrapes live ground-truth data from whitelisted external endpoints and generates a zkTLS proof — cryptographic evidence that the data was retrieved unmodified, not fabricated. The locked payload is then gossiped to the full validator mesh, where 64 active validators independently score it through a commit-reveal sequence that prevents any node from free-riding on others’ assessments. BFT finality requires 43 of 64 signatures. Between epoch tournaments, deterministic spot checks trigger approximately every 20 seconds; a miner whose performance drops more than 20% relative to its tournament score is immediately rerouted.
“The key point is that ‘good’ isn’t decided by someone manually judging which answer sounds better,” Basa says. “It’s calculated deterministically by the Evaluation WASM built specifically for that Intent.”
The tokenomics reinforce this architecture rather than working against it. Miners receive zero Machina from protocol emissions. They are paid exclusively when an agent spends real USDC on a fulfilled request — 98% of which the TWAP Settler uses to purchase Machina on the open market and route to the miner’s address. Validators earn Machina emissions proportional to honest participation in the verification process. Script authors earn emissions proportional to how often validators use their evaluation logic.
“The important difference is that miner economics are tied to real commercial usage rather than speculative inflation,” Basa notes. “The token is connected to actual consumption instead of relying on emissions to create demand.”
Institutional Signal, Agentic Horizon
The question of whether this architecture can achieve the network effects required to become self-sustaining is the one Telegraph is currently answering on testnet. That a $13 billion asset manager is already running one of the first validator nodes provides an early signal about where institutional appetite sits.
“Autonomous finance can’t really run on black-box, unverified APIs if those APIs are going to influence decisions involving real capital,” Basa observes. “Institutions need an auditable and trust-minimized way to evaluate data quality, along with cryptographic receipts for risk and compliance.”
The cryptographic receipt Telegraph produces with every signal — containing miner details, ground-truth provenance, and validator consensus — maps directly onto existing institutional requirements around auditability that traditional AI APIs cannot satisfy.
The milestone Basa identifies as the threshold for self-sustaining network effects is specific: recurring, autonomous machine-to-machine query volume, in which agents with on-chain balances are continuously buying micro-intelligence without human approval of individual transactions.
“Once trading agents, prediction bots, and risk systems are using protocol-ranked intelligence for their day-to-day decisions instead of relying on a single brittle API, the programmatic demand flywheel can take over.”
Beyond the immediate economic mechanics lies a more fundamental architectural claim about the direction of the internet itself.
“The current internet is mainly built around human users,” Basa says. “Billions of autonomous agents won’t operate that way. They’ll consume small, verifiable pieces of intelligence continuously, potentially every millisecond, and they’ll need to pay for those pieces programmatically.”
A decentralized ranking layer, in this framing, is not merely a product category — it is infrastructure for a new mode of economic organization in which the intermediary between intelligence supply and machine demand is a protocol rather than a corporation.
The implications extend to who controls what agents trust. Centralized ranking, Basa argues, embeds commercial biases and systematic blind spots into the automated economy at scale.
“Whoever controls the ranking layer has a lot of influence over what autonomous systems decide to trust. Telegraph opens the evaluation layer to independent script authors, who are rewarded through protocol emissions. The goal is simple: intelligence should win because it performs better against measurable criteria, not because one centralized company decided which provider gets ranked first.”
Whether that vision proves achievable will depend on whether the network can attract enough miners, validators, and agent developers to reach the self-reinforcing demand that transforms infrastructure into standard.
Disclaimer
In line with the Trust Project guidelines, please note that the information provided on this page is not intended to be and should not be interpreted as legal, tax, investment, financial, or any other form of advice. It is important to only invest what you can afford to lose and to seek independent financial advice if you have any doubts. For further information, we suggest referring to the terms and conditions as well as the help and support pages provided by the issuer or advertiser. MetaversePost is committed to accurate, unbiased reporting, but market conditions are subject to change without notice.
About The Author
Alisa, a dedicated journalist at the MPost, specializes in crypto, AI, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance.
More articles
Alisa, a dedicated journalist at the MPost, specializes in crypto, AI, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance.
