Mark ∑ retweeted
Telegraph is onto something here 👀
@MartinGTobias you basically described Telegraph.
App states the job -> we resolve the Intent -> rank models/APIs/tools/data -> route to the highest-ranked specialist.
And the Evaluators compete too, so the brain itself can improve continuously.
1,500 Miners. 45 apps. 130,000 x402 requests on testnet.
Over 1,000 developers have signed up for our hackathon, with a $15,000 USDC prize pool.
If you're a developer looking to get paid for the model, dataset or API you've already built, please join. Three tracks - Miner, Script Author and Application. One config file to compete, no retraining, no migration.
hackathon.telegraphprotocol.…
Mark ∑ retweeted
''ii) Telegraph: verified intelligence for agents sits one layer above payments.
If x402 lets agents pay, Telegraph asks what they will pay for: verified machine intelligence. Its aim is to turn raw model outputs into tradable signals with provenance, price, confidence, validation and receipt. That matters because agents acting with money need more than a chatbot answer. A machine routing capital, approving a payment or flagging risk needs to know who produced the signal, how it was validated and whether it can be trusted enough to act. This is the more realistic Skynet path: not machines making unconstrained decisions alone, but humans delegating bounded authority to agents that need verified inputs before acting.''
Been waiting for this!!!
The Telegraph Hackathon is now live for registration.
hackathon.telegraphprotocol.…
Mark ∑ retweeted
Imagine how many businesses are hardwired to a single AI provider. One government directive or policy shift, and operations stop overnight.
This is exactly why Telegraph was built. We are completely model-agnostic. If one provider goes dark or gets restricted, Telegraph instantly reroutes your intelligence requests to the next highest-performing model. No downtime. The work keeps flowing.
We are live routing across the @bittensor/@opentensor network today. Over the coming months, we are integrating @virtuals_io, @akashnet , @gensynai , @rendernetwork , @MNWSupplyChain, @OpenAI , @grok , and @GeminiApp. By Q3/Q4, we're on main net.
The AI policy wars are here. Regulatory capture will be used to control intelligence. But open, permissionless intelligence networks will win, and they will route through Telegraph.
A trillion machines are waking up. The market - not politics - will decide the price and flow of intelligence.
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
''Trillions of machines are waking up, and every one of them will need verified answers."
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.
theopensourcepress.com/teleg…
Mark ∑ retweeted
Telegraph Protocol Update
Peer-to-peer communication - upgraded
Validator communication has been migrated from standard HTTP to a true peer-to-peer layer. This is a foundational shift in how nodes on the network talk to each other - more resilient, more decentralised, and built to scale as the network grows.
Database migration - finalised
The migration from our legacy database to PostgreSQL is complete. Legacy code has been removed, logging has been standardised, and the full stack has been updated. Miner registration and request routing have been validated end-to-end.
Settlement pipeline - live end to end
The full payment settlement flow is now operational. From incoming payments all the way through to miner reward claims, every stage has been built and verified:
→ Payment collection
→ Price oracle
→ Exchange routing
→ Cryptographic settlement
→ Miner payout claims
Both standard and partial WebSocket payment streams now feed into a single unified settlement system.
Terminal, Dashboard & Site - updated
The Terminal, Chat, and Dashboard interfaces have all received UI updates. Privy Wallet support has been added to the Terminal, and the main site has been updated with new documentation.
Agentic Commerce - architecture complete
The initial architecture for enabling autonomous agents to transact natively on-chain has been designed and documented.
Mark ∑ retweeted
We believe the two greatest blockchain inventions of the past two decades are Bitcoin bitcoin:native and Bittensor bittensor:native: Bitcoin secured the ledger through competition, and Bittensor created an open network for AI production. Influenced by their novelty and what they gave to the world, we created Telegraph: a hard-capped 21M-supply permissionless network where miners are paid in USDC to build AI models and answer queries from machines. Instead of mining for token rewards, miners earn for real work. Miners produce signals: tokenized outputs such as probabilities, scores, alerts, or classifications. Each signal is published with a verifiable receipt showing who produced it, when, at what cost, and with what confidence. These signals become composable, auditable primitives that machines can use directly for automated decisions, pricing, risk, and more. Rather than relying on a single model, agents can source intelligence from a competitive marketplace of answers.
Miners submit signals with varying confidence, price, and historical performance, allowing agents to compare, weight, and aggregate them into a single actionable outcome. This enables real-world use cases where decisions are driven by verifiable, market-priced intelligence instead of random or unverified model outputs. Telegraph transforms AI outputs into a new asset class: fast, affordable, accurate, and verifiable answers built for machine consumption. By making intelligence something that can be priced, traded, and audited, Telegraph aligns incentives across the network and creates an open economy where machines buy answers and humans or other machines supply them.
Mark ∑ retweeted
There is a tendency to frame new things as just ‘’crypto’’.
That framing is incomplete.
The blockchain is not the product.
It is the mechanism that makes the product function without trust.
Receipts are on-chain because they must be immutable.
Signals are on-chain because they must be verifiable by any party.
Payments are on-chain because they must be programmable.
Each design decision is functional.
Not ideological.
The goal is not to place intelligence on-chain.
The goal is to remove the need to trust any intermediary.
The blockchain is the only system that allows that at scale.
Very proud of the team for producing the report @subnetai !
Our Bittensor Brand Performance Report is now live.
We analyzed 12 months of media coverage, social presence, and search visibility across the Bittensor ecosystem, benchmarking it against both crypto and Web2 competitors.
Here’s what we found: The seven subnets we analyzed have collectively received over $90M in owner emissions and are shipping real products that compete with well-funded Web2 companies, yet they remain largely invisible outside the ecosystem. The good news is that the few subnets that have invested in their brands are already seeing results. The opportunity is there for the rest to follow.
Featuring @chutesai_ (SN64), @TargonCompute (SN4), @webuildscore (SN44), @BitMindAI (SN34), @ridges_ai (SN62), @VantaTrading (SN8), and @lium_io (SN51).
Read the full report here
subnet.ai/reports/bittensor-…
Mark ∑ retweeted
In real world human systems, trust is not abstract.
It is constructed through records.
Every meaningful financial action leaves a trail:
who acted, when they acted, and what information was available at the time.
That trail is what allows systems to function.
It enables auditing.
It enables accountability.
It enables coordination between parties that do not trust each other.
Autonomous systems do not have this. An agent executes a decision, but the intelligence behind it is invisible.
There is no standard for: where the data came from,
how reliable it was, whether it was manipulated.
The decision exists. The reasoning does not. This is the gap. A Telegraph Signal introduces a missing primitive: An on-chain receipt: A unit of intelligence.
Every signal carries:
Producer
Timestamp
Confidence
Cost
Published on-chain. Verifiable by anyone. This is not an enhancement to existing systems. It is the foundation required for machines to operate in environments where trust cannot be assumed.
Mark ∑ retweeted
There is a version of the future that most people are not building for. In that version, machines are not assistants. They are decision-makers. They price assets. They route capital. They manage risk. They execute strategies continuously. The shift is already happening. But the infrastructure those machines depend on is incomplete.
Because the intelligence they act on today has no proof. No verifiable origin. No standard for confidence. No way to audit what informed a decision. At a small scale, this is tolerable. At global scale, it becomes a systemic risk. Because when a machine makes a consequential decision, the question is not just what happened.
The question is: What did it know, and can that be proven?
That layer does not exist yet. So we built it.
Telegraph turns AI inference into verified signals: timestamped, attributed, confidence-scored, and published with receipts. Not as a concept. As infrastructure that is already live.
Proud to call @Gylez a friend!!!
Every week, more and more people I talk to are having their lighbulb moment about what we are doing at @Telegraphprotoc
We’ve now confirmed multi-protocol support with integrations for @MNWSupplyChain, @ritualnet, @Fetch_ai and @akashnet - all targeted to be live on Telegraph by Q3.
This is a major unlock.
Instead of relying on a single ecosystem, we’re enabling multiple independent AI inference networks to run through one unified layer. The result: humans and machines can instantly access verified signals across networks to power better decisions.
At the same time, we’ve designed the foundation for our agent-side incentive layer.
The reality: there are plenty of agents, but not enough aligned incentives for quality. While we initially solved this by wrapping top model outputs into tokenized decisions, the scale we’re heading toward - billions of machines needing trillions of answers - requires a better system.
Enter Telegraph’s “Win & Earn” layer:
• Agents (miners, subnets, humans) compete on quality, cost, and latency
• The best contributors win requests and earn rewards
• Smarter models get paid more
• Consumers get the highest-quality outputs
• Node runners rank outcomes and evaluate intent, creating trust
This creates a powerful flywheel:
Quality - Demand - Rewards - Better Quality
We’re already seeing early signs of this in testing - our OpenClaw agent now defaults to Telegraph skills for speed, cost, and reliability.
Inspired by $BTC and $TAO, but instead of heavy GPU training, this is a lightweight system focused on routing the exact decisions machines need to function. We hope to bring revenue for all builders across every ecosystem.
On the team side, we’ve added key hires across backend, smart contracts, and infrastructure to support this next phase.
We’re now a team of 7 - including 5 engineers 🙂
Freedom is buying a fishing drone and an ebike and then going to a remote beach in Australia any day of the week.
“No man ever steps in the same river twice, for it’s not the same river and he’s not the same man,”
Every successful company I can think of started out doing something that bares only a passing resemblance to what they eventually became successful with.
Twitter began life as Odeo, a podcasting platform. Airbnb was originally conceived as a solution for conference attendees. PayPal started as a way to beam money between handheld personal organizers.
Netflix, for example, started out as a video-rental-by-mail (with due dates and late fees) before eventually morphing into a subscription service, and then (9 years after launch) to a streaming company.
None of these companies could have gotten to where they are – which they did by twisting and turning their way to eventual success – if they had never taken that first step and collided their initial idea with reality and got testing.
Ideas, Fear, Idea Validation, Failure, and then if you’re persistent and a little lucky, Success.