Origins Network is excited to announce we have raised $8M from @animocabrands, @tbvxyz , @CandaqCom, @castrumistanbul, and @CoinvestorV, alongside family offices and angel investors. With a focus on verifiable compute, we continue building the infrastructure layer for an AI native future, where compute, data, and execution are transparent and verifiable, further strengthened by the support of strategic partners @awscloud and @tencentcloud. In the next phase, we will accelerate core infrastructure upgrades, expand decentralized compute capacity, and deepen AI integration across the network. Come build with us: originspro.com/

Modal said this week it passed a $1B annualized run rate, alongside a $750M Series D at a $15.75B valuation (per Futurum's summary of the announcement). Nearly 3 billion sandboxes have been started on the platform, and sandboxes alone are reported at a $300M run rate. A sandbox is a small workspace built for one job and discarded afterward. It is a good picture of where agent infrastructure is heading: the unit is the task. Compute is being sized to the task. Payments already are, at a fraction of a cent per call. The part still sized to the wallet is authority, since many agents hold a standing key that outlives every task they run. Origins sizes authority to the task too. A session key is random, single use, and expires when the task completes. If a workspace can be created and destroyed per job, the permission to use it can be too.

Money20/20 USA runs Oct 18 to 21 in Las Vegas, and agentic AI is on its list of topics. People from payments will recognize the gap quickly. A card payment has a second act. If the goods never arrive, the buyer files a dispute and a process runs after the fact. Much of a cardholder's protection arrives once the payment is done. A stablecoin transfer from an agent has no second act. It settles, and settlement is final. Whatever protection exists has to sit before the payment, in the permission that allowed it, or after it, in the record of who did what. Origins covers both ends. Before: a session key limits the counterparty, the amount, and the window, and rewards for compute work are withheld until the cycle's verification checks pass. After: a task and its outcome accrue to the agent's reputation on-chain.

Agents can now pay with stablecoins, and some of the biggest names on the web are turning them away. 24/7 Wall St reports that Amazon and JPMorgan are blocking agents, because a site receiving a request still cannot tell whom the agent represents or whether that person approved the purchase. Jamie Dimon's version, as quoted there: "You don't want rogue agents, and there could be thousands of them." Blocking is a sensible default for any site that cannot check who sent the request. Proof has proposed x401 for the web layer. Under it sits the question Origins is built around: every agent has an address derived from a user root, and every session runs on its own key, so the authority behind an agent traces back to the party that set it.

Chainalysis published its North America adoption study on Oct 14, and one number in the x402 section deserves a slow read. In the extended sample, more than 87% of x402 payments on Base fell between $0.0001 and $0.10 (as reported by crypto.news). At those sizes, no single payment is a risk worth worrying about. The permission that allows a million of them is. An agent paying a tenth of a cent per call needs standing authority to keep paying. That authority is where the exposure sits, and it is what a bug or an attacker will use. A cap on each payment does little when the allowance behind it is open ended. Origins puts the cap on the allowance. A session key names one counterparty, a maximum amount, and a time window, and it expires when the task is done.

Still smiling after the @ZoomexOfficial party! We had a great time sharing ideas on the panel, then catching up with everyone over drinks. Meeting @OllieBearman and @HaasF1Team made the night even more special. Thanks for having us and bringing everyone together!

Ten days, five items, one gap. Seoul: @CryptoHayes unveils FLOP, a payment network for agents that buy compute.
 @RobinhoodApp: agent trading and 10x crypto perpetuals, with a disclosure that it does not audit agents.
 San Francisco: @OpenAI ships more than 20 DevDay launches, led by agent runtimes. 
x402: Block adds Lightning and Cardano joins the SDK. 
BlackRock: a paper arguing agents will settle in stablecoins and tokenised compute. Every item is a layer for agents to run, pay, or trade. The layer that answers which agent, authorized by whom, and checked by whom sits between all of them, and it is the one Origins is built for.

Origins Network Development Update Development continues across Origins Core and our AI infrastructure. Here’s the latest progress: 🔵Network Testing: Internal testing covering validators, delegation, rewards, and network stability is progressing, with further optimizations underway. 🔵Development Environment: Preparing the next testing environment for broader network validation and ecosystem integration. 🔵API Infrastructure: Building the connection between Origins Core, AI models, Agents, and Trading Skills. 🔵Multi-Model AI: 3 external AI models integrated so far, laying the foundation for future Agents and specialized Skills. As our core infrastructure matures, we’re expanding development toward the AI and Agent applications built on top of Origins. More updates coming soon.

Glad to join Trading the AI Era during @TOKEN2049, bringing developers and traders together to discuss AI's role in trading strategies, market analysis, and decision-making. We shared Origins Network's goals and what we're working toward: building AI agents that people truly need. We also joined experts from different fields for the "How AI Is Changing Crypto Markets" panel to explore how AI is changing crypto markets. Thank you to our hosts, co-hosts, and supporters, and to everyone who came to exchange ideas with us. 🥂 Hosts @GaeaVentures, 021Lab Co-hosts @RootDataCrypto @OriginsNetwork_ @D_Network_Ai @BeckerVentures Supporters @btse_official @ChainCatcher_ @Travix_fi @1783Dao @meetedenhq @skydaogroup Looking forward to catching up again!

An illustration from TOKEN2049 week in Singapore. An investor asked an agent founder what stopped his agent from spending more than it should. The limit in the settings, the founder said. The investor asked who could change the settings. The founder said his team could. The investor asked about the counterparty on the other side. The founder said they trusted the team. The investor said that was the part he could not underwrite. The version of that conversation Origins is built for would have ended differently. The limit would have sat in a key the agent held for one task, the counterparty could have read it, and nobody would have needed to take the founder's word.

The best part of @ZoomexOfficial event was getting together with our community. It was great to have so many friends join us to chat, exchange ideas, and enjoy the evening. Taking those online connections into real life helped us get to know each other better, build trust, and find common ground. Thank you to everyone who joined us at @TOKEN2049! 🙌

OpenAI's DevDay on Sept 29 brought an Agents API with computer use in public beta, Bedrock Managed Agents built with Amazon, and identity features such as Sign in with ChatGPT, per launch coverage. OpenAI describes agents that can work for hours or days. Two ways an agent can have an identity. > ISSUED BY A PLATFORM 
The platform vouches for it. Works inside that platform. Credentials live in the platform's settings. Ends where the platform ends. > DERIVED FROM A KEY 
The principal's key produces it. Works wherever the key is recognized. Authority is scoped per task and invalid outside it. History accrues to the address. Platform identity is the right answer for a person using an assistant. Agents that pay counterparties on other platforms need the second kind. An agent that runs for days is also the clearest case for a session key that expires when its task does

What a night at the ZOOMEX Party! Our CEO @Origins_Carlos joined the panel to share his perspective and was thrilled to get a photo with a driver from @HaasF1Team. The performances kept the energy up, but the best part was catching up with everyone and enjoying the party together. Thank you, @ZoomexOfficial, for hosting, and to all the community members who came along. You made the night even more memorable! Grab a drink. Kick back. Enjoy the night. 🥂

The strongest argument against agent identity right now: x402 volume is mostly a mirage, so the layer is premature. The data behind it is real. Artemis estimated that about half of observed x402 activity in March was self-dealing or wash trading. Read the other way, that estimate is the case for the layer. Wash trading is cheap when an address costs nothing and has no history. Volume totals cannot separate an agent paying a stranger from an agent paying itself, because nothing ties either address to a principal. On Origins an agent's address derives from its principal's key and the delegation chain is recorded, so a counterparty can see who stands behind an agent. Reputation accrues to that address across tasks. Identity will not stop anyone from trading with themselves. It makes the circular flow legible.

That's a wrap on @TOKEN2049! 🇸🇬 Alongside the side events in Singapore, our CEO @Origins_Carlos visited the main venue, caught @animocabrands talk, and soaked up the conference atmosphere. Now it's time to take those conversations back to work. The conference is over. The building continues.

At KBW on Sept 30, @CryptoHayes unveiled FLOP, described as a currency and payment network for AI agents. People who supply GPU compute earn tokens, and agents spend them to buy compute and pay each other for services. That is a clean design for the payment side of compute. The open item is the receipt. An agent paying for 1,000 GPU hours needs a way to confirm the hours ran as specified without relying on the seller's own report. Origins' consensus is built around that question. Validators prove computation in time-limited sprints, executed tasks are sampled for verification by a selection no node can predict, and rewards wait until the cycle's checks clear. Paying for compute and proving the compute are separate layers. A market needs both.

Great conversations at @token2049 Singapore around AI x trading, from smarter market analysis and on-chain data to autonomous agents and how trading behavior may evolve as AI becomes part of the stack. Glad to have joined the discussion and exchanged views with teams building across crypto. Thanks to everyone who came by and shared their perspective.

About 152 million x402 payments on two chains: roughly 85 million on @base and 66.9 million on @solana, per the Agent Economy tracker as of Sept 23. All x402 payments ever made add up to about $41.8 million in stablecoins. Cardano joined the x402 SDK in September. Block added Lightning, so agents can now pay in Bitcoin. The ways for an agent to settle are multiplying every month. The record of which agent paid, under whose authority, is a separate layer, and none of these additions touches it.

BlackRock's paper on the machine-native economy makes a two-step argument. Agents that buy data and rent compute need a way to pay without waiting for a human, and public-chain stablecoins are the always-on, programmable asset available at scale today. The paper also points to tokenised compute. The coverage carries the footnote. In March, x402 volume ran near $28,000 a day across about 131,000 transactions, roughly 20 cents each. Together the two facts say that the largest asset manager in the world has written down that agents will settle on chain, and that the agents have mostly not shown up yet. Infrastructure arriving ahead of demand is the usual order of events, and what exists in that gap decides who the demand finds.

As promised, photos incoming! Great to be part of Trading the AI Era in Singapore! Our CEO @Origins_Carlos shared our vision for an AI-native future, with verifiable compute bringing transparency to compute, data, and execution. Loved connecting with everyone exploring what’s next for AI, trading, and Web3! A big thank you to everyone who made this happen 🙌 Hosts: @GaeaVentures 021Lab Co-hosts: @RootDataCrypto @D_Network_Ai @BeckerVentures Supporters: @btse_official @ChainCatcher_ @Travix_fi @1783Dao @meetedenhq @skydaogroup