The Good Tech Companies - The AI Economy Needs a New Engine: Inside Ambient’s Radical Vision

Episode Date: April 12, 2025

This story was originally published on HackerNoon at: https://hackernoon.com/the-ai-economy-needs-a-new-engine-inside-ambients-radical-vision. Travis Good, co-founder of Ambient, explains how his AI-powered blockchain project is reshaping decentralized computing using verified inference, Proof of Work. Check more stories related to web3 at: https://hackernoon.com/c/web3. You can also check exclusive content about #web3, #travis-good, #behind-the-startup, #decentralizing-ai, #ambient-decentralized-ai, #settled-knowledge, #verified-inference-system, #good-company, and more. This story was written by: @ishanpandey. Learn more about this writer by checking @ishanpandey's about page, and for more stories, please visit hackernoon.com. Travis Good, co-founder of Ambient, explains how his AI-powered blockchain project is reshaping decentralized computing using verified inference, Proof of Work, and economic utility in this HackerNoon interview.

This audio is presented by Hacker Noon, where anyone can learn anything about any technology. The AI economy needs a new engine, inside Ambient's radical vision, by Aashan Pandey. In this exclusive edition of our Behind the Startup series, Aashan pandes it's down with Travis Goode, co-founder of Ambient, a decentralized AI network that's redefining how computation, verification, and blockchain consensus come together. From building mathematically optimal freight systems to pioneering AI-driven biotech efficiency, Travis now turns his focus to decentralizing AI at a time when centralized giants dominate the narrative.

In this candid and technically dense conversation, Travis explains why ambient rejected industry assumptions, why proof of work is making a comeback, and how AI inference can fuel a new, verifiable internet economy. A'Shawn Pondy. Hi Travis, welcome to our Behind the Startup series. You started AT Harvard, worked on optimizing complex systems in biotech and transportation, and now you're leading an AI-powered blockchain company. Can you walk USTHROOUGHTHAT journey? What drove your shift from traditional AI applications to BUILDIINGA decentralized AI network? Travis Good.

Thanks for having me, Ashan. I'm what you might call a deep tech contrarian. I am attracted to difficult domains wherein the experts claim that the correct approaches are settled knowledge. This has been the case in three fields as you've mentioned, biotech, transportation, and now blockchain. In biotech, I was in a meeting with a major pharmaceutical company when they were proudly stating that they had tested 2 billion compounds to replace a particular pesticide using a

completely brute force approach featuring lots of people and machines working 24-7. All I could think at the time was, that is so wasteful. At the company where I was CTO, we sought to radically reduce the need to screen so many compounds, using machine learning and computation alchemistry. Investors didn't like it at first because of the costs but ultimately the bet paid off because it produced huge productivity gains and the whole industry now takes that approach. Similarly, in transportation, I was told by experts that rules engines, which lacked performance guarantees but could work fast, and genetic algorithms, which lacked both performance and timing guarantees, were the best the freight industry could hope to achieve. My skepticism about that led me down a research rabbit hole that ended up with me recruiting some mathematicians affiliated with the University of Rome as

consultants who ultimately helped us build the world's first mathematically optimal freight railroad movement planner, which worked fast, had performance guarantees, and actually improved the whole railroady network's velocity. Crypto AI, which I have watched since 2017, has consistently made a couple of assumptions that I think are worth critically examining. The first is that, marketplaces of models, are the best for network economics, innovation, and performance. The second is that proof of stake is the best way to secure networks that feature a lot

of mining. I strongly disagreed with both of those ideas, but they kept being repeated by chain after chain. Then, centralized I entered the scene with, OpenAI, and Anthropic, and I began to really worry, because no one was building what I thought would be a viable decentralized economic alternative. I felt a moral obligation to enter the fray, so I created Ambient with my co-founder Max. Ambient is built counter to the two assumptions I mentioned, it's focused on delivering a single model really well using proof-off work.

Ashant Pandey. Ambient recently secured $7.2 million in funding LED by A16 Zee's crypto startup accelerator, along with Delphi Digital and Amber Group. INVESTORSTYPICALLY look for strong economic models and defensible modes. What aspects Ofambian's revenue model and long-term economic sustainability made this ACOMPELLING bet for them? Travis Good. I don't want to put words in anyone's mouth, but my belief is that our investors recognize that Ambient solves a fundamental economic problem that emerges as AI becomes the backbone of the global economy. Our revenue model revolves around what we

call useful proof of work, a system where miners earn both inflation and transaction-based rewards by performing verified AI inference that users actually need and pay for. Unlike traditional cryptocurrencies where mining creates no direct utility, Ambient miners produce economic value with every transaction. This creates a virtuous cycle as more users need AI inference, more miners join to provide it, strengthening the network and increasing its utility. The defensibility comes from three main sources. First, our technical innovations in verified inference. We've achieved verification with just 0.1% overhead

compared to competitors 10-1000x overhead costs. Second, our focus on standardization, by optimizing for one high-quality large language model rather than fragmenting resources across many models, we dramatically improve miner economics. And third, network effects, as more miners join, latency decrease sand performance increases, making the network more attractive to users. From a sustainability perspective, Ambient creates an economy where the currency is directly tied to the most valuable economic resource of the coming era, machine intelligence. As AI becomes essential to nearly every business function, Ambient provides both the computational infrastructure and the financial framework to support this transition.

The most compelling aspect I think was our ability to demonstrate, not just theorize, these innovations, for instance, at the time we raised our seed round we'd already implemented our verified inference system on models up to 400 billion parameters in size. Ashaan Pandey. Raising capital in the current market is challenging, yet AMBIENTSUCCESSFULLY secured backing from A1-6C, Delphi Ventures, and Amber Group. WHATKEY lessons did you learn from the fundraising process, and what advice WOULDYOU give to other I and blockchain startups looking to attract top-tier investors? Travis Good. Regarding lessons from the fundraising process, I think I learned that the more thoughtful technical diligence a fund does, the more impressed

and excited I get, because that effort showcases a fund's own technical capabilities, insights, and willingness to engage with the project. For example, I particularly appreciated that a 16z CSX brought a full panel of distinguished academics to vet our proof of logits in depth with us. The other lesson I learned is that not everyone has a thesis on everything. Swa's important and necessary to align with funds who have a thesis on your project's area and whose thesis aligns with your project. We've done well on that front with all of our funders, I think, and that feels good. Regarding my advice? I think intention matters.

Ambient's mission is to deliver decentralized AI at scale, to address the fundamental economic problems that centralized AI creates, while giving miners the best possible economic deal by bringing back and modernizing proof of work. We want to help not just our network, but serve web 2 and many other blockchains with the fastest, cheapest verified inference on the best open weights model. I strongly believe our investors resonated with that mission and wanted to invest in the innovations we proposed to accomplish it. My simple advice to other founders would be to not trend chase, to undertake projects that you sincerely believe in and think will change the world.

Ashant Pandey. How do you plan to allocate the new capital across R&D, infrastructure, and scaling efforts? Are there any technical milestones or go-to-market strategies this funding will accelerate? Travis Good. We're allocating capital across three main priorities, with technical development taking the largest share. First, approximately 60% is going toward R&D to perfect our verified inference system and enhance our fork of Solana. We want verified inference to be a seamless experience for miners large and small. Similarly, we'd like our API Toby as useful as possible for developers,

so we're working hard on the Solana portion. Second, about 25% is dedicated to infrastructure and testnet development. We're retargeting a testnet launch later this year, which will allow miners and developers to experience the platform firsthand. This includes building developer tools, documentation, and SDKs to make integration as frictionless as possible. The remaining 15% supports our go-to-market strategy, which has two tracks, miner recruitment and developer engagement. On the mining side, we're creating tools and resources to help GPU owners transition to ambient mining. For developers, we're focused on demonstrating compelling use cases for verified and chain AI. If you're a potential miner,

developer, or developer platform, like eNagentic Framework Provider, we'd love to hear from you. This funding accelerates several key milestones including our testnet launch, our onboarding of our first cohort of miners, our completion of our cross-chainbridge for interoperability, and the development of our public-facing API gateways. and the development of our public-facing API gateways. Aashan Pandey. AI inference at scale is a complex challenge, especially WHENBALANCING security, decentralization, and cost efficiency. How does Ambin's architecture solve the trade-offs between computational efficiency ANDBLOCKCHAIN verification?

Travis Good. At the heart of Am ambience architecture is our proof-of-logits system, which represents a fundamental rethinking of verified inference that eliminates the traditional tradeoffs. Most verification approaches force a binary choice. Either sacrifice efficiency for security, like ZK Snarks with 1000x overhead, or sacrifice security for efficiency, like optimistic verification systems. Our innovation was recognizing that the

fingerprint of an AI model's thinking, the raw numerical outputs called logits, could be used to verify model execution with minimal overhead. Here's how it works. When an AI model generates text, it produces a unique set of logits for each token. These logits reflect the model's internal state and can be hashed to create a compact representation. A key insight was that validators don't need to replicate the entire generation process.

They can verify individual tokens and the mathematical relationships among them at random points, dramatically reducing computational requirements while maintaining strong security guarantees. We've architected the system as a non-blocking proof-of-work consensus mechanism, meaning verification happens in parallel with transaction processing. This preserves the high throughput of our Solana-based foundation while adding the economic benefits of proof-of-work mining. For decentralization, we've implemented model sharding techniques inspired by recent academic breakthroughs. This allows us to distribute massive models

– 600B plus parameters – across multiple nodes, enabling even consumer-grade hardware to participate. For blockchain security, we give miners who have made the biggest verified contributions to network problem-solving on two timescales, short and medium term, the greatest ability to select and order transactions. In other words, we've replaced what, stake, means in Solana, money that you've locked up to earn rewards, with a proxy for hardware investment like what you see with Bitcoin, but in a way that allows us to reuse Solana's tower Byzantine Consensus. By solving these technical challenges, we've created a system where security, decentralization, and efficiency can coexist rather than compete, achieving 0.1% verification overhead instead of the 10-1000x overhead of alternative approaches.

Ashant Pandey. Your system aims to reduce training costs by 10x and INFERENCEOVERHEAD to 0.1%. Can you break down the technical innovations that enable THESEBREAKTHROUGHS? Travis Good. Please see my previous answer regarding how we approach inference. For training we use some similar tools for verification alongside some innovations in the use of something called sparsity, which refers to how the connections among weights are maintained during training. For a deeper dive, please have a look at our lightpaper at Ambient. XYZ. A'Shawn Pondy. What key technological O-R-E-C-O-N-O-M-I-C shifts do you anticipate in I and W-E-B-3? And how is ambient P-O-S-I-T-I-O-N-I-N-G-I-T-S-E-L-F for those developments? Travis Good. Over the next 3-5 years, I see 4 major shifts de tuile reshape the AI and Web 3 landscape.

First, AI will transition from being primarily human-directed to becoming increasingly autonomous. We're already seeing the emergence of agentic systems that can plan, execute, and learn with minimal human oversight. These systems will drive demand for trustless verification. You won't want to delegate authority to an AI agent unless you can verify it's using the model and constraints you specified. Second, computational requirements for frontier AI models will increase by at least an order of magnitude, creating significant centralization pressure. The most capable models

will require infrastructure that only the largest entities can afford, raising concerns about access and control. We're kind of seeing a peer review of this with the latest Lama 4 release. Third, traditional fiat currencies will face increasing pressure as economic activity shifts to digital realms where AI agents transact with other AI agents, often across jurisdictional boundaries. This will drive demand for currencies that are natively digital and tied to computational resources. Fourth, regulatory frameworks will mature around AI, with a likely emphasis on transparency, auditability, and safety.

Systems that can't provide verifiable guarantees about their operation will face increasing scrutiny. Ambient is positioning for these shifts in several ways. Our verified inference system already provides the trust layer needed for autonomous AI agents. Our model sharding approach directly tackles the centralization problem. By establishing machine intelligence as a currency standard, we're creating a native medium of exchange for the coming era of agent-to-agent commerce. Our entire architecture is built around verification and transparency. The ultimate vision is to create an infrastructure layer that allows machine intelligence to flourish without becoming centralized under the control

of a few corporations or governments. Don't forget to like and share the story. Tip Vested Interest Disclosure. This author is an independent contributor publishing via our business blogging program. Hacker Noon has reviewed the report for quality, but the claims herein belong to the author. Hashtag dyo. Thank you for listening to this Hacker Noon story, read by Artificial Intelligence. Visit HackerNoon.com to read, write, learn and publish.