There comes a time in every acquisition where you ask yourself 'was this worth it? Did we actually inflect the business?' This week - we can definitively answer that question: "yes, hell yes." Our new revenue growth 5x'ed QoQ, demand has been unprecedented, and a few weeks ago we did a thing I never thought we'd do: we made Metronome self serve. We are now onboarding more new users in a quarter than we did in the first several years of our life. Come join HuggingFace, Anthropic, Nvidia and the rest of the world on purpose-built billing software for AI. signup.metronome.com

At @getmetronome we got a few things right, but one that I'm particularly proud of was aligning our core product development principles to the problem domain that we were tackling. In billing, correctness is a P0 problem, so we oriented every product decision, every process, and every workflow around 'how do you make sure that everything we do is provably correct by construction?' You don't get infinite principles, so picking them correctly and following through is actually incredible alpha. Fun talk about this and more with @quentinclark

Yesterday at Sessions, we introduced the world to our vision for Streaming Payments. This work is the logical outgrowth of everything we’ve been building on Metronome for the past 6 years. Agents-as-users are different from traditional SaaS users. Humans tolerate latency. Agents don’t. Fundamentally this means that latency of traditional software operations need to shrink across the board. A human cannot detect the difference between a checkout page taking 100ms and 200ms to load. For an agent, every millisecond spent waiting is a wasted millisecond. Humans are also subject to decision fatigue. Ever wonder why every SaaS on earth offers ‘good, better, best’ pricing? It’s because humans are hardwired to choose the middle of three options. In an agent-centric pricing model, you should anticipate that pricing will get very complex and very dynamic. The best analogy I’ve seen is something like HFT where computers make millions of decisions a second to find the exact clearing price of complex products. It turns out that a huge amount of pricing and packaging design is about properly balancing the desire to offer choices that are economically efficient, with the fact that humans aren’t very good at making decisions. Agents don’t have these same foibles. In fact, agents love optimization problems with lots of variables. Agents will opt for pricing constructs that more accurately exchange price to value on a per use-case basis, even if they're significantly more complex than traditional SaaS models. So, what does this all mean for Metronome, Stripe, and Billing? Billing systems need to move to a realtime ledger. Every millisecond of lag between when an agent initiates a transaction and when that transaction clears the ledger is a wasted cycle. Even worse, if you aren’t operating in real time then you'll struggle with runaway agent spend or rampant fraud, or both. Transactions in billing need to be ledgered in a fine-grained way. Agents will have wallets, and they will use those wallets to spend money across a variety of vendors. Every agent action needs to be accounted for in real-time. If you don’t provide rich analytics, usage records, and audit-ready data then you’re going to drown in reconciliation problems and customer escalations. Finally, billing systems need to get more involved in entitlements. In an agent world, value exchange between vendor and agent will be discretized and continuously evaluated. This means that the system that governs access to data (entitlements) will need to be tightly coupled to the budgeting system (wallet). Technologies like MPP, stablecoins, agent wallets will need to be incorporated into core product gating flows in order to prevent runaway spend and fraudulent usage. That’s what we’re building with Streaming Payments. The combination of real-time payments, low latency usage tracking, and Stripe’s world-class infrastructure enables a new class of business models purpose-built for the agentic future.

So glad this episode went out into the ether. Jesse is smart af and a great friend. Hope you enjoy it.