Sapiom

1/ We've acquired @fewsats, the payment infrastructure for AI agents. The industry over-invested in reasoning and ignored access. Models can plan, but the second an agent has to act, there is nothing beneath it. Intelligence without access is a demo.

Just finished listening to Adam Mosseri on Lenny. A lot of useful nuggets. I'm glad he agrees that token leaderboards are a terrible idea. When it comes to tokens and spending, my thoughts: A lot of people's instinct in this industry is for managers to count every token. Reactively. As if there wasn't a better way for people to be spending their time. What builders need is a way to give agents a budget, set the rules once in the infrastructure, and get the breathing room to build.

@NotionHQ how is this agent native?

Introducing Ship OS: The agent-native way to ship software. Run your entire product development cycle in Notion, from customer feedback to a merged PR. Agents handle the triaging, routing, and summarizing. Your team handles the judgment calls. Set up Ship OS → notion.com/ship-os

Remember the 90's movie Flubber with Robin Williams? My team and I couldn't agree on whether it was called Flubber or Blubber so we looked it up. Turns out, it was a remake of a movie from the 60's.

The @sapiom team is hosting a small builder night in SF tonight. We're keeping it intentionally tight. If you're building something real with agents, DM me what you're working on. I'll send details to a handful of folks it makes sense for.

Want to build the execution layer, and not just another brain? We're hiring. 👇 sapiom.ai/careers?utm_so…

Our founder, @i_zerbib, is originally from France, but has been busy building in sunny (or foggy, depending on the time of day) San Francisco. 🇫🇷🌁 Naturally, our team spans both, with folks in between the two cities. When a community like Intertech Talents has spent the last year building bridges between the two, it was an easy yes to sponsor. Nikita from our technical team helped bring our story to Paris last week, chatting with builders passionate about the future of autonomous agents. We couldn't be more excited about what's ahead.

@gregisenberg I know a guy named @i_zerbib 👀 👀

Who wants to come on my podcast this week? 2,000,000+ listens per month. 1. You teach one AI tool, skill, or framework that helps people build a business 2. You can have 1 follower or 1M, doesn't matter 3. You come prepared Tag someone or tag yourself. Lets go.

Karp’s right that enterprises are “paying for tokens that create no value.” IMO, he’s partially wrong about the fix. His answer is discipline. Data sovereignty, more careful spend, tighter governance. But that puts the burden on the enterprise to manage something they can’t actually see. You can’t ask a team to exercise judgement over a system they have no visibility into. It's like sending your 4 year old off for the day with no guidance, no boundaries, no idea where they even went, and expecting them back with no scrapes, no bruises, and definitely no new pet snake in their pocket (swap to a cheaper babysitter and the kid may still come home with a snake). And watching from the front porch doesn’t fix it either. A camera pointed at the front door tells you the kid left. It doesn’t tell you where they went or why they came back with a snake. Watching a black box fail more slowly is still watching it fail. What actually changes the outcome is infrastructure that gives you real access at the task level: what ran, what it cost, whether it succeeded, how the routing decision got made. Not a dashboard. Control. That infra? It’s been built. And I’m damn proud of the @sapiom team for building it.

👉bit.ly/4gj1Yt7

We don't need a proposal-worthy stunt to prove we deliver. 99% transaction success. 84,000+ agents in production. That's just a Wednesday for us. Hi, we're Sapiom — the runtime layer AI agents actually run on. 👋

Most agent infrastructure today is a horseless carriage. In 1886, Benz's first automobile was literally a carriage with a small engine that took over for the horse that used to propel it. It took fifteen years for the car to get its own shape. Agent infra is living those fifteen years right now. Almost everything agents run on was built for a world with a human at every step. Every API key provisioned by one. Every billing relationship approved by one. Every permission signed off by one. Agents break that assumption. The industry's answer so far: put "agent" in front of the old tools. It doesn't change what they are. The engine is new…but the carriage is the same. The horseless carriage test, for your stack or a vendor's: 1. New capability = new vendor, new key, new billing relationship, or another sprint building it yourself? Or one step? 2. Governance designed for software that acts on its own, or inherited from human IAM? 3. Can you see what an execution did, step by step, or just the aggregate bill? "New vendor every time." "Inherited." "Aggregate." That's a horseless carriage. Building it yourself just means you're the carriage maker. We started @sapiom by refusing to inherit it.

Every team building agents in production ends up rebuilding the same boring-but-critical infrastructure: sandboxes, state, cost controls, capability access, visibility, and control. Yesterday, Sapiom handled 1.5M+ agent transactions across ~40K tenant sandboxes for ~40K active agents including 662K governance checks and 654K paid transactions. That’s the stuff teams shouldn’t have to spend months wiring together before they can ship the actual product. Sapiom is the platform where agents can actually perform.

I left Shopify in August 2024 after 4+ years. $100B in GMV through Shop Pay. I left knowing exactly what world-class infrastructure looks like when it works. Before that, I built Earny, an automated agent doing price refunds for 4.5M users. It was acquired in '21. Two different companies. Same ceiling. At Earny, nothing like this existed, so we built it ourselves. At Shopify, the infrastructure was world-class, but it assumed a human was always the one pulling the trigger. Either way, the moment something needs to act on its own, the floor gives out.

This is the whole game. The layer between the work and the intelligence is where ROI actually comes from. Aaron’s key insight is that every company building that layer alone will not scale. Understanding the workflow, routing every capability and model to the right path, enforcing budget, recovering when a step breaks, and knowing exactly what happened — that is not something 10,000 companies should each rebuild. It is heavy infrastructure. Most companies can't build it well themselves. So it becomes the thing they run on, not the thing they rebuild. Horizontal first. Verticals on top. Most interesting infrastructure problem in tech right now.

Let’s get you from demo to production ready. sapiom.ai

Anyone can get an agent working in a demo. The hard part is the 10,000th run at 2am, when a provider's down and something fails three steps deep and nobody's watching. We helped over 86K agents run in production.

Building agents? Let's chat. bit.ly/4g7lmct

Don't let your agent startup accidentally become an infrastructure startup. You build the agent. Then production shows up, and suddenly you own access, spend limits, approval flows, retries, routing, and a receipt for everything it did. Works as a demo, breaks at production scale. Builders shouldn't have to rebuild the operating platform for agents just to ship. That's the problem we're building Sapiom to solve.

@Bencera @deifosv 🫶 send us a dm

@deifosv @sapiom

Uber had a $500M Anthropic bill in a single month. Mine was over $1M. So I rented GPUs and now use open-source models. 100x cheaper. The future of AI is Open-source.