# Day 119 of the singularity: Patrick Collison's parabolic-chart sermon

URL: https://www.thedeepfeed.ai/posts/2026-05-02-day-119-singularity-stripe-keynote/
Category: Business
Published: 2026-05-02
Author: the-deep-feed
Tags: stripe, agents, payments, macro, ai-economy
Kind: deep

> Stripe opened Sessions 2026 with a chart of new firm formations going vertical and a familiar thesis — but the through-line was a single line from a Meta VP: payments are pivoting from a moment to a policy.

## TL;DR

- Patrick Collison opened Stripe Sessions 2026 with a chart he had **never shown publicly before** — new businesses launching on Stripe each month, going parabolic since January 2026 — and framed the conference as taking place on **day 119 of the singularity**.
- The macro evidence is unusually clean: **$1.9T TPV up 34% against world GDP at 2%**, **Stripe Atlas at 100,000 lifetime incorporations with Q1 2026 +130% Y/Y**, and the **Atlas class of '26 tracking 5x the revenue** of the class of '25 at the same point.
- Sam Altman, in the afternoon fireside, accepted the framing on the record — *"Day 119 feels like a reasonable enough guess. I won't fight it"* — the first time he has endorsed a specific date for takeoff.
- The actual editorial through-line of the keynote was not Patrick's chart. It was Meta VP Ginger Baker's line that **"payments will pivot from being a moment to being a policy"** — a sentence that ties the 45+ launches into one phase change.
- The position: the chart is real, the singularity framing is doing too much work, and the "moment to policy" frame is the most useful conceptual shift in payments since the magnetic stripe — because it's the one bet that survives whether the parabola continues or not.

![Patrick Collison's "parabolic" chart of new businesses launching on Stripe — the Day 119 framing.](/post-images/2026-05-02-day-119-singularity-stripe-keynote/hero-parabolic-chart.jpg)

# The chart

The Stripe Sessions 2026 keynote opened with a long montage of news clips, founder one-liners and pandemic-era video calls. It ended with **Patrick Collison** walking on stage and saying, deadpan: *"It's April 29th, otherwise known of course as day 119 of the singularity."* That joke, which the keynote returned to repeatedly, is the spine of everything that followed. ([Stripe's "everything we announced" recap](https://stripe.com/blog/everything-we-announced-at-sessions-2026) ran the line straight, no quotation marks.)

A few minutes in, after the *"5 million businesses, 86% of the Forbes AI 50, 90% of the Dow Jones, 80 of the NASDAQ 100"* throat-clearing, Patrick said: *"I want to show you a chart we've actually never publicly shown before."* On screen: the number of new businesses launching on Stripe per month from 2019 forward. The 2020 pandemic spike never returned to baseline. From 2021 to 2025 the line drifts upward at a polite rate. Then, starting in January 2026, it bends. ([The US Census Nonemployer Statistics](https://www.census.gov/programs-surveys/nonemployer-statistics.html) puts ~33M solopreneurs in the underlying baseline; Sands' segment puts ~5M of those above $100K.)

> "Since the beginning of this year, the beginning of 2026, we have started to see this incredible inflection, this parabolic rise in new firm creation."
> — Patrick Collison, Stripe Sessions 2026 keynote

A few hours later he sat for a TBPN interview and gave the part of the chart that wasn't on the slide: *"March was an all-time record for new business incorporations with Stripe. New businesses joining Stripe in Q1 were up 71% Y/Y. It's parabolic. It's exponential. I'm running out of mathematical terms, but it's really very striking."* TBPN clipped that as their headline of the week ([449 likes, 76K impressions](https://x.com/tbpn/status/2049948458648547394)). And it wasn't just more tiny businesses — Patrick told them [the median business is also doing better](https://x.com/tbpn/status/2049958588135948407): *"Somehow, the median business is also doing better."*

Two days later he posted the single number behind the chart:

> Stripe Atlas just hit 100,000 all-time incorporations. Q1 2026 is +130% Y/Y.
>
> — [@patrickc](https://x.com/patrickc/status/2050206647793144098), May 1, 2026 (918 likes / 70K impressions)

The "Day 119" framing is a joke. The chart is not.

# The macro stats, on one page

![The cleanest macro signal in the keynote: $1.9T processed, +34% Y/Y, against a global economy growing ~2%.](/post-images/2026-05-02-day-119-singularity-stripe-keynote/tpv-vs-gdp.jpg)

The keynote stuffed a remarkable amount of macro evidence into about ten minutes. Stripe doesn't usually share this much of its own data on stage. Five numbers carried the load.

| Stat | Stripe value | Comparable | Source |
|---|---|---|---|
| Total payment volume, 2025 | **$1.9 trillion** | ~1.6% of world GDP | Patrick Collison, opening |
| TPV growth, 2024→2025 | **+34% Y/Y** | World GDP growth ~2.3% | Keynote vs. World Bank |
| Stripe Atlas all-time incorporations | **100,000** | Hit ~6 months earlier than internal forecast | Sands segment + [Patrick on X](https://x.com/patrickc/status/2050206647793144098) |
| Atlas class of '26 revenue trajectory | **5x class of '25** | Class of '25 was already 2x class of '24 | Sands & Konstantopoulos, talk 03 |
| Stripe sellers earning most of revenue abroad | **22.9% (2x in 5 years)** | Was 11.6% in 2020 | [Jeff Weinstein on X](https://x.com/jeff_weinstein/status/2049979058465935514) |
| API availability, 2025 | **99.9996%** | Industry SLA bands ~99.95–99.99% | Patrick, opening |
| Production PRs merged per day | **>1,000** | "GitHub-recently-reported"-tier velocity | Patrick |

Two points to lean on. First, the **TPV vs. world GDP delta is the strongest macro signal in the keynote.** Stripe processed roughly 1.6% of global GDP and grew its share of it by ~17x the underlying rate. That delta has to come from somewhere — either share-shift away from offline payments and other PSPs, or genuinely new economic activity that didn't exist in 2024. The [World Bank's 2025 Global Economic Prospects](https://www.worldbank.org/en/publication/global-economic-prospects) put global growth in the low-2% range, which makes the comparison fair. Patrick's [meta-reflections thread](https://x.com/patrickc/status/2049705418436600244) (1,174 likes, 129K impressions) called it both: *"It feels that the entire economy is replatforming right now. Many charts at Stripe are inflecting in quite dramatic ways. What GitHub recently reported for commits we are seeing in economic activity."* John Collison, on Bloomberg the same day, [used the same word](https://x.com/collision/status/2049841675061903544): *"the economy is replatforming around AI."*

Second, the **Atlas class-of cohort comparison is the headline number for the *new firms* part of the story.** [Replit's Amjad Masad](https://x.com/amasad/status/2049934937688854993) summarised it in two characters: *"Prompt ➡️ LLC."* The stat behind that is from Emily Glassberg Sands' afternoon talk: companies incorporated through Atlas in 2025 are doing 2x the revenue of the 2024 cohort at the same maturity, and the 2026 cohort is tracking to 5x the 2025 cohort. That last number is internal and unverifiable from the outside; the 2025-vs-2024 doubling is consistent with [Stripe's published Atlas year-in-review](https://stripe.com/blog/stripe-atlas-startups-in-2025-year-in-review).

The "globalised by default" stat from [Jeff Weinstein](https://x.com/jeff_weinstein/status/2049979058465935514) (*"22.9% of Stripe users now make most of their money abroad, up 2x from 11.6% in 2020"*) is the second cohort signal under the chart. The top 100 AI startups on Stripe sell into 55 countries within their first year, per Sands. That makes the parabolic chart not just a US story.

The new-firm creation chart, the +130% Atlas Y/Y, the TPV growing 17x faster than world GDP — they are all telling the same story, from three different points of view. The story is real. The harder question is what it is evidence *of.*

# Sam Altman accepts the date

The day's most editorially loaded moment did not happen during the keynote. It happened a few hours later, in the **Sam Altman / Patrick Collison fireside** (talk 04 in the Sessions playlist, transcript publicly available).

Patrick opened with a setup: *"We've kind of arbitrarily decided that the singularity started on January 1st and thus today is day 119. What do you think of that?"*

> "It does feel like we are somehow in the takeoff. Day 119 feels like a reasonable enough guess. Yeah, I won't fight it."
> — Sam Altman, Stripe Sessions 2026 fireside

Sam has been studiously vague about *when* takeoff begins for years — usually *"agents in 2025,"* *"AGI before end of decade,"* *"every week feels different."* This is the first time on record he has accepted a specific *date.* He matched Patrick's diagnosis on what changed: *"the models got really good especially for coding starting late last year, very early this year. Every week is now a little bit different than the week before."*

The parabolic chart and the Sam endorsement land as a single artifact: a CEO whose company is downstream of the AI economy and a CEO whose company *is* the AI economy, on the same stage, agreeing that something measurable changed in January. That forces the rest of the keynote to be read as engineering done in service of a prediction.

# The chart is real. The 'singularity' framing is not.

Time to take a position. The Stripe Sessions 2026 keynote was a **strong macro claim wrapped in a weak rhetorical wrapper.** The numbers underneath are sound — the chart, the +130% Atlas Y/Y, the 22.9% globalised-sellers stat, the TPV-vs-GDP delta. None of them needs the word "singularity" to be interesting; the word *weakens* the argument by importing science-fiction connotations that the chart cannot support.

Three reasons to be cautious about the framing without being cautious about the numbers.

**First, parabolic curves are not parabolic for very long.** Stripe inflected in early 2026 and ran hot through Q1. That is one quarter of data. The 2020 pandemic spike, as Patrick notes, *"never went back to baseline"* — but the slope flattened within months. A two-quarter line that bends sharply upward is the *same shape* as the early phase of every previous adoption S-curve: card e-commerce in 1999, mobile commerce in 2010, freemium SaaS in 2014. Each looked vertical and then revealed the rest of the logistic. The "singularity" framing prices in that the slope continues. The chart only shows the slope changed.

**Second, the AI-elite reaction was muted.** The dossier flagged this, and the X data confirms it. Searching across the post-Sessions window, **none of Garry Tan, Paul Graham, Greg Brockman, Logan Kilpatrick, Sam Lessin or Will Brown tweeted about Sessions 2026.** Garry's only Stripe-adjacent post that week was about [Saikat Chakrabarti, an unrelated Stripe alumnus](https://x.com/garrytan/status/2049713080385536416). Sam Altman appeared on stage and *said* it was Day 119, but he didn't post about it. This is the *opposite* of the [AI Ascent 2026 cohort dynamic](/posts/2026-04-28-20-days-that-changed-the-ai-agent-market/), where the same names dominated the discourse for a week. The "singularity is here" framing landed inside Stripe's own audience, not the AI priesthood it was implicitly addressed to.

**Third, the credentialed counter-take is sharp.** [Soren Iverson](https://x.com/hypersoren/status/2050001428173746359) (43L / 71K imp) pushed back on John Collison's [Coasean framing](https://x.com/collision/status/2049974499278528952) (728L / 115K imp): *"AI modularizes execution, which sends scarcity to context and trusted operations. Transacting externally is now (counter-intuitively) more expensive. Stripe benefits from a Coasean Singularity, but market incentives do not support it."* BU's [Andrey Fradkin](https://x.com/AndreyFradkin/status/2049933001300258856) added the deadpan academic note: *"Stripe citing Coase + AI in their sessions talk, but not the singularity."* And [@kognise7's snark](https://x.com/kognise7/status/2049528040171204851) (83L), *"Patrick says we are >100 days into the singularity, but availability has gone down and pull requests have stayed the same,"* captured the obvious follow-up: if the singularity is here, why is the API still occasionally returning 502s?

None of these critiques touch the parabolic chart. They touch the *story Stripe is telling about it.* The chart can be true and the story can be over-fit at the same time.

The way out of the trap is to read the keynote against itself. There is a quieter, more durable thesis embedded in the same script — and it didn't come from a Stripe employee.

# The phase change: from a moment to a policy

![Ginger Baker's Sessions 2026 phase change: payments pivoting from a discrete moment into a standing policy.](/post-images/2026-05-02-day-119-singularity-stripe-keynote/policy-not-moment.jpg)

The most editorially important sentence at Sessions 2026 was delivered by **Ginger Baker**, VP of Product at Meta, in a six-minute fireside about a third of the way through the keynote. Asked what is going to feel obvious in two years that sounds radical today, she said:

> "I think that payments will pivot from being a moment to being a policy. And I think it feels radical today because that policy will be something that you set with your agent most likely, and that policy will be scripted with guardrails that say *please spend only up to $50 on groceries this week, always preference this card, never do an authorization for a payment over $500.* And those policies and guardrails are things that someone other than yourself, or something other than yourself, is going to be operating by."
> — Ginger Baker, VP Product, Meta

Will Gaybrick's response (*"Moment to policy. I like that"*) was the only time in the whole keynote where a Stripe executive plainly endorsed a non-Stripe person's framing as the better one. He repeated it again 25 minutes later in the Treasury segment: *"As you heard from Ginger…"*

This is the actual through-line. **Every payment system from the magnetic stripe through Apple Pay to one-click checkout has been a *moment*** — a discrete authorization, with a user attached to it, that the network treats as atomic. A *policy* is the opposite shape: a pre-approved standing order, scoped by rules, executable autonomously by something that is not a human, against funds the human controls, for as long as the policy is alive. Ginger described the consumer-side version (your agent grocery-shopping). The Stripe ship list is the merchant-and-platform-side version.

If you read the 45+ keynote launches through that lens, the keynote stops looking like a recap and starts looking like a single architectural change.

| Sessions 2026 launch | Old payments shape (moment) | New payments shape (policy) |
|---|---|---|
| **Link wallet for agents** | One-time CVV at checkout | Single-use credentials your agent draws against, with synchronous human approval per use |
| **Machine Payments Protocol (MPP)** | HTTP 402 was a meme | A live HTTP-native handshake by which any service tells *any* agent how to pay |
| **Agentic Commerce Suite (ACS)** | Static checkout page rendered for a human | Server-side endpoint that lets an agent buy on behalf of a logged-in user, scoped by the merchant's policy |
| **Universal Commerce Protocol (UCP)** | Each marketplace owns its own checkout | A shared catalog and intent surface that any agent can query under merchant rules |
| **Stripe Console (agentic dashboard)** | Operator opens dashboard, configures, hits save | Operator describes intent ("set up a Pix A/B test in Brazil") and confirms an action plan |
| **Treasury via MCP** | Treasurer logs in to bank dashboard, manually pays invoice | Agent drafts the payment from an attached PDF; operator approves under a treasury policy |
| **Streaming payments on Tempo (mppx)** | Per-charge minimum economics ($0.30 card fee) | Pay-per-token, settled on-chain, $0.001 per unit, sub-cent over millions of transactions |
| **Radar 2026 (multi-account abuse, token theft, custom signals)** | Score a transaction at the moment of authorization | Score *the policy issuing the transaction* — the identity, the trial cohort, the agent fingerprint |
| **Smart Disputes (evidence guidance, +3x wins)** | Win/lose a single chargeback | Tune the policy that produced the dispute; reduce future ones |
| **Custom Objects + Workflows** | Stripe knew about charges and customers | Stripe knows about *your* shipments, claims, contracts — the data the policy needs |
| **Stripe Database (private preview)** | Read-only Sigma queries on transaction data | A real-time managed Postgres of your Stripe data, agent-queryable |
| **Stripe Capital — lines of credit** | One-shot term loan with a fixed amortization | A standing draw against a programmatic credit policy |
| **Stripe Treasury (multi-currency, FDIC, MCP)** | Bank account operated by a human via dashboard | Bank account operated by code under a treasury policy |
| **Meta + Stripe + Tempo creator stablecoin payouts** | Wire transfer, FX, country-by-country | A creator gets paid by their platform under a single policy across 160 countries |

The unifying property: **every one of these collapses an authorization-shaped surface into a policy-shaped surface.** Will Gaybrick's framing near the start of the keynote (*"agents are autonomous economic actors responsible for most internet transactions, and just as agents work faster than we do, they're going to spend money a lot faster, too"*) only makes sense if the substrate is policy. Agent throughput at moment-shaped checkout is fraud, by definition.

Ginger said it in one sentence. Stripe has been building it for two years and shipped a lot of it on April 29th.

# Why this is the thesis to commit to

![The Atlas class-of-'26 trajectory: +130% Y/Y in Q1, tracking 5x the revenue of the class of '25.](/post-images/2026-05-02-day-119-singularity-stripe-keynote/atlas-curve.jpg)

The "moment to policy" framing is more useful than "Day 119" for three reasons. It is more *engineerable*: the [TDF developer guide](/posts/2026-04-30-stripe-sessions-2026-developer-guide/) walks the concrete stack underneath it (MPP, ACP, UCP, x402, skill.md, Tempo, Link agent wallets). It is more *durable*: new firm creation could flatten next quarter and the policy primitive would still be the right substrate for agent payments. And it is more *true*: the shift from a discrete authorization to a standing rule is the actual delta between Sessions 2025 and Sessions 2026, the same shift Coinbase's x402, OpenAI's ACP, Google's UCP, Meta's creator-payouts integration and Visa/Standard Chartered's MPP work are all converging on.

[John Collison's Coasean-lens thread](https://x.com/collision/status/2049974499278528952) reaches the same conclusion from the opposite end. The Coasean argument is that AI lowers within- and inter-firm transaction costs, redrawing the firm boundary in real time. Lowering an inter-firm transaction cost is, operationally, *replacing a moment with a policy.* You don't redraw a firm boundary by issuing more credit-card authorizations; you redraw it by giving an agent a standing rule to transact within. John's networked-Stripe follow-up makes the practical version of the same point:

> Increasingly, the best part of using Stripe is the millions of other companies using Stripe.
>
> — [@collision](https://x.com/collision/status/2049547995625205797), Apr 30, 2026

And on TBPN, Patrick named the [skill stack that operates inside it](https://x.com/tbpn/status/2049988742530879543): *"high-agency people."*

The concrete tests over the next four quarters:

- **The chart.** A single-quarter inflection is interesting; four quarters is a regime.
- **Long-tail adoption of agent-shaped transaction surfaces.** Soren Iverson's adverse-selection critique is sharp: *"Scaled retailers have shut out third-party agents."* Watch whether Wix, Commerce Tools, Best Buy and Kate Spade actually ship ACS to GA.
- **The Atlas class of '26 trajectory.** Internal and unverifiable today; eventually shows up in the [Stripe annual letter](https://stripe.com/annual-updates/2024).
- **Treasury conversion.** [Stripe's Treasury tweet](https://x.com/stripe/status/2049621560743608481) (5,153 likes / 1.45M impressions, ~6.7x the headline 288-launches tweet) was the most-engaged of the conference. The audience cared more about the bank account than the agent stack; either the bank account funds the agent bet, or the agent bet is downstream of it.
- **Streaming payments off the demo stage.** TBPN's [breakdown of the live demo](https://x.com/tbpn/status/2049963646470758553) (*"pay for each token as it's delivered with stablecoins"*) was the tightest framing of the world-first claim. Per-token billing on Tempo + mppx in shipped products is the test.
- **Radar's identity-shaped fraud thesis.** Patrick's [token-pilfering clip](https://x.com/tbpn/status/2049968787894722674) and the *"1 in 6 AI signups are in multi-account abuse"* stat is the policy-side answer to fraud. If policy is the new authorization, identity is the new transaction.
- **Link adoption beyond OpenAI and Anthropic.** The [Link agent wallet](https://x.com/stripe/status/2049529444092838116) and [Link CLI](https://x.com/patrickc/status/2049535449484644702) tweet pulled 5,978 likes / 3M impressions, the biggest single-tweet number of the day. Engagement is not adoption.

# What to bet on

![The position the post is committing to: bet on the policy, hold the chart at arm's length.](/post-images/2026-05-02-day-119-singularity-stripe-keynote/hold-the-chart.jpg)

Patrick Collison stood on the Moscone West stage, in front of a crowd that had [crossed 10,000 attendees for the first time](https://x.com/patrickc/status/2050212647539507627), and showed a chart that, if it sustains, is the strongest visible signal of an AI-driven economic phase transition that any private company has produced. He named the chart "day 119 of the singularity." That naming is the part that will look silly later. The chart will not.

The framing the post is committing to: **the parabolic chart is real, the singularity wrapper is doing too much work, and the most important sentence of the keynote came from a Meta VP, not Stripe.** Payments pivoting from a moment to a policy is the actual phase change Sessions 2026 documented — and the right test of whether the rest of the framing was prescient or premature is whether, two years from now, your agent has a budget you set in a config file and Stripe is the substrate it spends against.

[Patrick already conceded the joke part](https://x.com/patrickc/status/2049705418436600244): *"We are being a bit tongue-in-cheek with this, of course, but only a bit."* The bit that wasn't tongue-in-cheek is the bit worth quoting back at him at Sessions 2027. The story Stripe wants — **the chart bends, every business runs on a policy, agents account for most internet transactions, and Stripe owns the rails** — requires both halves. The policy half is the bet worth taking. The chart half is the part that should be held at arm's length until the next print.

---

*See also Part 1 of this series, the [complete developer's guide to Stripe Sessions 2026](/posts/2026-04-30-stripe-sessions-2026-developer-guide/), for the full list of 288 launches and what is actually shippable today, and Part 5, the [Stripe agent-native API playbook](/posts/2026-05-02-stripe-agent-native-api-playbook/), for how to build against the moment-to-policy primitives.*

## Sources

- [Stripe Sessions 2026 keynote (full transcript)](https://youtu.be/e13-s0p1tfE)
- [Stripe — Everything we announced at Sessions 2026](https://stripe.com/blog/everything-we-announced-at-sessions-2026)
- [Stripe Atlas 2025 year in review](https://stripe.com/blog/stripe-atlas-startups-in-2025-year-in-review)
- [Stripe 2024 annual letter](https://stripe.com/annual-updates/2024)
- [Patrick Collison on X — Atlas hit 100,000 incorporations, +130% Y/Y](https://x.com/patrickc/status/2050206647793144098)
- [Patrick Collison on X — Sessions meta reflections](https://x.com/patrickc/status/2049705418436600244)
- [Patrick Collison on X — Link CLI for agents](https://x.com/patrickc/status/2049535449484644702)
- [Patrick Collison on X — 10,000+ Sessions attendees](https://x.com/patrickc/status/2050212647539507627)
- [Patrick Collison on X — Meta creator stablecoin payouts](https://x.com/patrickc/status/2049853787645636951)
- [John Collison on X — the Coasean lens on AI](https://x.com/collision/status/2049974499278528952)
- [John Collison on X — Bloomberg replatforming interview](https://x.com/collision/status/2049841675061903544)
- [John Collison on X — networked Stripe](https://x.com/collision/status/2049547995625205797)
- [TBPN — Patrick Collison: Q1 incorporations +71% YoY, "parabolic"](https://x.com/tbpn/status/2049948458648547394)
- [TBPN — median Stripe business is doing better, not just more startups](https://x.com/tbpn/status/2049958588135948407)
- [TBPN — streaming token payments demo recap](https://x.com/tbpn/status/2049963646470758553)
- [TBPN — Patrick on token-pilfering fraud](https://x.com/tbpn/status/2049968787894722674)
- [TBPN — Patrick on high-agency people and double majors](https://x.com/tbpn/status/2049988742530879543)
- [Stripe on X — new Stripe Treasury](https://x.com/stripe/status/2049621560743608481)
- [Stripe on X — Link wallet for agents](https://x.com/stripe/status/2049529444092838116)
- [Jeff Weinstein on X — 22.9% of Stripe users earn most revenue abroad](https://x.com/jeff_weinstein/status/2049979058465935514)
- [Amjad Masad on X — Prompt → LLC](https://x.com/amasad/status/2049934937688854993)
- [Soren Iverson on X — Coasean Singularity counter-thread](https://x.com/hypersoren/status/2050001428173746359)
- [Andrey Fradkin on X — Coase, but not the singularity](https://x.com/AndreyFradkin/status/2049933001300258856)
- [kognise7 on X — Sessions 2025 vs 2026 critique](https://x.com/kognise7/status/2049528040171204851)
- [World Bank — Global GDP growth 2025](https://www.worldbank.org/en/publication/global-economic-prospects)
- [Census Bureau — Nonemployer Statistics (solopreneur baseline)](https://www.census.gov/programs-surveys/nonemployer-statistics.html)
- [Sam Altman & Patrick Collison fireside transcript (Stripe Sessions talk 04)](https://youtu.be/e13-s0p1tfE?t=4500)

---

Canonical: https://www.thedeepfeed.ai/posts/2026-05-02-day-119-singularity-stripe-keynote/
Site: https://www.thedeepfeed.ai
Full corpus: https://www.thedeepfeed.ai/llms-full.txt