Anton

The Financial Engineer | CTO @GetStigg

[Anton @thefineng]

Just 2 hours after I published the case of a $1M token leak and the lesson it teaches about subsidization, this happened.

https://t.co/9k3hoBw5zU

[Tibo @thsottiaux]

To make sure our current users have an incredible experience and continued access to Astra, we are going to pause subscriptions to our $200 Pro plan. These put the most strain on our systems and we wanted to take the smallest step that allows us to continue giving the broadest access possible. All other plans and the api remain available.

There is no impact to existing accounts and we are working on adding more capacity as fast as we can. Thanks!

[Anton @thefineng]

@hanrui_w @nebiustf @nebiusai The fact that 'token != compute cost' is a game-changer for so many startups. Now they just need a reliable metering layer between inference and application (yeah, shameless plug, I know 😉)

[Anton @thefineng]

token != compute cost

[Ryan Hanrui Wang @hanrui_w]

GLM-5.3-Flash is flying on Nebius: 294 output tokens/sec -- the #1 result in Artificial Analysis.⚡️ @nebiustf @nebiusai

[Anton @thefineng]

@jjacky tmux with a 43" screen

[Anton @thefineng]

@dylan522p Have you every visited JFK?

[Anton @thefineng]

@mitsuhiko I have an AliExpress habit of buying 3 of the same cheap products, so one of them will for sure be the one I need

[Anton @thefineng]

The biggest token economics engineering problem isn't recording that 2,741 tokens were consumed. It maintains the relationship among consumption, entitlements, pricing, and revenue.

On this and more thought about @stripe's @OpenRouter deal: https://t.co/wSAjmsDQq8

[Anton @thefineng]

@Inomsxbt No reason to justify. The question is usage vs. outcome..

[Anton @thefineng]

To all the engineers who just finished the journey to usage-based pricing and are now being asked to move to outcome-based pricing: my heart is with you.

[Anton @thefineng]

@yabarich It indeed seems like a big number, but network metering tools measure much larger packet capacities and have done so successfully for years. I agree that outcome-based pricing is important from a financial and product perspective, yet we need good on-the-token metering.

[Anton @thefineng]

With all the GH downtime on Monday, I really think they are in trouble now that people (agents) are using actions for everything, costing them way more than ever.
I bet someone at @github is working really hard right now on better metering and pricing changes. Here to help 😉

[Anton @thefineng]

@vral Now it's time for advanced metering and enterprise contracts...

[Anton @thefineng]

@MillionInt and the art of explaining your mom that Claude will fix the printer better than you

[Anton @thefineng]

The OpenRouter acquisition story makes me think a lot about the role it plays as the AI infrastructure piece that Stripe is missing. Sharing them raw; will try to compile it better later:

1. OpenRouter currently seems to struggle with large enterprise deals and mostly acts as a (super successful) consumer platform. How will it align with Stripe’s B2B strategy?
2. Stripe sees Tokenomics as a key integrated part of the future economy. Stripe’s main revenue source is taking a fee on money transactions, and OpenRouter takes a fee on token transactions. It is more of a day-after-acquisition thing than something they would do now.
3. Assuming Stripe is in true need of someone who sits on making money from inferences, OpenRouter sits exactly between HuggingFace (which is too deep for Stripe to dirty their hands) and “consumer” AI gateways (Venice, for example). With its access to builders on the one hand, and the simplicity of its architecture on the other, Stripe gets a “shortcut” to making it another income channel for its “bank”.
4. Alex Atallah (OpenRouter’s CEO) knows how to create marketplaces; he has been into it for the last ~10 years. Stripe is now in an interesting moment with their moat. On one side, they have Visa and Mastercard that are squeezing Stripe’s “multi-currency” advantage. On the other hand, OpenAI and Anthropic are capturing the bulk of token consumption, which processes the largest amount of money in tech ever (40b/y). Excelling in a marketplace is Stripe's biggest opportunity to create a moat that would survive this squeezing. With Alex’s experience scaling OpenSea and now OpenRouter, Stripe couldn’t wish for better talent for their marketplace moment.

What else am I missing here that you think should make this acquisition happen?

[Anton @thefineng]

We acquired https://t.co/dXsNj5Hxs6 to complete one of the critical missing pieces of AI billing infrastructure: enterprise contracts
Shared some of my thoughts and the journey here: https://t.co/nVLT2Wu01G

[Anton @thefineng]

@maxleiter Stripe needs an infrastructure layer to manage the largest bank of token economics. Yes, it sounds like a collection of buzzwords, but this is usually why acquisitions happen. OpenRouter is their best option today for supporting multi-currency.

[Anton @thefineng]

@patrickc 💯 I'd add that there's a limitation on how much data our brain can process in a passive, text-based chat interface. To learn, understand, and memorize, we need various interfaces that help with focusing and context switching. CLIs are great, but really overrated nowadays.

[Anton @thefineng]

@GergelyOrosz Don't you think a small ClaudeCode/Codex system prompt with Serpapi/Browserbase MCP would do better for the research?

[Anton @thefineng]

I really don't understand this whole watermark thing.
Like, ANY text Claude ever wrote has a very strong watermark that Claude made it.