We built this because we were tired of the eventual consistency and “webhook hell' of traditional billing wrappers when trying to implement real-time usage tracking for AI tokens and seats.
The Core Idea: We provide a deterministic billing engine designed for complex models like metered usage, credits, and seat-based pricing, acting as a single source of truth for your app’s permissions. Beyond the logic, we also act as the Merchant of Record (MoR), taking on the global tax liability (VAT/GST) and compliance so you can offload the entire billing and tax infrastructure.
Tech Stack: Next.js, Go, and PostgreSQL. We focused heavily on idempotency and sub-100ms event ingestion.
We have a sandbox at sandbox.commet.co where you can test the API/SDK without an account. I’d love to hear your thoughts on the architecture, especially regarding how we handle multi-jurisdiction tax logic.
This was inevitable. We're seeing the "Vercel-ification" of the edge. For a long time, Cloudflare had the superior infrastructure (Workers, R2, KV), but lacked the cohesive "onramp" that Next.js provides for Vercel. By acquiring Astro, Cloudflare finally owns the full stack, from the local dev experience to the global edge delivery.
The real question is whether Cloudflare can maintain Astro's "framework agnostic" soul while being incentivized to push everyone toward Workers/Pages.
IMHO if I were you, I wouldn't take the deal. This structure is a red flag for most tier 1 VC. Here's why:
1. Equity Standards: 5% is astronomical for an advisor. Standard advisor equity is 0.1% to 0.25%, maybe 1% if they are practically a part-time Co-founder. 5% is what you give a late-stage Co-founder or a very early C-level hire, not someone just for 'making intros.'
2. The Signaling Problem: VCs invest in you, the founders. If an advisor is leading the pitch and handling the fundraising, it tells VCs that the founders can't sell their own vision. VCs want to see the CEO pitching, not a "hired gun".
3. Vesting: Any equity given must be on a 4-year vest with a 1-year cliff. If he gets 5% up front and then his "health doesn't permit" him to join, you've just dead-capped 5% of your company for a deck and some emails.
If he really believes in you, he should invest his own money or take a much smaller advisor slice (1% max) with zero cash fee.
Disposable architecture is a luxury of stateless systems. Once you deal with financial ledgers or tax compliance, the cost of "disposing" and rewriting a component is astronomical because of the audit trail requirements.
Reliability in fintech isn't just about a clean interface; it's about the historical context of every transaction. If you treat your core logic as disposable, you aren't just swapping code, you're risking the integrity of the data history, which is the actual product.
Regarding the separate control plane: I don't think it's overkill if you're aiming for multi-agent orchestration. A safety mesh needs to be centralized to maintain a global state of permissions. If you bake the safety logic into each worker, you end up with the same "flimsy logic" problem you're trying to solve.
Curious, how are you handling latency in the CAP v2 protocol when the control plane has to intercept every intent before execution?
The point about documentation is spot on, but there's another technical hurdle often overlooked: payment infrastructure. In Japan, B2B is still heavily dominated by bank transfers (Furikomi). If your SaaS billing stack only supports credit cards, you'll hit a wall with larger enterprise clients who expect to pay via monthly bank invoices. Localizing the billing logic is almost as important as localizing the UI.
I'm curious how you're handling the feedback loop from the device's own speakers. If the phone plays back the audio, doesn't the mic pick it up again and create an infinite echo? Or does the 2-second delay just turn it into a slow-motion oscillation?
This is a great deep dive. Most ASCII renderers feel "muddy" because they treat intensity as the only variable. Treating characters as structural embeddings (the 6D vector approach) is much closer to how our eyes actually perceive edges. It reminds me of how font hinting works at low resolutions. Truly impressive work on the contrast enhancement pass too.
We built this because we were tired of the eventual consistency and “webhook hell' of traditional billing wrappers when trying to implement real-time usage tracking for AI tokens and seats.
The Core Idea: We provide a deterministic billing engine designed for complex models like metered usage, credits, and seat-based pricing, acting as a single source of truth for your app’s permissions. Beyond the logic, we also act as the Merchant of Record (MoR), taking on the global tax liability (VAT/GST) and compliance so you can offload the entire billing and tax infrastructure.
Tech Stack: Next.js, Go, and PostgreSQL. We focused heavily on idempotency and sub-100ms event ingestion.
We have a sandbox at sandbox.commet.co where you can test the API/SDK without an account. I’d love to hear your thoughts on the architecture, especially regarding how we handle multi-jurisdiction tax logic.
Happy to answer any questions!