fellow of infinite jest, of most excellent fancy // CTO @useorb (hiring!). powering billing for @vercel, @glean, @fal, and a ton more. cs/philosophy @caltech.
Building AI infrastructure means building systems that can scale alongside the product.
Tom Wallenstein, Staff Software Engineer at @youdotcom:
“We needed a scalable and flexible billing and metering system that lets us move fast today and keep moving as we grow and things change. Orb gives us exactly that, one system that serves engineering, analytics, and RevOps alike.”
On initial glance, AI blows our minds. But after repeated use, it disappoints. This pattern with new model releases has proven itself over years: they seem smart until we spend enough time with them. Then we notice all their warts.
Capabilities seem mind-blowing when first demoed, but fall short after weeks/months of exposure.
Why is this?
At AI infrastructure companies, pricing is part of the product.
Sid Shanker, Engineering Manager at @baseten, helps the team launch new models and hardware without turning monetization into a bottleneck.
“We launch new models and hardware constantly, and pricing them is half the product decision. Orb lets us move at the speed of our roadmap and prioritize delivering value to our customers.”
these guys are the real deal! congrats on the launch and we are excited to power this team @useOrb
(also s/o to one of the best product feedback givers in the game @jordanalexmeyer)
Over 1 billion PDFs are created every day, but your agents still can’t read them reliably.
Today we’re releasing Parse 2.0, the most accurate document parsing API in the world.
Extend already processes millions of pages daily for leading AI teams like Brex, Mercury, Opendoor, Flatiron Health, and hundreds of others. Now, its even better.
Parse 2.0 is SOTA quality on RealDoc-Bench, our open source benchmark that measures agent success rate on real world docs that agents actually encounter in production.
We trained Parse 2.0 on 1M+ pages of the hardest documents seen in production. Here’s how it stacks up:
- #1 in healthcare, real estate, logistics, and financial services
- 95.7% agent Q&A accuracy on 581 docs (next best: 92%)
- 0.847 F1 on layout (next best: 0.759)
Give it a try today and build production-ready document agents with Extend.
every vendor → vendor escalation involves a datadog screenshot. you can't escalate without showing datadog. canonical proof of problem
free alpha: datadog should build inline cross-vendor tracing with at-mentions. would overtake slack connect overnight
it's 2026. anthropic is printing money. github is down. installing dependencies is a dice roll. the rocket company is going public. the rocket company is buying the coding company. github is down. there's a thermal event in us-east-1. login with google is unsafe. $100M is not enough in sf anymore. github is down...
@maynkag what's your sense of the upper limit capability right now of AI-in-prod? curious if there are boundaries you think we'll cross in 6mo, 1year, etc
I love to see companies and customers excited about stuff like this. it's technically tricky + product clarity challenging + interacts with a *ton* of surface area