Soria is an AI financial terminal built for healthcare. Its agents pull hundreds of public and private sources into a single always-updated sector model: CMS star ratings and rate announcements, NAIC statutory filings, SEC 10-Ks and XBRL, state Medicaid portals, DMHC flash reports, hospital cost reports, Part D enrollment. Analysts review the output, and it reaches banks, hedge funds, and asset managers through a terminal UI, an API, MCP, warehouse sync, and a research feed.
Founder and CEO Adam Ron spent eight years at Bank of America as a VP in equity research, covering the same sector Soria covers today. The company is backed by Y Combinator, won BattleFin's #1 New Data Vendor award, and is used by 30+ banks, hedge funds, and asset managers.
Being first is the product
Sector research runs on an asymmetry: the same filing is worth considerably more to whoever reads it first.
CMS posts a rate announcement. A state Medicaid portal quietly swaps out an enrollment file. An insurer drops an 8-K revising guidance. None of it arrives with a push notification, and none of it lands on a schedule anyone else controls. Somebody has to be watching the page.
So Soria watches. Hundreds of sources, continuously, waiting for the moment a page changes and a new number appears. That watch loop isn't back-office plumbing for Soria — the whole pitch to a hedge fund is that Soria sees it before the analyst does. However fast the pages get checked is how fast the product is.
Buy-versus-build math, from someone who had already built it
Most teams evaluating a scraping API are estimating what the thing would cost to run in-house. Adam wasn't estimating.
"I've self-hosted the same functionality on AWS, so I know what it costs. I basically needed Firecrawl-like features at a more reasonable price, and I honestly didn't want to manage it. Context was a much better deal."
That's an unusual position to buy from. He had already paid the build cost once: proxies, headless browsers, retry logic, queues, the autoscaling, and the pager that comes with all of it. The stack worked. It just stayed his to operate, and every hour spent keeping it healthy was an hour not spent on healthcare data, which is the part of Soria nobody else can build.
The pricing question and the operations question came out the same way. A managed scraping layer only makes sense if it costs less than the sum of the infrastructure and the attention, and Adam was one of the few buyers in a position to check that claim precisely.
He found Context.dev the way a lot of our customers do:
"Got referred by another YC founder."
Notably, none of this was about brand data. Soria doesn't need logos or colors — they needed the Markdown, Crawl, and Sitemap endpoints underneath their own monitoring pipeline, at a price that survives running them on hundreds of sources on a loop.
One shot with Codex
The migration was a single sitting.
"Basically one shot the migration with Codex and an API key."
Scraping is one of the cleanest interfaces in a data pipeline: URL in, clean content out. When the shape of the call barely changes, an agent can do the swap end to end, and the review is short enough to read in one pass. No spike, no parallel run, no staged cutover.
Faster pages, faster customers
"Faster page monitoring means we get data for our customers faster, which can be a key differentiator for us versus our competitors."
The sentence moves from an infrastructure property to a competitive one with nothing in between, which is exactly what it means for page monitoring to be load-bearing. Soria's customers don't buy a scraping layer, and they will never see one. They see a number that shows up in the terminal before it shows up anywhere else.
Key benefits
- A one-shot migration: Codex and an API key, done in a single session.
- No infrastructure to operate: the proxies, browsers, and queues Adam used to run on AWS are somebody else's on-call now.
- Pricing checked against a real build: evaluated by a founder who had already run the same stack and knew its true cost.
- Faster page monitoring: hundreds of regulatory and filing sources checked sooner, so new data reaches customers sooner.
- Engineering time back on the moat: healthcare coverage and analyst review, not crawler maintenance.
The outcome
Soria stopped operating a scraping stack it had already proven it could build, and got faster page monitoring out of the trade. For a research product whose value is measured in how early a number arrives, that speed goes straight to customers.
Monitoring pages that don't announce their changes? Context.dev turns any page, crawl, or sitemap into clean markdown, and Monitors can watch a page or a whole site on a schedule and webhook you the moment it changes. Start at docs.context.dev.
P.S. If you invest in healthcare and still track the sector in a spreadsheet, Soria puts CMS, NAIC, SEC, and 30+ state portals into one terminal that updates itself.