Document infrastructure for agents. Trusted by startups and F500s to turn their most important documents into data.
Joined March 2023
- Tweets252
- Following21
- Followers3.3K
- Likes480
One of OpenAI’s latest models looked at this invoice and called it legitimate with 95% confidence. The math was correct.
Except the fraud wasn’t in the math. Someone altered the PDF and inflated the PO balance by $30k.
Verify caught it by checking the file structure, not just by looking at what's on the page.
If you’re building agents for financial services, lending, or insurance, you can’t blindly trust an approval like this. Your agent can reason perfectly over a forged doc and still approve it, and one bad approval can cost tens of thousands of dollars and your customers’ trust.
Joon (@joon_kim96) from our engineering team just published a breakdown on how we made table parsing 4x faster with speculative decoding. Read it below!
we optimized inference for our table model and made it 4x faster
on a single H100, median latency on our Light table model dropped from 3.35s to 0.76s
our ML engineer @joon_kim96 wrote up our learnings and how we ultimately ended up using DFlash:
Operator-1 came out a month ago, and here are some snippets from customer calls about their experience with it!
2
9
455
Here are a few cool public projects built with our open-source UI kit!
1) @kortix CompanyOS - AI agents that work across your company’s tools, tasks, and shared knowledge
2) OpenMeld by @felixlu1018 - a shared workspace where people and AI agents collaborate and pick up where others left off
3) Jevbox, built by our own @andrewlu0 - a document library that organizes your files and answers questions with page-level citations
4) Neutree Agent Platform - a self-hosted platform for building and sharing AI agents, with Extend UI powering document previews
we open-sourced the Extend UI kit in June.
1,600+ stars later, our OSS documents are seen in:
> agent workspaces with inline spreadsheets
> research tools with source doc previews
> enterprise apps in production
we've kept improving it too including PDF annotations, redaction, forms and signatures, richer spreadsheet editing, live search in CSV previews, better doc navigation, and more.
it's been great to see where devs will take something once you give them the building blocks! if you're building a doc agent, you shouldn't need to build the UI from scratch.
repo in reply ↓
We launched Operator-1 earlier this month and we’ve been getting questions nonstop.
Here are the most common ones, answered by @bolau_
introducing operator-1, an autonomous document extraction agent
operator-1 saturates multiple 3rd party benchmarks, and delivers the most accurate performance on complex tasks that other solutions can't handle:
> long tables with thousands of rows
> lengthy files with context scattered across hundreds of pages
> dense diagrams and images
> much more!
operator-1 is a specialized agent we built that:
- runs within a custom workspace
- knows how to use dedicated tools (parse, search, zoom, code execution, validation) to handle any document
- learns via persistent memory
We’re going to start highlighting some of the amazing customers we work with and what they’re building with Extend.
First up: Checkr, a longtime Extend customer and a verification platform trusted by more than 140,000 customers.
Checkr runs many of its document workflows on Extend at 95%+ accuracy, with some approaching 100%, on par with their human Ops team.
Thank you to Adam and the @checkr team for the continued partnership. We’re excited to keep building together!
Here’s Operator-1 running on an 11-sheet MEP drawing set. It returned structured records for:
- 18 air terminals
- 23 luminaires
- 4 plumbing fixtures
Each one is linked from its floor plan tag to its schedule, with dimensions, model numbers, connection sizes, and the room it’s in.
Operator-1 is built for documents where the data lives across pages, like construction drawings, financial statements with long tables, legal agreements, and healthcare charts.
Learn more about Operator-1 and try it out today: extend.ai/resources/introduc…
We’re teaming up with our friends at @tryramp and @LangChain for a movie screening in SF this Thursday, Oct 22!
RSVP here: luma.com/27b59c8i
super excited to announce that we’re partnering with @tryramp and @LangChain for this thursday’s movie screening in sf!!
we're also hosting an after party from 10:30 pm till late to debrief the movie (so come join even if you've already seen it)
Think you can spot a fake document?
Test your fraud-detection skills live at @money2020 USA. The Extend team is at booth #6085 with a real-life demo of the Fraud Arcade.
AI-generated fake documents grew 500% last year, and humans can no longer tell the difference
That's why we’re launching Extend Verify, the first document fraud detection API built for agents
Verify catches the fastest growing types of document fraud, including:
- fully AI-generated documents
- localized AI edits (down to a single modified character)
- patched text, overlays, and in-painting from Photoshop, PDF editors, and coding agents
- digital tampering after a document was signed
- and many more
In our internal evals, Verify detected over 96% of AI-generated and tampered documents.
Verify is available today: extend.ai/verify/request-acc…
Introducing Extend Verify, the world’s first document fraud detection API built for developers and agents.
Request access today! extend.ai/verify/request-acc…
AI-generated fake documents grew 500% last year, and humans can no longer tell the difference
That's why we’re launching Extend Verify, the first document fraud detection API built for agents
Verify catches the fastest growing types of document fraud, including:
- fully AI-generated documents
- localized AI edits (down to a single modified character)
- patched text, overlays, and in-painting from Photoshop, PDF editors, and coding agents
- digital tampering after a document was signed
- and many more
In our internal evals, Verify detected over 96% of AI-generated and tampered documents.
Verify is available today: extend.ai/verify/request-acc…
Come say hi!
the @ExtendHQ team is at AI Engineer NYC this week! stop by booth #39 for a live demo of our SOTA parsing API, Operator-1, and the best merch in the game.
7
789
Should we do a bagel reranking?
Announcing Bagel-Bench: the definitive guide to NYC bagels.
Every Friday for the past 2 years, we’ve ordered bagels for the whole @ExtendHQ team. 1000s of bagels later, we’re publishing our findings:
S-tier: Apollo and Bagel & Schmear are SOTA, pushing the frontier on the taste × reliability Pareto curve. two of NYC’s leading applied bagel labs
A-tier: Brooklyn Bagel, NYC Bagel & Coffee House, and Murray’s are strong production-ready alternatives. The gap to the frontier is real, but we remain bullish.
B-tier: solid baselines, but we’d like to see more bagel R&D
C-tier: these shops have impressive distribution. We’d love to see that same investment in the core model
D-tier: significant work to do before we’d recommend consumption
like all benchmarks, this one is biased. unlike most benchmarks, we’re admitting it upfront.
what bagel should we eval next?
9
1,494
Meet Operator-1, an autonomous extraction agent that writes and runs code against your documents and gets better with every run.
It’s live today on every plan. Try it on your hardest docs.
introducing operator-1, an autonomous document extraction agent
operator-1 saturates multiple 3rd party benchmarks, and delivers the most accurate performance on complex tasks that other solutions can't handle:
> long tables with thousands of rows
> lengthy files with context scattered across hundreds of pages
> dense diagrams and images
> much more!
operator-1 is a specialized agent we built that:
- runs within a custom workspace
- knows how to use dedicated tools (parse, search, zoom, code execution, validation) to handle any document
- learns via persistent memory
1
21
2,235
