engineer @ExtendHQ | prev @hindsight_ai @ycombinator @theteamatx
New York, NY
- Tweets 204
- Following 300
- Followers 543
- Likes 1,061
Excited for this launch - Verify runs on specialized tamper and AI fraud detection models trained on 1m+ pages of fake documents
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…
1
2
7
326
Andrew Luo retweeted
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
35
38
331
53,129
You can now use Cloudflare's Clef model with Jevbox
every company has a shared drive nobody can find anything in
we built a doc library that files itself and answers questions with the exact page + citation
it's called jevbox. built on extend parse + jev. and it's open source.
a few ways to use it:
- contract library for procurement and ops
- due diligence dataroom for VCs
- policy and compliance reference for HR
- clean context for your agents
MCP included with authz built in (think google drive permissions)
self-host it or one-click deploy to render, instructions are in the repo.
link in the comments!
3
2
12
985
every company has a shared drive nobody can find anything in
we built a doc library that files itself and answers questions with the exact page + citation
it's called jevbox. built on extend parse + jev. and it's open source.
a few ways to use it:
- contract library for procurement and ops
- due diligence dataroom for VCs
- policy and compliance reference for HR
- clean context for your agents
MCP included with authz built in (think google drive permissions)
self-host it or one-click deploy to render, instructions are in the repo.
link in the comments!
12
5
58
5,382
Today we're shipping Parse Auto-Router: it picks Light or Performance Parse for every page in a document, so you don't have to.
Clean pages go to Light, hard pages go to Performance. Set engine to parse_auto to try it today.
5
3
20
11,116
Edit fills PDF forms programmatically, with or without embedded fields.
We've improved detection on low-quality scans and skewed pages, made field labels more consistent, and redesigned filling to better handle corrupt files, revisions, and signatures.
The new form detection model is fine-tuned from the layout model behind Parse. Great to see our AI/ML team's work compound across products! @ameyaajoshi @joon_kim96
Our Edit API got a major upgrade with a new forms model delivering:
> better field detection, labeling, and filling
> improved corrupt PDF handing
> verifiable signatures that survive revisions
check it out and try it on your docs!
4
4
22
1,214
added a PDF editor to Extend UI
annotation, signing, forms & redaction, built on embedpdf
also a paper inspired color picker :)
extend.ai/ui/docs/components…
2
1
14
1,111
maybe just me but noticing coding agents have been patching libraries more frequently
fable found a bug with older versions of next js swc and regex code, components would render fine locally but not in production builds
209
today, we're shipping Extend Templates: 50+ document types, ready to go.
every template comes with the pipeline, processor settings, schema, sample document, and expected output already figured out. until now, teams rebuilt all of that every time, based on vibes and whatever expertise was locked in someone's head.
point your coding agent at one, import it into your Extend account, or fork it on GitHub.
browse the library and find your document type!
just shipped @ExtendHQ Templates with @andrewlu0!!!
50+ pre-built doc processing pipelines + a smart search library. We noticed AI teams ran into similar challenges, like the 12-page EOB w/ nested tables, or the bad receipt scan.
Now they can solve them with Extend in <1 minute.
— every page is agent-readable, so you can point Claude Code or Codex at one and the template ships itself
— import into your Extend account in <10s
— fork it, or grab the client code in TS / Python / Java / Go
Check it out! It's live now at extend.ai/templates
1
4
13
1,481
Andrew Luo retweeted
our workflows product has executed 100,000,000+ runs, and companies like Opendoor, Zillow, and Mercury use it to run their document pipelines. today, we're shipping Workflows v2.
very few document problems are a single step. the most common thing teams build is a chain: split → extraction → validation → routing (→ sometimes a human in the loop).
most teams end up writing orchestration code, retry logic, and state management to build this flow. workflows exist so you don't have to rebuild that infrastructure every time.
what's new in v2:
- a GitHub app for CI/CD: review, test, and deploy workflow changes like any other production code
- multi-file extraction: treat a packet as one unit of work instead of a pile of disconnected calls
- validations: catch bad output at the step where it happens, not three systems downstream
- routing logic: branch on extracted fields, confidence, or your own business rules
- agents: describe what you want in natural language, and agents can manage the exact same workflows as code
6
6
36
6,032
