New York City, NY
- Tweets 162
- Following 175
- Followers 98
- Likes 655
hiring someone to throw parties for a living!
@ExtendHQ doesn't host boring events. We've hosted intimate founder dinners, rage rooms, Hot Ones wing tastings, and an upcoming full theater rent-out for the Nathan Fielder / Elizabeth Holmes documentary (shoutout @bolau_)
we’ve grown revenue 6x YoY to 8 figs with 200%+ NRR, powering document ingestion for ai teams at Square, Zillow, Brex, Mercury, and Checkr
apply below! dm me if you have any questions :)
6
11
345
I talked to every person on their way out, trying to make sure every person got a goodie bag. So many people said this was the best office-warming event they've been - here's to the next one!
Scenes from our office warming party at our new Union Square space last month.
Thanks to everyone who came by!
7
219
Darren Chow retweeted
today, we're launching our new Parse Router
extend now runs a classifier on every page, and decides whether to run our performance or light engine for parsing (on a page-by-page basis)
tldr:
- most documents have a mix of complex pages intermingled with simpler ones. The router classifies and routes each page so you get the best blended performance x cost for any given document
- you pay per page for whatever was used: 0.5 credits for our Light engine, 2 credits for Performance
- our classifier works by scoring each page on layout, text quality, tables, form fields, and checkboxes
- on RealDoc-Bench, enabling the parse router scored 89% (close to Performance's 90.9%) while using significantly fewer credits
Parse Router is now our recommended default for most customers with mixed document payloads in mortgage, insurance, healthcare, and legal
it's live today, set the engine to `parse_auto` and give it a shot on your own docs
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.
9
10
30
7,841
parse auto-router is a game changer for teams with mixed complexity docs. Auto-router picks from our SOTA light parse or performance parse options for every doc so you can valuemaxx to the fullest!
try it today!
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.
1
4
201
as ex-consulting, while there will always be outliers, I think being a founder requires a degree of delusion and risk-seeking that someone who started in banking/consulting generally won’t have
v bullish on early hires from this profile tho—disproportionate gain embedding their learnings into an early co
A lot of VCs (myself included) used to shit on the ex-banker/consultant founder archetype.
But I'm increasingly realizing I was wrong. In an age where building is easier and distribution/execution is still hard, these programs are the perfect training grounds.
2 years at McK/GS/etc. teaches you how to grind, how to sell to enterprises, how to work effectively in a team, and gives you a deep understanding of business models. Exactly what it takes to succeed as a founder in 2026.
2
186
Come mingle and learn from some of the smartest teams building the multi-modal future! Reach out to @bolau_ if you're interested!
we're hosting a technical panel in NYC at the end of this month with @brexHQ and panelists from @Baseten and @vercel!
we’ll dig into open-source models, model routing, keeping token costs down, and when smaller, task-specific models are the better choice
rsvp below!
1
2
187
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!
2
4
16
2,248
Darren Chow retweeted
come stop by our booth at YC demo day!
we have:
- signature coffee drink menu
- custom 3d printed logos of your startup
- best merch in the biz
- you can say hi to @bolau_
8
5
55
9,866
Darren Chow retweeted
Astra shows a significant improvement over the GPT-5 family on long-array extraction, one of the hardest document extraction tasks.
here's what we're seeing:
> 77.7% bank, 72.8% legal, and 61.9% clinical mean per-document accuracy. this is +24–36 percentage points over GPT-5.6 Sol.
> Astra maintains high accuracy on larger arrays before performance drops. on clinical docs for example, astra maintains 100% accuracy through roughly 6,500 expected extracted values. Sol drops from 84% at roughly 3,500 values to 14% at 5,000.
> but it still struggles with the largest docs. Astra gets only 9–28% of expected values correct on the largest docs, compared with roughly 0–3% for earlier GPT-5 models.
3
3
14
686
the @ExtendHQ team is coming to SF for demo day on the 8th and 10th!
we’re going to be giving out coffee and some exclusive YC themed drinks 👀 (dw we have actual baristas on site) and a bunch of our cool merch to hand out && @Jingg_n_Tonic has been working HARD to 3d print some of the current batches logos as memorabilia <3
3
2
22
1,908
The new Extend MCP is now live, allowing developers to work with any API available in our platform in all agents that support MCP.
With a simple config change or prompt, your agents can directly work with our platform to parse, extract, split, edit documents and build complete workflows.
As a developer-first company building for devs and their agents, we made sure our MCP is designed specifically to support AI agents.
Try it on your docs today! docs.extend.ai/mcp
5
1
16
1,478
Light parse is a game changer for so many customers with high volume use cases. Give it a try today!
today, we're launching Light Parse
Extend can now parse documents up to 70% cheaper than other solutions
- light parse starts at $0.00625 per page, and scales down significantly with volume discounts
- it's SOTA on Databricks OfficeQA Pro (a 3rd party agent benchmark) among all non-agentic parsing providers
our ML/AI team trained smaller and more efficient layout, table, and form vision models to enable significant cost reduction without impacting accuracy on many types of documents
live today for all users!
4
153