marcon680

Generally our use cases are completely different -- if a user is doing scraping, it's been structured scraping on a small (<50) set of sites where they need to be able to pull data from a website as if it was an API call, not as a way to web-crawl and get masses of training data.

We gate full access to the platform partially for this reason. We debated giving fewer than 50 free browser sessions, for example, and have already banned a few accounts from our self-serve today that were unidentifiable/"company accounts" without a legitimate web presence.

We have paying customers in production at the Growth and Enterprise plans :)

We've generally seen the "easy" flows not actually be easy. Workflows that have complex branching logic (shown when filling out the Aetna form in the first example in the video), structured scraping (the second example we showed in the video), and login/2FA/intelligent multi-agent scraping (shown in the last example in the video), are all things that are difficult to impossible to do with traditional automation libraries.

We also have an example of a complex, multi-agent workflow here that might be useful for you to look at: https://www.simplex.sh/blog/context-engineering-for-web-agen...

Yep we actually cache flows after first run! This makes flows that are closer to traditional RPA pretty much the same as using Playwright/Puppeteer.

From my understanding, Browserbase mostly provides remote browsers for their users. We also provide remote browsers, but with a lot more infrastructure on top (single/multi web agents, caching/reliability support, an agentic IDE/editor, etc).

Will fix, thanks for flagging!

Good questions! Will add these and put a data processors page tonight. We use Anthropic/OpenAI models.

Could I ask why the flow starts with a human logging in? Is it because you're using their credentials and/or have some sensitivity around storing their credentials? Or is it something to do with 2FA (we handle 2FA)? Or are you just storing the session data after they log in so you can re-use it for those few months you mentioned?

Re: Puppeteer automation as part of the script -- we have a feature we wrote for one of our customers that we didn't promote to production where you can define a deterministic action in the dashboard that allows you to paste in JavaScript, but we're likely not to push that to prod anytime soon. Could you explain your reasoning for wanting to use Puppeteer still? We've generally seen customers fully switch over to Simplex instead of relying on their original Puppeteer/Playwright scripts -- since we have action caching, the underlying script (click on div locator with this div id, etc.) is pretty similar to what you'd get using Playwright.

Yes, it's a huge pain from what we've seen. That portal is actually running in production for some of our AR/AP AI customers, and getting browser agents to properly parse that page was difficult -- Coupa injects all the DOM elements through a JavaScript <script/> tag like 2-5 seconds after the page loads, non-deterministically. :)

Currently no, it's just the three options (free vs. $2500/month vs. larger enterprise plans). We do usually offer 2-3 month ramp up periods for the $2500/month plans that are usage based.

To be transparent, we're an early startup, and a big part of that is user validation. We've been lucky to have companies as small as 5 people sign onto our $2500/month plan -- it shows some commitment on their side and helps us understand whether it's a real problem for our users. That's the same reason the $2500/month plan isn't self serve (you have to talk to us first).

We're definitely thinking of adding a pay as you go plan! But we aren't there yet re: our understanding of the market, if that makes sense.

Some of the things we're trying to differently are 1) making flows deterministic by using vision models that have a consistent output given an image and an input query and 2) breaking up flows into these smaller atomic actions in order to improve consistency as well.

RE: workflows w/ complex forms and tabs -- do you have some sites that are good examples of this? We'd love to see how Simplex does.

Thanks John! We should make that more clear in the docs. You can set browser=None on initialization and Simplex will create a local browser instance that can run on your local websites. We're not planning to be open source right now since a large portion of our product is custom vision models + inference speedups through hosting.

Browser Use is another YC company. Probably the biggest difference is that they're more agent focused while we're more lower level -- in the Claude Computer Use camp like you mentioned.

Thanks! We're happy to add more REST API functionality -- could you email us at founders@simplex.sh if you're up to talk further about what you'd like to see? RE: pricing, we're still figuring this out!

Yeah sure! We mentioned a little bit in the post that we stumbled upon this problem when working on a synthetic data contract. Internally, we were building software at the time to fully automate the process of creating synthetic data, all the way from purchasing assets to building scene layouts to rendering images. We realized after 2 months of building that there wasn't a great need for this problem nor could we build something minimal but feature-complete that we could give to people and iterate off of. We also saw the writing on the wall re: building 3D world models (see WorldLabs and NVIDIA's new 3D world models). We're excited to see where this goes!

So sorry about this, it looks like our servers went down overnight. Just fixed, should be good now!

We finetune our own VLMs for this -- unfortunately prefer not to share which ones we use specifically! ClickClickClick looks awesome, have you heard of FerretUI (https://arxiv.org/pdf/2404.05719)? Pretty similar idea.

Yep, any websocket URL works. I see that Browserless offers a websocket URL, so you can use them! You just need to pass in a Playwright browser object into the Simplex constructor.

Yep, we also run all our python code in sandboxes and a few other security measures!

Site loading is pretty now slow due to a combination of 1. traffic we're currently getting, 2. running a remote session, 3. running a few large vision language models, and 4. adding waits to allow pages to load/allow you to view your search results. We're working on cutting our latency significantly since it leads to a better development experience.

RE: feedback mechanism -- yep, a feedback mechanism is definitely something we're thinking of adding. Since we use VLMs that are trained to always output coordinates (i.e., they don't have a way to say "not on the page"), we're probably going to try fine tuning with some negative examples to see if we can build that feedback mechanism in.

One way to hack the scrolling to an element is to first run extract_bbox on a natural language description (in your case for GitHub it might be "follow button") then take the Y coordinate of that element and scroll that number of pixels. I just wrote this bit of code that I tested and it brings the contribution graph into full view:

    simplex.goto("github.com/spro")
    coords = simplex.extract_bbox("follow button")
    simplex.scroll(coords[1])
    simplex.wait(2000)
    image = simplex.extract_image("green tile graph")

But then it incorrectly picks the code review/submissions/etc. graph as the green tile graph -- we'll look into it!

re: frequency graph typo -- just pushed a fix, thanks!

Thanks, we hope so! That's an interesting conversation -- I'd argue that the graph-based no-code solutions you're referring to are for a different set of people than those commonly found on HN. As a developer I didn't particularly want to work with those tools, so we built this to supercharge our own code-based flows instead. I actually don't think the node-based flows are horrible at all since they successfully enable non-developers/less technical people to use agents/build easy automations.

Ah, we've whitelisted some websites to prevent abuse -- amazon.in wasn't one of them. I just whitelisted amazon.in for you and tried the search query -- looks like it works!

Hey thanks! We think we're pretty different from Skyvern as Skyvern provides a full agent loop for users + is no code. We wanted to build something at a lower level (i.e., no high-level planner that chooses what tasks to do for you) and we wanted to be able to program our own logic alongside the intelligence part because using code is what's natural for us as developers.

I hadn't heard of UI Vision but just took a look at it -- it also looks like a no-code solution that's a Chrome extension, so I'd say the main differences are the same as the differences w/ Skyvern -- we're lower level and meant to be used by developers.

I'd add that we're also able to directly extract parts of websites that have no official API -- for example, an image of the GitHub contribution graph like I show in the video demo.