suchintan

7 ms·

31.
by suchintan 1y ago
That would be amazing!!
32.
by suchintan 1y ago
Very cool. The benchmark can be found here if you want to take a look at it: https://github.com/Halluminate/WebBench
33.
by suchintan 1y ago
We definitely plan to expand it. I want to get to ~10,000 for a reasonable benchmark. 15 blew my mind -- it's too easy to overfit that dataset
34.
Web Bench: a new way to compare AI browser agents(blog.skyvern.com)
34 points by suchintan 1y ago|9 comments
35.
by suchintan 2y ago
Definitely. It gives you a lot more control on which models you want powering these browser agents, which is the important part
36.
Show HN: MCP Server to let agents control the browser(github.com)
14 points by suchintan 2y ago|3 comments
37.
by suchintan 2y ago
Hey HN, we were playing around with MCPs over the weekend and thought it would be cool to build an MCP that lets Claude / Cursor / Windsurf control your browser: https://github.com/Skyvern-AI/skyvern/tree
38.
by suchintan 2y ago
I've seen people use semantic versioning to version APIs and SDKs https://semver.org/ We at Skyvern are still doing patch versions only
39.
Raw Dump of Skyvern's navigations and decisions while evaluating WebVoyager(eval.skyvern.com)
2 points by suchintan 2y ago|0 comments
40.
by suchintan 2y ago
I wonder if more companies should open source their eval model outputs alongside the eval results We tried doing that here at Skyvern (eval.skyvern.com)
41.
by suchintan 2y ago
Definitely need a newer benchmark. I couldn't find where browser-use published their run results (expected to see it here https://github.com/browser-use/eval ) We went ahead and published our full run at https:&#x
42.
by suchintan 2y ago
I'd love to chat to see how we can help! Here's my email: suchintan@skyvern.com We're working on 2 major improvements that will get cost down at scale: 1. We're building a code generation layer under the hood that will s
43.
by suchintan 2y ago
This is a great point -- the example we chose was meant to be a consumer example that we could relate with.. however a similar example exists for the enterprise which may be more interesting Let's say that you are a parts procurement s
44.
Skyvern Browser Agent 2.0: How We Reached State of the Art in Evals(blog.skyvern.com)
49 points by suchintan 2y ago|31 comments
45.
by suchintan 2y ago
It creates a feedback loop for the LLM to help it correct hallucinations or misunderstandings about how it's planned actions actually played out!
46.
Show HN: Skyvern 2.0 – open-source AI Browser Agent scoring 85.8% on WebVoyager(eval.skyvern.com)
9 points by suchintan 2y ago|3 comments
47.
by suchintan 2y ago
This is very interesting, thank you for sharing it What are your opinions on lists like this? https://github.com/mmccaff/PlacesToPostYourStartup
48.
by suchintan 2y ago
Yep. Makes a lot of sense. Your side project is super noble -- thank you for that
49.
by suchintan 2y ago
This is really cool. Do you have any interest in helping people auto apply to them? We can help you set it up with a really simple API call via Skyvern ( https://github.com/Skyvern-AI/skyvern )
50.
A practical example of YC's advice to launch early and often(blog.skyvern.com)
2 points by suchintan 2y ago|1 comments
51.
Using Cedana to cut time to first token by almost 50%(docs.cedana.ai)
1 points by suchintan 2y ago|0 comments
52.
by suchintan 2y ago
we started this last year. Back then, autogpt was more of a prototypical framework -- I'm sure it's improved dramatically since then. We ran into too many issues while developing with it (how do you run tasks effectively? How do y
53.
by suchintan 2y ago
Yep. This is totally fair feedback -- we're still a super early product and haven't had a chance to optimize the phone experience.. largely because it's tough to see the magic from the phone We'll improve it soon!
54.
by suchintan 2y ago
Yes -- in theory. You'd need to use our workflows feature to get that set up and chain a few tasks together to collect that information!
55.
by suchintan 2y ago
Great to see you on here Anton! Just curious, how do you differentiate from other open source competitors like n8n and Active pieces?
56.
by suchintan 2y ago
You're making some really good points here 1/ the current prompt + payload structure is definitely on the complicated end of the spectrum, but we've found that we can use an LLM to help generate this payload for our users The
57.
by suchintan 2y ago
1. Yes absolutely. But the issue is a little bit more nuanced than that. Websites without APIs don't have them for one of two reasons: (1) They want to protect their data (LinkedIn) or (2) can't be bothered to make an API (boutiqu
58.
by suchintan 2y ago
Depends on the scope of the changes. What did you have in mind?
59.
by suchintan 2y ago
We have an open issue for this right now -- we would LOVE some contributions here. The biggest problem until Llama 3.2 came out was that most (good) open source llms were text-only, and Skyvern needs vision to perform well This isn't t
60.
by suchintan 2y ago
Give it a try! It's very capable of doing simple tasks like logging in and clicking around. You'll need to prompt assertions like "Complete if..." and "Terminate if..."

More ›