DanyWin

7 ms·

1.
Show HN: Reduck GEO, open source Skill to measure and optimize Claude citations(github.com)
1 points by DanyWin 1mo ago|0 comments
2.
Show HN: Connect Claude to LinkedIn, Twitter, etc. with Reduck MCP(reduck.ai)
2 points by DanyWin 2mo ago|0 comments
3.
Show HN: An MCP to create and call Playwright scripts as tools on your extension(reduck.ai)
3 points by DanyWin 2mo ago|0 comments
4.
Show HN: Duck Talk – Real-time voice interface to talk to your Claude Code(github.com)
6 points by DanyWin 7mo ago|0 comments
5.
Vibe coding as a VC(kevinkuipers.substack.com)
24 points by DanyWin 1y ago|30 comments
6.
Local Embeddings and LLMs Can Outperform OpenAI and Gemini for Web Navigation(blog.lavague.ai)
2 points by DanyWin 2y ago|0 comments
7.
by DanyWin 3y ago
Yes, we are working on that! We are preparing to release a feature for people to enable telemetry to contribute to a decentralized and open dataset to train and evaluate models for Selenium code
8.
by DanyWin 3y ago
Exactly! In the future, testers could just write tests in natural language. Every time we detect, for instance with a vision model, that the interface changed, we ask the Large Action Model to recompute the appropriate code and have it be e
9.
by DanyWin 3y ago
Very interesting indeed! We are thinking of developing an extension that would connect the browser to LaVague so that actions can be sent to the extension and be executed locally, thus bypassing their barriers
10.
by DanyWin 3y ago
You are exactly right! As I wanted to have a solution that works with many LLMs out of the box, I focused on chain of thoughts and few shot learnings. Lots of paper show that fine-tuning only helps with steerability and form ( https:/&
11.
by DanyWin 3y ago
Thanks! Funny thing, we did not use Vision models but text only with the HTML of the current page. However, we intend to add it to boost performance
12.
by DanyWin 3y ago
Thanks a lot! Love the support <3
13.
by DanyWin 3y ago
This is just the beginning, but it is indeed on the roadmap! Once we solve browser automation, we intend to support other integrations to further facilitate automation of workflows
14.
by DanyWin 3y ago
It could indeed have an impact on jobs, just like any productivity gains have destroyed jobs. However, the net gains, in my humble opinion, could be phenomenal. Imagine all the time, mental energy and money spent on navigating through the l
15.
by DanyWin 3y ago
Here we just provide natural language instructions and the LLMs generate the code appropriate at a given time. If the site changes, we can regenerate the code using the same instruction, so unless the site changes a lot, it is quite robust
16.
LaVague: Open-source Large Action Model to automate Selenium browsing(github.com)
378 points by DanyWin 3y ago|95 comments
17.
Show HN: LaVague – Automatic Selenium code generation from natural language(github.com)
1 points by DanyWin 3y ago|0 comments
18.
LaVague: Text2Action AI pipeline to turn natural language into browser actions(github.com)
3 points by DanyWin 3y ago|0 comments
19.
'sleeper agent' AI assistants can sabotage your code without you realizing(theregister.com)
2 points by DanyWin 3y ago|1 comments
20.
Best framework to create synthetic data for finetuning small models?
1 points by DanyWin 3y ago|0 comments
21.
Show HN: OSS, private and in-browser Conversational AI with local model(github.com)
2 points by DanyWin 3y ago|0 comments
22.
Zero-trust AI APIs serving Llama 2 70B inside enclaves(blindllama.mithrilsecurity.io)
35 points by DanyWin 3y ago|8 comments
23.
BlindLlama: Open-source Zero-trust AI APIs for transparent and private AI(github.com)
2 points by DanyWin 3y ago|0 comments
24.
Is retrieval necessary/sufficient to solve “hallucinations”?
1 points by DanyWin 3y ago|1 comments
25.
Show HN: TCO Calculator to compare on-prem LLM deployment vs. OpenAI and Co(huggingface.co)
6 points by DanyWin 3y ago|0 comments
26.
On demand vs. Reserved instances for LLM fine-tuning
1 points by DanyWin 3y ago|1 comments
27.
Big vs. small GPU clouds for fine-tuning LLMs
12 points by DanyWin 3y ago|16 comments
28.
Why open-source is necessary but not sufficient for AI transparency(blog.mithrilsecurity.io)
1 points by DanyWin 3y ago|0 comments
29.
by DanyWin 3y ago
There is still a design decision to be made on whether we go for TPMs for integrity only, or go for more recent solutions like Confidential GPUs with H100s, that have both confidentiality and integrity. The trust chain is also different, th
30.
by DanyWin 3y ago
Exactly! It's not sufficient but it's at least necessary. Today we have no proof whatsoever about what code and data were used, even if everything were open sourced, as there are reproducibility issues. There are ways with secure

More ›