You can give Prism a try at https://prismvideos.com - I'm excited to hear your feedback.

Nevertheless, things are trending more in this direction, and AI influencers will soon become the norm. Brands should be required to disclose when their marketing is AI.

It's worth mentioning that AI videos on Prism (and on any platform) do not have to be purely prompt to creative. For example, a brand designer can take an existing creative for a billboard for example and then use AI to generate images of this creative at a train station, in the Louvre, at a bus stop etc (without actually going there and shooting images).

We've found that Seedance is good at photorealitic faces, Kling is fantastic at generating audio (highest quality model in terms of syncing character's face to the words they say imo), and Sora is great at UGC.

Would be curious to see your script.

- Code Compass indexes a GitHub repo with AST parsing, enabling agents to understand where developers are developing codebase context [1]

- Given a current linear ticket, Code Compass's agent provides context about the relevant codebase locations, implementation patterns, and starting points needed to complete the task and feeds those instructions directly to your coding agent [2]

- Given a current ticket, Code Compass’ agent identifies the next ticket to work on that will maintain as much of the engineer's existing codebase context as possible [3]

Why we built it

As engineers, we often struggled with deciding what to pick up next, especially when deep in a particular section of the codebase. Each context switch slows us down, and a lot of valuable attention can be lost when switching too often between unrelated tasks. Context switching is also expensive for AI models: each switch requires re-ingesting the same files for reindexing, driving up token costs and more importantly wasting time. We built Code Compass to help engineers seamlessly flow from one meaningful task to the next, maintaining momentum and minimizing friction, and to save both humans and models from unnecessary context resets.

Tech implementation details

- Code Compass indexes your entire GitHub repo using AST parsing to capture the code structure and developer context.

- It also syncs with your Linear workspace, mapping tickets to relevant code areas based on historical changes.

- Our agent combines this insight, leveraging up-to-date context to ensure suggestions always build on your current work.

- We utilize modern LLM architectures and lightweight vector databases to enable real-time, context-aware recommendations.

- Use Cursor command tools and manage branches and create PRs with each linear ticket.

Challenges we overcame

Visualizing code graphs proved much trickier than we thought. Representing complex dependencies in a meaningful way for both the AI and human user took multiple iterations. We took inspiration from projects like GitHub Next's repository visualization (https://githubnext.com/projects/repo-visualization/), eventually building a visualization layer that helps users quickly see project hot spots and context flows.

Potential Extensions

- Codebase + Developer understanding - Find who has expertise and triage tickets accordingly.

- Project planning - Architect new projects with developer and codebase understanding

- Expanding integration to include additional issue tracking systems (such as Jira and Trello), further streamlining workflows for a wider range of teams.

- Other languages (currently only supports typescript)

Try it yourself

You can sign up at https://codecompass.sh. We truly welcome your feedback and are eager to see how engineers use it to streamline their workflows.

[1] https://youtu.be/0EJl8b3fWYg

[2] https://youtu.be/hpxFagxHqu0

[3] https://youtu.be/fGLIlh3q1-k

Comments URL: https://news.ycombinator.com/item?id=45737812

Points: 2

# Comments: 0

One risk with these new standards for agent auth - which we will of course support if our customers want it - is that the websites that need them the most are the least likely to adopt them.

The main use cases for browser agents are for paying utility bills on old government websites or finding receipts for an expense report on a website without an API. There is a no reason to use browser agents on a website like Linear for example. A developer is better off integrating via API or MCP.

Therein lies the main challenge; the websites where browser agents are most useful are the same websites that are least likely to adopt new technology (it was their not adopting new technologies that made them good candidates for this browser agents in the first place).

I think this new standard is awesome, but I fear that the websites that support it will be those websites that didn't need it in the first place (because they could just as easily add an API).

When our agent signs in, we input the forwarded otp code to get access.

Right now, we're focused on building connectors for our customers, which has not yet involved Captcha solving.

We spoke to browser agent developers and saw people copying and pasting credentials and even credit card numbers directly into model system prompts. We were surprised that there wasn’t a better way to give agents access to websites on a human’s behalf. Moreover, we noticed that every company had to build infrastructure to manage OTP, TOTP, and MFA and that auth remained a significant hurdle in agent reliability. We wondered if this was a boring part of the problem of building web automations that someone could automate away.

We started working with Casco, an autonomous security testing company, to enable their agent to access customer sites. Before a pentest, Casco makes a request to Prism’s API specifying test user credentials, a domain, and a login method. For example, give me an authenticated session for the account rajit@prismai.sh for OpenAI via OTP code over email. Our agent logs in on their behalf (without exposing credentials to a model), and we download the cookies and send them back in the response.

To maintain speed and reliability, we use playwright in most cases to login (which gives us speed), and we fallback to AI on failure (which gives us reliability). We have a number of websites we support out of the box and add new scripts as the number of websites we need to support grows. We are working on a way for the agent to update the existing playwright script on failure, so our scripts always stay up to date.

To try our api, you can use our API playground docs.prismai.sh/api-reference/endpoint/login to sign into x.com with the following API key: pk_54abb1cd0a637eb973ed690416e71a953e98f2ea839cf16529bbfa41a41bc016 .

We’d love to learn more about how other developers give agents access to their accounts. We look forward to everyone’s feedback and comments.

Comments URL: https://news.ycombinator.com/item?id=45377810

Points: 21

# Comments: 15