Currently your prompts are processed by our server hosted at www.gopeek.ai. This was meant for velocity and early iteration while we're getting data models right.

We're already working on whitelists and blacklists for files, topics, memory exports, and can even work on self-hosted/locally hosted versions so please let us know what is a non-starter on this front.

Sensitive data is already wiped and never stored such as keys, tokens, and PII.

Sensitive data is wiped and never stored such as keys, tokens, and PII.

If it's a non-starter for most users, we'd definitely build out encryption, and local storage formats.

Cloud was for product velocity for ourselves, not to be malicious. Again we put it together in 4 days.

Appreciate the warning here.

My cofounder and I have gotten tired of CC ignoring our markdown files so we spent 4 days and built a plugin that automatically steers CC based on our previous sessions. The problem is usually post plan-mode.

What we've tried:

Heavily use plan mode (works great)

CLAUDE.md, AGENTS.md, MEMORY.md

Local context folder (upkeep is a pain)

Cursor rules (for Cursor)

claude-mem (OSS) -> does session continuity, not steering

We use fusion search to find your CC steering corrections.

- user prompt embeddings + bm25

- correction embeddings + bm25

- time decay

- target query embeddings

- exclusions

- metadata hard filters (such as files)

The CC plugin:

- Automatically captures memories/corrections without you having to remind CC

- Automatically injects corrections without you having to remind CC to do it.

The plugin will merge, update, and distill your memories, and then inject the highest relevant ones after each of your own prompts.

We're not sure if we're alone in this. We're working on some benchmarks to see how effective context injection actually is in steering CC and we know we need to keep improving extraction, search, and add more integrations.

We're passionate about the real-time and personalized context layer for agents. Giving Agents a way to understand what you mean when you say "this" or "that". Bringing the context of your world, into a secure, structured, real-time layer all your agents can access.

Would appreciate feedback on how you guys get CC to actually follow your markdown files, understand your modus operandi, feedback on the plugin, or anything else about real-time memory and context.

- Ankur

Comments URL: https://news.ycombinator.com/item?id=47371284

Points: 18

# Comments: 7

We connected the Context 7 MCP Server and it tends to miss a lot for us. Maybe we're doing something wrong?

I'm surprised there isn't something out there that can write perfect integrations for any API yet.

Comments URL: https://news.ycombinator.com/item?id=45279527

Points: 2

# Comments: 0

Aka, I'd argue that elon musk drive gives you a greater chance to become an L8 engineer, where as elon musk drive may make you an NBA player, but at a significantly lower chance.

Basically, you are born with the required genetics + surrounded with the right coaching from a very early age or you won't get into the NBA right?

Are there other factors you think contribute? What percentage weight would you put on each factor if you had to guess.

I think you're likely correct here. There's definitely exceptions to this, but those exceptions fit into the 99.9% figure you have here.

2) We aren’t complaining. We’re curious how others view this topic and space because it’s a contentious topic. We recognize that we might be able to address the larger issue of lack of compensation for websites being scraped by facilitating a win-win marketplace (only loser is proxy providers).

We recognize it as an unrewarding, tedious, and time-consuming thing humans have had to do until the latest abilities of browser agents.

As we built and learnt more about the industry we started to understand the underlying problems. For 99% of web sites web scraping isn’t the problem, the lack of compensation is.

We think there’s actually a better way to do this. If there’s enough demand, we can facilitate a rev share between agent scrapers and websites. Scrapers will pay less than what they pay for proxies and websites get a new revenue stream.

These are our thoughts at least so far. We aren’t ashamed of what we’ve built by any means in the way your comment implies lol. We want to see if we can benefit both parties in a win-win marketplace.

Maybe we don’t need to protect websites that rely on advertised content? Maybe it makes a better internet?

Based on what you’re saying here it sounds like good content will inevitably go behind paywalls in this case, and the number of paid subs people need to have will only continue to explode. Aka no free internet. Maybe that was bound to happen one way or another.

And all LLM providers will pay their data providers directly by API or something?

original internet (DARPA) was meant for communicating research and science. Maybe we return to that being the primary thing that will continue to be free?

I’ve invested more research and engineering into organizing users browser tabs than I’ve seen any other team do really. I’ve always wondered why big teams haven’t invested more time into this. They must have data that shows it’s not a profitable problem, idk.

Either way, 10 or more research institutes (including Harvard, CMU, and WashU) have shown how severe and widespread the problem of a poorly organized browser workspace and how detrimental it is to productivity. So I took a crack at building a tab auto-pilot system that doesn't require any pre-set categories, runs continuously, and is accurate across any topic. I didn’t build a new browser because viewing a webpage isn’t the problem—finding information nested in your browser and organizing your browser tabs into ideas and action items efficiently is.

The autopilot system handles about 80% of the initial browser tab organization that you either hate doing yourself. Then, I built a workspace that provides a better UX for organizing, ideating, and navigating your daily explorations and work online. We combined the two into a single tool and called it Peek.

Our AI autopilot system:

- Organizes tabs into suggested groups as you work. - Summarizes web page collections with one click. - Finds related pages from your browser history as you type. - Keeps related tabs grouped together.

With the workspace users can:

- Quickly prioritize, organize, and switch between tasks using drag-and-drop. - Add notes to topics, keeping ideas and tabs together. - Search across all notes and web pages. - Close and open all tabs related to a topic. - Save tabs for later easily from your browser.

So far, our suggestion system seems to blow Arc's "Tidy" feature and Chrome's Experimental "Group Tabs" feature out of the water. I'm curious how valuable an integrated workspace + built-in autopilot system like this is for HN!

I think the main thing you'll need to figure out is how to get users from a youtube video to this software. Maybe a Chrome Extension?