Some design decisions I made:

1. Local first: OCR uses Apple Vision, summarization supports local AI providers via Vercel AI SDK. Nothing leaves your computer. 2. Multiple Provider: exposes MCP so any coding agents can use it. 3. Swift menubar app: efficient, low-footprint 4. Blacklist apps: password managers, messaging apps (Slack, WhatsApp, Messenger), mail clients are on default blocklist.

Current Limitations: 1. Mac only. Mac-first is a feature. 2. Small local models with weak structured-output support will fail on generateObject. 3. Retrieval is LIKE-query keyword search. FTS and optional embeddings are on the list.

Demo video (6s): https://youtu.be/V75tnvIdovc

Curious what you think the right balance between exclusionlist allowlists. Happy to answer anything.

Comments URL: https://news.ycombinator.com/item?id=47858398

Points: 5

# Comments: 1

The sync engine feature looks very interesting to me. There have been quite a few products available on the market today, but none has achieved a dominant share yet. So if this is your main strength, I'd like to see more demos built local first.

Curious if you considered shipping the engine itself as a standalone infra piece.

Curious how much savings do you observe from using runbook versus purely let Claude do the planning at first. Also how the runbooks can self heal if results from some steps in the middle are not expected.

Comments URL: https://news.ycombinator.com/item?id=47114233

Points: 2

# Comments: 0

SOC 2 audit prep costs 20K/year, so I wrote a Agent skill to automate most of it. It goes beyond policy generation from templates by leveraging the abilities to access your codebase:

- asks contextual questions about your company and specifics about each policy - scans your source code - uses aws, azure-cli, gcp to check your cloud infra - writes shell scripts to fetch configs from your SaaS: Okta, Datadog, PagerDuty, Jira etc.

As a result you get:

- A set of policy documentation tailored to your company - All evidences are automatically versioned with your repo - Automatic evidence collection through Github Workflow Actions - Leverage Claude Code to add new integration scripts easily - No secrets leave your environment

This is targeted at small companies and startups who’d rather spend an afternoon with an agent than $20K/year on a compliance platform. It doesn’t replace the auditor and various aspects of compliance platforms, but handles a good trunk of the grunt work for free.

Some design choices I made: Each evidence collection script is a readable shell you can run locally: make it easy to unit test each script rather than relying on runtime agents. Organize frameworks & integrations so it’s easy to add new ones. Progressive Disclosure: only load context when needed

Repo: https://github.com/screenata/compliance-automation

Demo Video: https://www.youtube.com/shorts/EevpE6bKwhA

I'd love your feedback on:

If you’ve been through SOC 2, what controls were hardest to evidence?

PRs with new frameworks, evidence collection scripts with your SaaS tools

What would you like to see next

Comments URL: https://news.ycombinator.com/item?id=46959220

Points: 2

# Comments: 0

Our team has been working on a suite of tools that help developers to better document their work. We released an open source tool [DeepGuide CLI](https://github.com/DeepGuide-Ai/dg) that records terminal and automatically convert it to SVG which can be copied and pasted, and allow search engine to index.

Why dg?

* Demo GIFs and Videos are large. dg replaces them with animated SVG, up to 50–100× smaller than equivalent GIFs, ensuring your README docs stays lightweight and fresh. * Demos aren’t just illustrative; they’re re‑run and validated in github CI, so if something breaks, the PR fails. * Works across any Node.js 18+ environment: macOS, Linux, and Windows (via WSL). We tried to use bun to build it into a binary program, but it was flakey due to the reliance on foundational programs. * Open source: Built on solid and awesome OSS foundations like asciinema, node‑pty, and termsvg, all with permissive licenses.

It's easy to use:

```bash # Install via npm (recommended): sudo npm install -g @deepguide-ai/dg

# Or run without installing: npx @deepguide-ai/dg init ```

Once installed, you can just run `dg capture` which will start a pty that allows you to run any terminal programs. After exiting, the recording will automatically be converted to a SVG in your local folder. It streamlines the process of downloading asciinema, installing it, recording your workflow, and converting the cast to SVG into a single command.

Moreover, the recordings were saved into CI action folder. You can easily add them to your Github actions to re-run the workflow as part of CI.

If you’re tired of outdated CLI screenshots or bloated GIFs, dg might be just the tool for you — lightweight, automatic, and CI-friendly.

Do you think such tool is useful? Do you think it would help console devs to better document their work? Or is it useful to record CLI tools such as Claude Code or Gemini CLI sessions and share with others as SVGs? Would love your feedback, issues, or PRs!

Comments URL: https://news.ycombinator.com/item?id=44953244

Points: 4

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=2915867

Points: 4

# Comments: 0