Feed aggregator

Article URL: https://angelina-yang.medium.com/bolt-news-coo-on-why-vibe-coding-is-the-wrong-word-and-what-non-technical-founders-should-d9e20ca1be7d

Comments URL: https://news.ycombinator.com/item?id=46935386

Points: 1

# Comments: 0

I historically have switched between Apple Notes and Obsidian.

I built mdnb as a hybrid of both— it's a fully native mac app that lets you edit markdown with the same interaction as Obsidian. I plan to keep it minimal & have no plans to add a plugin system like Obsidian. Instead will build a curated first class featureset into the editor, not unlike Zed editor.

It will always be free to use. Feedback is welcome!

Comments URL: https://news.ycombinator.com/item?id=46935379

Points: 1

# Comments: 0

Article URL: https://jrfernandez.com/mdserve-1.0/

Comments URL: https://news.ycombinator.com/item?id=46935374

Points: 1

# Comments: 0

Article URL: https://trevoro.net/posts/2026-01-02-software-cognitive-terminal-velocity

Comments URL: https://news.ycombinator.com/item?id=46935353

Points: 1

# Comments: 0

Article URL: https://ossama.is/blog/disparity

Comments URL: https://news.ycombinator.com/item?id=46935349

Points: 1

# Comments: 0

Article URL: https://iconstack.lovable.app/

Comments URL: https://news.ycombinator.com/item?id=46935347

Points: 1

# Comments: 0

I built this tool because I kept seeing random terminal windows flash on my screen for a split second. I could never open Task Manager fast enough to catch them or see what they were executing.

ShellScope is a local flight recorder for these transient processes. It monitors the creation of terminal instances (cmd, powershell, wt) and logs their activity to a local SQLite database.

The architecture consists of two parts: 1. A Python backend that uses WMI (Windows Management Instrumentation) to hook into Win32_ProcessStartTrace events. This allows it to capture the process name, the parent process ID, and the full command line arguments before the process terminates. 2. A Flutter frontend that visualizes the logs in real time.

Current features in this MVP: Captures processes that live for less than 100ms. Logs full command line arguments to help identify hidden scripts. Highlights suspicious keywords (like "-enc" or "hidden"). Stores all data locally with no cloud dependency.

I am currently using WMI for the monitoring layer. I am aware this has some latency compared to a kernel driver, but it was easier to implement for an MVP.

I would appreciate feedback on the architecture, specifically on moving from WMI to ETW (Event Tracing for Windows) for better performance.

Project Page: https://shiks2.github.io/shellscope

Comments URL: https://news.ycombinator.com/item?id=46935337

Points: 1

# Comments: 0

Article URL: https://futurisold.github.io/2026-02-08-what-bleeds-through/

Comments URL: https://news.ycombinator.com/item?id=46935307

Points: 1

# Comments: 0

I'm a solo inventor in rural Pennsylvania. Over four days last week, I filed 10 provisional patents (70 claims, $650 total) describing a complete hardware specification for a standardized AI compute module for robots -- what I've been calling a "robot brain."

The spec defines three standardized sizes (drone to surgical robot), a universal connector system (the Manufacturer Interface Module, or MIM) so any brain works with any robot body, and a hardware safety architecture where a dedicated safety processor on its own power rail physically controls whether the AI processors receive electricity.

That last part is the point. Every AI safety system I've seen is software watching software. This design puts a hardware kill switch between the AI and the power supply. The AI can't prevent its own shutdown -- it's the same Safe Torque Off principle that industrial motor controllers have used for decades, applied to AI compute for the first time.

The formal spec name is the Standardized Autonomous Safety Module (SASM). Think ATX for robot brains -- standardized form factor, universal connector, safety built into the hardware.

I built all of this working alongside AI, using an open-source context management system I created that gives AI assistants persistent memory across sessions. The tool I built to solve AI's memory problem became the tool that let me design a robot brain. Several of the patents describe the context management system itself.

Open-source tool: https://github.com/RobSB2/CxMS

Happy to answer questions about the hardware spec, the safety architecture, or the experience of filing 10 patents in four days as a solo inventor.

Comments URL: https://news.ycombinator.com/item?id=46935294

Points: 1

# Comments: 0

Article URL: https://www.frdmtoplay.com/ledger-lsp/

Comments URL: https://news.ycombinator.com/item?id=46935286

Points: 1

# Comments: 0

Article URL: https://github.com/TadPath/PUMA

Comments URL: https://news.ycombinator.com/item?id=46935272

Points: 1

# Comments: 0

At this point not only the front page of hacker news, but the new page of hacker news is totally inundated with blog posts about people complaining about or otherwise worried about about generative AI.

Are there any other general technical forums, where this is not the dominant narrative?

I say “generally technical” because places like less wrong and lobsters are not general in the same way as HN

Comments URL: https://news.ycombinator.com/item?id=46935259

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=6v4Juzn10gM

Comments URL: https://news.ycombinator.com/item?id=46935248

Points: 1

# Comments: 0

Article URL: https://spectrum.ieee.org/3d-modeling-blind-programmers

Comments URL: https://news.ycombinator.com/item?id=46935240

Points: 1

# Comments: 0

Hi HN,

I’m building an AI video playground where people can try different text-to-video models in one place. We recently added support for Seedance 2.0, ByteDance’s latest video model, and I thought some of you might like to experiment with it.

https://laike.ai/tools/seedance-2

A few things that stood out when testing Seedance 2.0:

It often generates multi-shot sequences instead of a single static shot

Motion can be quite smooth compared to earlier T2V models

It supports native audio generation

Prompts with camera movement or cinematic language tend to work well

The goal of the site isn’t just Seedance — it’s to let people compare different video models under a similar interface. I found it frustrating that every model lives in a separate UI with different credit systems and prompt formats.

Some honest notes:

Outputs are still inconsistent

Prompting quality matters a lot

Long narrative coherence is hard for all current models

Running these models is expensive, so there are usage limits

If you’re experimenting with generative video, I’d love to hear:

What prompts work well for you

Where current models fail

What tooling you wish existed around text-to-video

Happy to answer questions about the product or setup.

Comments URL: https://news.ycombinator.com/item?id=46935234

Points: 1

# Comments: 0

Article URL: https://www.forrester.com/press-newsroom/forrester-impact-ai-jobs-forecast/

Comments URL: https://news.ycombinator.com/item?id=46935223

Points: 1

# Comments: 0

Article URL: https://k2xl.substack.com/p/planning-at-meta-is-waterfall

Comments URL: https://news.ycombinator.com/item?id=46935216

Points: 2

# Comments: 1

Article URL: https://www.nature.com/articles/d41586-026-00388-0

Comments URL: https://news.ycombinator.com/item?id=46935214

Points: 3

# Comments: 1

Article URL: https://robot-daycare.com/posts/2023-07-18-balancing-leg/

Comments URL: https://news.ycombinator.com/item?id=46935196

Points: 1

# Comments: 0