It takes a screenshot, analyzes it through its Vision capabilities and returns historical-related feedback in under ~6 seconds. The smaller models even faster.

All done locally, privacy-friendly and offline available.

Tried including good installation instructions, if any is interested in trying it out. Using LM Studio.

Comments URL: https://news.ycombinator.com/item?id=47700476

Points: 1

# Comments: 0

We are currently researching automating a lot of the 'specs -> design -> code' pipeline with AI cloud agents.

Basically, we make new tasks in Notion that are shaped by a human through a /shaping skill, that then gets picked up by an AI Agent doing a design prototype with our design system and all that.

Here we add a designer/ux/pm/dev to see if the results are good enough, and then it gets send of to AI cloud agents again to build the entire thing into our codebase.

After build a PR is created that is reviewed, and of course re-triggered with prompt if not good enough or merged.

We are trying to get the design into Figma, since we are ending up doing a lot of iterations here to make the UX and UI fit better.

Also the /shaping skill feels very crucial in all of this. It needs to be really spot on and specific.

Any have experience with this orchestration or pipeline?

So far the idea it is for more "simple tasks", and larger tasks designers/uxers take directly before ai agents go in.

Comments URL: https://news.ycombinator.com/item?id=47530901

Points: 2

# Comments: 0

Video is speed up. I ran it through LM Studio and then OpenCode. Wrote a bit about how I set it all up here: https://www.tommyjepsen.com/blog/run-llm-locally-for-coding

I ran a fun experiment trying to beat the danish C25 index with just Gemini 2.5 in simulation.

I fed Gemini 2.5 with stock data, danish news articles, Wikipedia & Google trends top lists and much more like Eurostat data to see if I could get Gemini to beat the danish C25 stock index. (Novo Nordisk, Carlsberg, Mærsk etc.)

It could only see the current days data and backwards and was prompted to not take in forward data from its training data. It was prompted to follow these strategies among others::

- Take a longer term perspective - Be conservative with position sizes - Only trade when seeing a clear opportunity

It did pretty well, but ended up doing a couple bad moves selling at bad times - it almost feels like it panicked.

Data is from (01-01-2025 to 11-12-2025).

-

Any ideas on improving it? I think next step is using more a Chain of Thought process and multiple personas to evaluate each investment.

Comments URL: https://news.ycombinator.com/item?id=46807536

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=45365119

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=43842074

Points: 2

# Comments: 0

I didn't find a decent priced solution to add subtitles to my videos, so I ended up exploring how to use Whisper and FFMpeg to do it and found a pretty decent solution, that is both free and fast.

Hope someone else can benefit from it to

Comments URL: https://news.ycombinator.com/item?id=42682348

Points: 2

# Comments: 1

To advance my Tailwind CSS skills, I made a little quiz game to help me remember e.g. the height of h-48, how add line height etc.

Thought maybe others would find it funny as well and potentially help them become better at Tailwind :)

Comments URL: https://news.ycombinator.com/item?id=41233783

Points: 5

# Comments: 3

I've developed a simple tool to animate and design Bento Grids from scratch to final mp4 file.

I was as a product designer struggling in my job to create these Bento Grids using Figma and After Effects, so I made this tool to solve my own pain and thought maybe others would like it too.

It is using React, Vite, FFmpeg, Shadcn, Tailwind, Firebase, Cloudflare R2 and lots of other cool libs and tools that was a ton of fun to both learn and integrate.

There might be some bugs and issues here and there Im smoothing out, and currently the best experience can be experienced using Chromium, so let me know if you find any.

Hope you like it - check it out on https://motiongrids.com

Comments URL: https://news.ycombinator.com/item?id=41117910

Points: 1

# Comments: 0

Performance diff: https://ffmpegwasm.netlify.app/docs/performance

Yea makes sense with file naming and button. Let me try to iterate on that.

The mp4 to webm was actually the hardest to get working. Seemed like FFmpeg-wasm would stall at different times with some settings. I've updated it now to produce a bit better quality, but will definitely keep improving on this.

Comments URL: https://news.ycombinator.com/item?id=40994898

Points: 52

# Comments: 23

Makes more sense this way.