anyg

Isn't /btw meant to be ephemeral?

I have found Fable 5.1 to be a much more natural communicator than prior Claude models

The biggest issue i found was the gap between the tire and rim. Otherwise, the max version is the best so far!

I've found that adding the words - "tell me in simple words" manages to improve the output. But, i have to keep repeating that

Wow, is this really a side-gig?! I would have assumed this would need a team of engineers to run

Not dumb at all. It's a whole field of active research - Speculative Decoding. A recent paper goes one level deeper with Speculative Speculative Decoding - https://arxiv.org/abs/2603.03251

Framer Motion has recently launched CSS Studio. It's a library that runs on your browser and gives you the ability to tweak the site directly in browser which fires off instructions for your coding agent to commit to code.

I have been thinking about better and faster ways to communicate with agents for UI related changes and this is the right direction I believe.

If it is a little bigger to incorporate a bigger chip antenna and some GPIO pins, it is going to be very useful for a lot of IoT projects!!

Couldn't we get a low-res version of this info by tracking the active window using a cli tool? For linux, there are several options. Not sure about Mac.

Another approach is to run OCR on 1FPS screenshots. Everything runs locally without draining the battery like an LLM would.

Off topic, but I assume the name braintrust comes from Creativity, Inc. Amazing book by Pixar co-founder Edwin Catmull.

Relevant XKCD - https://xkcd.com/86/

I had a similar experience.

It did not change the text on a hat (ended up changing 1 of 3 words).

On one occasion it regenerated the same image again, ignoring my instructions to edit.

I get the feeling that this model is optimised for images with people in it than objects or drawings etc

> from idea to solid working product

What platforms have you been using for this? Did you face any challenges in getting the backend setup?

How are the RNA sequences used? Are there any clinical trials running?

What is this?

This is pretty cool! What kind of users are you targeting? Technical folks or those dipping their toes into dev with vibe coding?

This is quite similar to what I was trying to explore with [1] & [2] yesterday. Currently, the only option I found is to use localStorage and give the user the option to manually export and import.

Hyperclay has given me some ideas. What I want is something like [3] but that the user only needs to install once. One electron app that can load our mini-apps.

[1] - https://news.ycombinator.com/item?id=44930814

[2] - https://news.ycombinator.com/item?id=44933897

[3] - https://www.electronjs.org/fiddle

He has two main projects - A & B. Currently A takes up 80% of his time and B takes up 20%. He uses the timers to track time on each project with the goal of transitioning to 80% on B.

This thought is what triggered me to try and build an offline web app - https://news.ycombinator.com/item?id=44930814

After building this, I realise that we need a standalone electron that can download and run offline apps that are more complex.

My colleague uses 2 kitchen timers to track daily progress on his two main projects. I was inspired by this to create an open source app but for yearly goals - https://github.com/anyg-me/goals-tracker

Linear is mostly used in collaborative setting right? Don't you use any of its integrations or workflows?

Figma-lite maybe for solo designers?

This is useful but not to decide whether to work on an idea or not. Most initial ideas are going to exist somewhere around the world but they may not focus on the angle that we are interested in. So, I think this is more useful as a way to understand existing approaches and see if our insights can lead to a meaningfully different solution.

I'm implementing a personal AI agent that is fully local inference with external memory that has access to my meetings, notes, Slack and Google Docs. Initially, it replaces Fireflies to capture notes from meetings and Slack.

Good to know - > Knowledge cut-off is September 30th 2024 for GPT-5 and May 30th 2024 for GPT-5 mini and nano

Also, the code demos are all using GPT-5 MAX on Cursor. Most of us will not be able to use it like that all the time. They should have showed it without MAX mode as well

Those who give it a try, I would really appreciate it if you upvote or downvote your result. I'm thinking of fine-tuning a small model as the next step to improve it. Would be interesting to have a GAN-like setup (?) where the fine-tuned model suggests comics and a large model uses its world knowledge to accept or reject

Pretty cool! I've been meaning to work on something similar. Congrats on the launch.

How are you planning to market this?

Yes. People like competing with people

I always go back to chess when I have these kinds of doubts. Deep down, we like human creativity and that's what made AlphaZero's chess entertaining for us as well as it wasn't as unnatural as traditional chess engines. But, we still enjoy playing and watching chess. Many new approaches have been discovered using powerful engines which probably would not have been tried out without them. The same thing can happen with math problems. We may have fresh takes on problems previously considered unsolvable, etc.