jjcm

8 ms·

91.
by jjcm 3mo ago
yea this was just straight into claude desktop / its standard tool usage, on "high" thinking. The ramen website is on "extra" thinking.
92.
by jjcm 3mo ago
Designs are here if you want to play with em: https://diffui.ai/app/canvas/68c5bb3d-467e-4841-b49d-c008e72... This was just from a prompt "A cyberpunk themed ramen food cart website. Should feature menu, loca
93.
by jjcm 3mo ago
All of these are using a build skill which specifies rules for building it, requirements to create a pixel perfect implementation, and tooling to help in that process. Here's the build skill / instructions I pasted in to both of t
94.
by jjcm 3mo ago
Here's another test of a cyberpunk ramen shop website. One thing I've found LLMs have a lot of difficulty with is angular cuts / elements that aren't easily representable with CSS. Cyberpunk aesthetics are generally a gr
95.
by jjcm 3mo ago
That's from my own tool. I left figma to build a diffusion-based UI tool. Here's a show hn post with some more info: https://news.ycombinator.com/item?id=48995754 Edit: Generation was down, back up now. Apparently
96.
by jjcm 3mo ago
I agree the fable version looks nice - the rounded hero image for instance. Opus though followed the source of truth better imo. The details are more present. Fable filled in the gaps for things it wasn't able to do (ie in the design t
97.
by jjcm 3mo ago
Doing testing with it now, specifically for image->html conversion. Previously Fable was the best at this, followed by Gemini 3.1 pro (a surprising #2, but Google has great vision models). Opus' results seem to be more accurate than
98.
by jjcm 3mo ago
> especially with OpenAI making so much progress with the efficiency of their models To be fair though, Sol tends to go off the rails sometimes. It's much less reliable than Fable in its outputs. It tends to be overzealous in its re
99.
by jjcm 3mo ago
Exactly right. You design with the diffusion model, then hand those designs off to an agent to implement. It's a lot like having an architect create plans for you before handing it off to a builder. In my (obv biased) experience, you e
100.
by jjcm 3mo ago
Thank you! That effect is easier than you'd think. Normal maps/depth maps are actually fairly easy to generate via diffusion models. Once I had the designs for the tarot site in diffui, I just copied the build plan, pasted it into
101.
Show HN: I left Figma to build a diffusion-based UI design tool(diffui.ai)
80 points by jjcm 3mo ago|13 comments
102.
by jjcm 3mo ago
For those curious, the LLM they provided participants with was Step 3.5 Flash: https://huggingface.co/stepfun-ai/Step-3.5-Flash
103.
by jjcm 3mo ago
The voice activity detection alone here is compelling - very useful for doing things like highlighting a speaker who's transmitting in realtime. At that rate the impact on perf will be so minimal that you could easily run it in the bro
104.
by jjcm 3mo ago
> taking that to a design system is a great way to make something more personal and original Indeed, that's why the "brands" feature exists. Once you have 5 screens on canvas you like, select all of them, right click and s
105.
by jjcm 3mo ago
They've unfortunately been radio silent since their release, and aren't active on socials at all. I've tried to contact them for updates / api usage, but haven't heard anything back.
106.
by jjcm 3mo ago
It's been really interesting seeing how LLMs perceive things differently than humans. I'm working on image->html conversion pipelines right now, and there are glaring issues LLMs run into that are obvious for humans. Any subtle
107.
by jjcm 3mo ago
This is really cool. What learnings would you say you could extract on the design side from the Watergate scandal, and how do you think StyleSeed applies to those learnings?
108.
by jjcm 3mo ago
Oh one warning - I didn’t optimize any of these for mobile. The hydroponics one does use some device motion for a cool effect though.
109.
by jjcm 3mo ago
I’d recommend a different approach if you’re looking for unique, non-LLM looking styles. Consider trying diffusion as a starting point, then feeding that into an LLM to build. Ie here are some results: https://html.non.io/ta
110.
by jjcm 3mo ago
I'm not sure I fully agree with this being a major vuln. There's a lot of up front scary text which was raising a lot of red flags until it actually discussed the "what". An actor has to place a malicious .exe in the use
111.
by jjcm 3mo ago
Yes you do. The teams that typically win the Molokai'i Hoe (the race between the islands) are the ones that ride the waves the longest. You'll stay on the same wave for 20-30min if you're good.
112.
by jjcm 3mo ago
The last thing you'd want is a balloon or an inflatable hampering your movements. Getting into the canoe as it's moving is a tricky, practiced and precise motion. A hyperbolic example would be imagine a pole vaulter with a safety
113.
by jjcm 3mo ago
two other unneccessary details: 1. that 15m one was the most fun I've ever had on a wave, ever. Think of it like going down a ski slope in a canoe, except the ski slope moves with you for miles. 2. we had one of the motorized lead boat
114.
by jjcm 3mo ago
Quoting this comment: https://news.ycombinator.com/item?id=48875676 > Don't think of these waves like the ones you encounter at shore. Open ocean waves are moving mountains. > It isn't this: /( > It&
115.
by jjcm 3mo ago
One fun thing you get to do in long distance outrigger canoe races in hawaii is crew changes. Generally, outrigger races have 6 people in the boat and a 9 person team. An escort boat will hold your reserve people, and then drop them in fron
116.
by jjcm 3mo ago
I've done open ocean crossings in canoes between the hawaiian islands, so I can answer from experience. Don't think of these waves like the ones you encounter at shore. Open ocean waves are moving mountains. It isn't this: &#
117.
by jjcm 3mo ago
This is no joke. I've done the crossing from Moloka‘i to Oahu (~45 miles) in a canoe several times, and those open ocean waves can get very nasty (largest I've dealt with were around 15m tall). I can't imagine the mental endu
118.
Show HN: Typeflag. Can you guess which Country's design system a font belongs to(typeflag.com)
3 points by jjcm 3mo ago|0 comments
119.
by jjcm 3mo ago
Not quite. These cost-per-task benchmarks report the cost of the task after the model gives its initial answer. The total cost is irrelevant, and isn't factored into the model's decisions - a run of the full benchmark for somethin
120.
by jjcm 3mo ago
I want a new bench - given $100 of api spend, how much can a model accomplish for a suite of benchmark tests? Give us something that measures a combination of efficiency and intelligence. I think this would allow for some interesting tactic

More ›