Luke

6.8K posts

Luke banner
Luke

Luke

@ldo_dev

Building @LogoLattice to make the best in browser logo creator 🎨 Building @FlowInvoice so you can get paid faster and stress less 💰

Присоединился Aralık 2010
459 Подписки2.1K Подписчики
Закреплённый твит
Luke
Luke@ldo_dev·
Creating a new logo shouldn't take ages. I wanted something fast, simple and free. So I built Logo Lattice. 100% free. 100% browser based. 100% ownership of creations. Link in first comment 👀
English
108
76
1.6K
244.9K
Luke
Luke@ldo_dev·
@vaademola Yeah the tests are weird too. Not at all what I would expect 🤣
English
1
0
1
4
Victor Ademola
Victor Ademola@vaademola·
@ldo_dev Nice. It's better to do real testing than relying on assumption
English
1
0
1
5
Luke ретвитнул
Luke
Luke@ldo_dev·
I'm integrating LLMs into Logo Lattice, but not for image generation It gets a brief ("minimal fintech mark, teal + charcoal") & our schema It return JSON: shapes, points, grid snaps, fills Testing GPT/Claude/Gemini What would you try for structured spatial JSON?
Luke tweet media
English
8
1
12
433
Luke
Luke@ldo_dev·
@bas_fijneman GPT5.5 and Opus 4.7 are by far the best two so far. I’m mid write up of my tests. I also will run some more benchmarks on text based logos too. This is just the start of the can of worms I’ve found 👀
English
0
0
0
7
Bas Fijneman
Bas Fijneman@bas_fijneman·
@ldo_dev Hmm I would advise staying with 4.7 for this for now! I see to many inconsistencies atm. Also a lot of work done that I didnt ask for
English
1
0
1
17
Luke
Luke@ldo_dev·
LLM testing update for Logo Lattice: Opus 4.8 is struggling the most. Inconsistent coordinates, breaks schema rules, can't hold geometric precision across attempts Complex shapes: decent. Simple shapes: inconsistent. Maybe I need model routing based on logo complexity 🤯
Luke tweet media
Luke@ldo_dev

I'm integrating LLMs into Logo Lattice, but not for image generation It gets a brief ("minimal fintech mark, teal + charcoal") & our schema It return JSON: shapes, points, grid snaps, fills Testing GPT/Claude/Gemini What would you try for structured spatial JSON?

English
1
0
4
93
Luke
Luke@ldo_dev·
@yusukelp I actually got it to generate me a Rubik’s cube the other day. I’ve done so much work on the LLM integration for Describe!
Luke tweet media
English
0
0
1
4
Yusuke
Yusuke@yusukelp·
Good morning builders☀️ Pitch your app in 5 words👇 I’ll go first
English
10
0
13
221
Luke
Luke@ldo_dev·
@MythThrazz 55k views a day on average is so hard to do without consistently posting engagement farming or striking gold on a super viral post 😂
English
1
0
2
22
Marcin Dudek
Marcin Dudek@MythThrazz·
I think its not a problem to get 5M impressions overall. It’s a challange to do so without: - spamming - cheating - crypto - sports - ragebaiting - politics What’s left is unique trivia and wholesome content.
English
4
0
9
130
Luke
Luke@ldo_dev·
@zuess05 Have a great day man!
English
1
0
1
27
Luke
Luke@ldo_dev·
@stijnvwx It'll come! You got this, just stay at it :)
English
1
0
2
26
Stijn
Stijn@stijnvwx·
Goal was 1K by end of May. Hit 750 today. Missed the target, but I'm way further than if I'd aimed low. That's the thing about big goals: You either hit them or you grow into someone who can. Still counts as a small win today.
Stijn tweet media
English
9
0
14
159
Luke
Luke@ldo_dev·
@theo It's really bad at converting text prompts to an editable, grid snapped vector though. Like really bad.
Luke tweet media
English
0
0
0
15
Theo - t3.gg
Theo - t3.gg@theo·
It is possible that Opus 4.8 is a much better model than I give it credit for, and it is held back terribly by Claude Code. Sadly I cannot confirm this myself without spending thousands of dollars because their API prices are absurd and you can't use your sub anywhere good.
Theo - t3.gg@theo

I haven't used Claude Code much recently. Are these write failures just, like, expected? I haven't seen this many random tool call errors since I last used a Gemini model

English
77
6
619
52.4K
Luke
Luke@ldo_dev·
@stijnvwx Yeah I agree. I've just finished my benchmark and I definitely think 5.5 is the best at my use case. It's interesting though, because other models you think would be good, are really bad. I'll write up my thoughts on them
English
1
0
1
10
Stijn
Stijn@stijnvwx·
@ldo_dev I would go for GPT-5.5. Think it would work best
English
1
0
1
20
Luke
Luke@ldo_dev·
@vaademola Benchmark is being run!
Luke tweet media
English
1
0
1
12
Victor Ademola
Victor Ademola@vaademola·
@ldo_dev Looking forward to that, especially the logo/text cases.
English
1
0
0
23
Taniya
Taniya@Taniyatweets_·
which programming language you learned once but never touched again?
Taniya tweet media
English
198
5
186
13.3K
Luke
Luke@ldo_dev·
@edytwithme Yeah it definitely is miles ahead. I’ll have to create a logo and try lots of them. And yes the cost of me is about £0.17 per generation at the moment. So definitely not cheap
English
0
0
1
7
emily
emily@edytwithme·
@ldo_dev imo GPT 5.4/5.5 is still king for anything related to visuals im still trying to balance between cost and performance for my ai (the concept is similar to yours i think)
English
1
0
1
71
Luke
Luke@ldo_dev·
@adelbucetta Yeah I definitely think it’s the way forward. It also comes with the added benefit of being able to edit it after it’s been generated. At least in the way that I handle it
English
0
0
0
10
Adel Bucetta
Adel Bucetta@adelbucetta·
@ldo_dev the honest answer is that most people overcomplicate lattices with unnecessary image gen. your approach to embedding schema directly into lattice points is the way forward, imo
English
1
0
1
17
Luke
Luke@ldo_dev·
@zuess05 Nooo anything below gpt 5.4 sucks at it 🤣 5.5 is the best so far. Opus 4.8 is also trash. I think I will plan out a kind of benchmark thing and write up my findings!
English
1
0
1
42
Suhas
Suhas@zuess05·
@ldo_dev very interesting, im still wrapping my head around it but any decent llm should be good it doesnt have to be ultra good imo
English
1
0
2
32
Luke ретвитнул
Luke
Luke@ldo_dev·
Logo Lattice has a proper home Not just a free editor. A product you’ll come back to Explore is your hub for guides, features, and updates, with a cleaner site and in-app support Same obsession with snapped grids and clean marks. Now the site matches x.com/ldo_dev/status…
Luke@ldo_dev

✏ Copy. Paste. Repeat. Why do logo tools force you to manually enter hex codes over and over? Brands just shipped in @LogoLattice - Save palettes - Load them instantly - Keep every logo consistent No more "wait... which blue was the right one?" 😳 x.com/ldo_dev/status…

English
3
3
17
710
Luke
Luke@ldo_dev·
I definitely think I'm going to have to write something up with my findings. Especially considering how easy OpenRouter makes it to try different models. I'll have to give the same prompt to the models and see how they fair. Especially with different types of logos with marks/text
English
1
0
1
13
Luke
Luke@ldo_dev·
@farrux_hewson I'll have to have a play with them as a language model. On OpenRouter they're listed as Image generation ones. I shall topup my account and report back 🫡
English
0
0
1
18
Farrux Hewson
Farrux Hewson@farrux_hewson·
The way you described me the internal logic last time, it’s similar how SVGs work. If you open any SVGs source code you will see different paths, colors and different properties. You probably know this already. Recraft looks like a great model for Vector graphics, they have image creation capabilities as well, but that’s not their focus. If you look at SVGs that it creates, they are very polished. I am guessing that since SVG technically are paths with different properties, it should perform good within LogoLattice as well. Not an expert in this though and might be totally wrong here :)
English
1
0
1
35
Luke
Luke@ldo_dev·
Yeah I've been bashing 5.5. I've spent quite a bit on it and it's by far the best in my testing so far. It can create real 3d shapes in space (it also creates the back of 3d objects which is fun). 5.4 struggles slightly more but is still usable. Below this it really starts to fall over. I've also not been that impressed with my findings with Opus 4.8 either. It seems to fail quite drastically at spatial structures
English
1
0
1
24