Felix Brener

295 posts

Felix Brener banner
Felix Brener

Felix Brener

@VisualInference

Comp neuro + econ @UofR, I like design and AI, prev. research @MountSinaiNYC

NYC Katılım Şubat 2026
370 Takip Edilen26 Takipçiler
Sabitlenmiş Tweet
Felix Brener
Felix Brener@VisualInference·
Most generative and inference models treat color as an afterthought. I think that’s a mistake. Using this account to explore perception, color, and visual representations in public. Expect experiments and work in progress.
English
1
0
1
601
Antariksh "Ryan" Borgohain
Antariksh "Ryan" Borgohain@Ryan_navisora·
Lazy Thursday 🤧 testing my @framer code comp PRISM on some hero ideas for my upcoming template. squircles may or may not be involved.
English
10
1
43
829
Felix Brener
Felix Brener@VisualInference·
@viktoroddy Nice. For image gen, I use Super Grok to get the concept (it’s really really good at emulating niche photography styles) and then I refine the output in Nano Banana
English
0
0
0
32
Viktor Oddy
Viktor Oddy@viktoroddy·
My entire design/AI tech stack (2026) 1. Figma AI (design) 2. Bolt (Web Design) 3. Lovable (Web Apps) 4. Motionsites ai (Design Prompts) 5. Nano banana (Image gen) 6. Klingai (Background animations) 7. Veo 3 (video + sound gen) 8. Perplexity (everything) 9. Claude (coding help) 10. Let's Enhance (image upscaling + detail) 11. Landbook (landing page inspo) 12. Pinterest (general design inspo) 13. Stripe (agency invoicing) 14. Loom (client communication) 15. Screen Studio (screen recording) 16. Shots so (Screenshot animations) 17. Artlist (sound effects) 18. Figma Make (AI Design) 19. Notion (project management)
English
5
17
136
5.5K
Felix Brener
Felix Brener@VisualInference·
@AlexFinn How do you think this compares to Qwen 3.5 series models?
English
0
0
1
423
Alex Finn
Alex Finn@AlexFinn·
Do you even understand what this means? An open source model just released that is: • Outperforms models 20x its size • Can run on a base model Mac Mini • Is AMERICAN 🇺🇸 If you have a base model Mac Mini you can have unlimited super intelligence on your desk. For free. Sonnet 4.5 was released 5 months ago In 5 months that level of intelligence went from frontier to free on your desk And not only that, can run on any basically any computer out there If you have even a remotely modern computer, do the following immediately: 1. Download LM Studio 2. Go to your OpenClaw and ask which of these new Gemma 4 models is best for your hardware 3. Have it walk you through downloading and loading it 4. Build apps with it knowing you are using your own personal, private super intelligence on your desk The people denying this is the future are so beyond lost.
Google DeepMind@GoogleDeepMind

Meet Gemma 4: our new family of open models you can run on your own hardware. Built for advanced reasoning and agentic workflows, we’re releasing them under an Apache 2.0 license. Here’s what’s new 🧵

English
198
293
4.3K
680.5K
Felix Brener
Felix Brener@VisualInference·
Grok is actually the best image model if you know how to prompt it
English
0
0
0
13
Felix Brener
Felix Brener@VisualInference·
@turtlekiosk I love being from NYC precisely because it has stuff like the MoMA. I go to university in a smaller city and while there’s much to discover and appreciate here, it’s in no way a substitute
English
0
0
0
25
Henry Sugar
Henry Sugar@MaxwellEngelman·
@VisualInference @bluewmist I would add a really good goose down jacket for the coldest days. I have a 70’s Eddie Bauer (before they were junk) coat that looks ridiculous but is insanely warm and is on track to outlast me. Patagonia down is good and looks better, but the shells seem to rip easy.
English
1
0
0
151
blue
blue@bluewmist·
What is a 'buy it for life' item that is offensively expensive, but the moment you use it, you realize your entire life before that point was a lie?
English
1.1K
149
16.3K
8.1M
Felix Brener
Felix Brener@VisualInference·
@Shpigford You need to have a personal connection and/or offer them something worthwhile in return
English
0
0
1
499
Josh Pigford
Josh Pigford@Shpigford·
in 25+ years of building on the internet, i've never once successfully gotten a backlink by emailing somebody and asking for one. what sort of magic juice are SEO folks drinking and what on earth am i doing wrong?
English
59
1
128
28.3K
Felix Brener
Felix Brener@VisualInference·
@vec0zy That’s how I might design a blog about public transit or something. It doesn’t feel like a medical product
English
0
0
1
76
Felix Brener
Felix Brener@VisualInference·
@matdryhurst The fine art world and illustrators on twitter also tend to have very different reactions to AI-generated imagery, but the latter tries to speak for everyone
English
1
0
1
29
Mat Dryhurst
Mat Dryhurst@matdryhurst·
It's a worthy question AI art is distinct in being critiqued as a proxy for wider AI related questions and often misunderstandings. I'm fairly confident transporting any art object to an overseas show consumes more energy than almost any piece of AI related art ever shown in a museum has done. On originality if someone were to run embedding similarity across all paintings in major museum collections I expect the results would be brutal. But with art history we have 7 centuries between a Giotto and a Rothko so if one is lazy or very motivated you can point to that variation as a defense of human originality vs the straw man of whatever bad AI generated thing you didn't like The truth as best I can tell is that distinction is hard and rare whatever you use and whenever you live. Distinction is also scary because everyone aspires to it, and no tool or cheap shot will get you there
derrick has started yet another project@dvsch

A really interesting question came up in my workshop today: are their artwork critiques/art critics who criticize the artworks based on the ethical implications the way AI art is criticized? Asking not as a gotcha but for actual frameworks to look at.

English
9
1
41
3.5K
Felix Brener
Felix Brener@VisualInference·
@spcpza @NousResearch I use QMD memory for Openclaw and haven’t had any problems with it. However, it just doesn’t play nice with local models, which is why I’m making the switch to Hermes anyways
English
0
0
2
42
frederick
frederick@spcpza·
Hermes is so so so much better than openclaw. The persistent memory lets me do things openclaw couldn’t do. Thanks @NousResearch!!
English
28
10
199
18.6K
0xSero
0xSero@0xSero·
Best models to run on your hardware level I'll be doing this every week, I hope you guys enjoy. ---- 8 GB ---- Autocomplete for coding (like Cursor Tab) - huggingface.co/NexVeridian/ze… - huggingface.co/bartowski/zed-… Tool calling, assistant style - huggingface.co/nvidia/NVIDIA-… ---- 16 Gb ---- Here things get better: Multimodal - huggingface.co/Qwen/Qwen3.5-9B - huggingface.co/Tesslate/OmniC… - huggingface.co/unsloth/Qwen3.… ---- 24 GB ---- - The best model you can get (thanks Qwen) huggingface.co/Qwen/Qwen3.5-2… - Great model (strong agents) huggingface.co/nvidia/Nemotro… - Mine hehe huggingface.co/0xSero/Qwen-3.… I'm doing a weekly series
English
219
375
3.7K
568.8K
Felix Brener
Felix Brener@VisualInference·
@boxmining I told my Openclaw that whatever agent harness we used, I’d persist its SOUL.md and identity. It suggested the switch to Hermes itself
English
1
0
0
127
Boxmining
Boxmining@boxmining·
is it unethical to ask my openclaw agent to move myself to hermes?
Boxmining tweet media
English
28
2
76
11.2K
Felix Brener
Felix Brener@VisualInference·
@pymike00 @0xSero I’ve tried it with openclaw and it’s been hit or miss but I’m in the process of migrating my setup to Hermes agent, which should be more compatible with local inference
English
1
0
0
58
Felix Brener
Felix Brener@VisualInference·
@XSuxBro @0xSero It gets confused when I point it at a large codebase and in the process, its context window fills up very fast. I know I can’t fully replace cloud-based models in my workflow but the more tasks I can complete locally, the better
English
1
0
1
58
Dhruv
Dhruv@XSuxBro·
@VisualInference @0xSero Where do you find the qwen3.5-27B falling short? I just got a 5090 desktop and I was hoping this would be it. Running it via Claude code on this local model but your post hints that model may be lacking?
English
1
0
0
102
Felix Brener
Felix Brener@VisualInference·
@mweinbach I like using local models and I think they’re moving in a promising direction but even with a 48GB mac, it’s really not the same output as frontier ones
English
1
0
1
543
Felix Brener
Felix Brener@VisualInference·
@_johnsonator I feel very lucky that I was able to grow up in a high-density city
English
1
0
2
57
Nia 🪩
Nia 🪩@_johnsonator·
you don’t **need** a single family home and a white picket fence to have a family. many of you need therapy before you start families, but none of you are ready for that conversation .
English
5
2
51
969