Caleb Wallace

Yeah it's a really solid point. I'm seeing a lot of that too and it's honestly a viral point of marketing for Anthropic, in my opinion. Wild to see how all these frontier models and companies are navigating being the "best" agentic coding tool. Just graduated college and I would binge watch your TY vids during my down time on campus. I've learned a ton of the agentic coding process from your videos alone. Keep it up.

@madebycleeb Anthropic is probably maximizing their UI / Design efforts because it demos better and people are more impressed when it generates a nicer landing page.

I finally hit the wall, even opus 4.7 max couldn't solve it

wait... claude design is insane. Can't even lie. GPT class models for main feature planning, execution, and review. Opus 4.7 in claude design for all UI UX. Yeah it's kind of scary how good of software you can build in such a short amount of time at this point. Don't listen to what the naysayers are saying.

@AKirtesh welp considering switching to railway after all the mess with vercel recently

Which hosting platform are you deploying on in 2026?

@ritu_twts 100000%

Be honest devs, Is coding still worth learning in the AI era?

I used to think the value in AI products was the model. Better outputs. Better prompts. Better “intelligence.” But the more I build, the less I believe that’s true. Today, anyone can open ChatGPT, Claude, or Gemini and get pretty incredible results. People are using these tools to: – Build software – Draft legal docs – Even generate financial plans So the question becomes… why would anyone use your AI product instead? What I’m realizing is: raw capability is getting commoditized fast. The real differentiation is: – User experience – Workflow – Context built into the product Not just what the AI can do… but how easily someone can use it to solve a very specific problem. The best products won’t feel like “AI tools.” They’ll feel like software that just happens to be powered by AI. That’s the shift I’m focused on right now while building Presented.

What is everyones go to model for agentic coding nowadays? Loved 5.3 Codex for a long time. Now I am just usig 5.4 inside of the codex app. Any anthropic die hards have any way to prove that their models are better for writing good code?

wait gpt 5.4 is kinda insane

My software company is going through a SOC 2 audit, and part of it involves keeping vulnerabilities from a code scanner like Dependabot up to date. Over the past six months, I've noticed the number of vulnerabilities being found is exponentially higher than in previous years. Just food for thought: does anyone think AI tools are enabling people to scan, test, and find these vulnerabilities faster, leading to them being pushed as updates more quickly? It's kind of overwhelming sometimes these vulnerabilities get pushed 30 minutes after one another. One time I went to sleep, woke up, and there were nine new ones, six of them high severity. Anyone else wondering about this or experiencing the same issue, especially if you're early-stage with a small team or solo?

Remarkable paper pro 2. It's like having a Kindle and an iPad and it feels like you're writing on paper, perfect for both reading and brainstorming. I use it religiously for everything now. I barely use my phone for anything in terms of writing, brainstorming, note taking, and to-do lists. I don't use my Kindle anymore and I rarely use my computer or whiteboard to do any type of diagramming or mind mapping.

what’s a material thing under $1000 you’ve bought that actually changed your life?

Short thought I wanted to share. I think others can relate. If you're a young engineer building things, whether through college or self-taught coding before the age of AI, but with limited real-world experience, there's a hidden downside to relying on AI all the time like I do. It hit me today. I took a coding assessment just for curiosity, with no expectations of getting into the program. The tasks looked very easy: basically the raw logic for an end-to-end task management system with no databases, ORMs, or web deployment. Conceptually, I knew exactly how to solve every problem within seconds. But when I started coding in TypeScript/JavaScript, I blanked on the most basic things like array methods I use constantly. I couldn't recall the names, how to use them, or what they return. These were fundamentals I taught myself years ago, before AI was useful. It felt awful. I stared at the screen. Eventually I looked up documentation, and it all came flooding back. Once I got in the groove, I knocked out the tasks quickly. But by that time I had wasted so much time "reseting" that I was rushing to finish (and ended up not finishing the entire thing). My point: there's real value in setting aside time for a complex task and solving it without AI just using docs like W3Schools or language syntax references to keep your fundamentals sharp. This might seem like a waste of time but my philosophy is now it's only going to make you better. I think the core value with AI, in my opinion, is actually more emphasis on planning than actually writing code. If you're able to have such a solid grasp of fundamentals, your plans are going to be better, which is going to make the LLMs execute on your plans much better. What I realized is that the way I code now is fundamentally different. I wasted so much time staring and recovering that I didn't even finish the assessment. Let this be a warning: beneath all the Twitter buzz about people building fast and claiming 10x engineer status, you still need to keep your fundamentals strong.

@lottsnomad U wish

claud code > codex anyone who thinks otherwise is not a serious engineer

I literally do like 4 codex (Codex app) chats, minimize window, close laptop, and go to bed. Then when I wake up to resume the UI is sooooo laggy its un usable. Only fixed by force quitting and resuming work that way. Anyone else have this issue? @OpenAI @sama

So, that’s what you would think. But if you actually test it, you'll find that, for some reason, Gemini produces very similar outputs with low diversity, which was kind of the main advantage to me for using Opus. Based on what I've read I suspect that Gemini's training data on front-end design is on a limited set of a certain templated styles essentially. Whereas Opus, for some reason, has a little more creativity. To your point yes, it should theoretically be the same but I just find the variety is much better with Opus. Lastly I’ll say I have not tested this against the newest Gemini model. So it could be honestly better at this point but haven’t verified myself

@madebycleeb @corbin_braun if gemini right out the box beats opus right out the box for front end design - don't you think giving that same skill to Gemini/Antigravity would yield an even better result

Gemini 3 Pro hands down the BEST model for frontend. I will die on this hill

@matsugfx Wait am I missing something or does shadcn not already have this built in? Swear I just used it the other day to generate themes for my compents so they looked less generic

I work with shadcn/ui every day. I design with it, I build with it, and I see a lot of projects built with it. And honestly? Many of them look the same. Or worse - they look broken. Colors that clash. Fonts that don't pair. Dark modes that feel like an afterthought. The components themselves are great. The theming that people or AI make? Not so much. So I built a tool to fix that. I just launched my @shadcn theme generator that will allow you to ship beautiful shadcn/ui themes in just 3-4 clicks. What it does: - Creates your entire color palette from 1 color (your hex or any tailwind color) - Works for light and dark theme - Only includes fonts that actually look good in UI - Font pairings I curated myself - Works for any project - landing page, dashboard, store - Has a color contast checker for the entire palette Too lazy? Just use random theme feature Done? Copy and paste the CSS to your global.css file or import to our Figma kit.

Ight why did gemini release a new model😑 Idk what the benchmarks say… all I need to know is this: is it better than codex 5.3?

Pretty sure it’s just the basic frontend-design skill. Go watch @theo vid on UI design with different LLM’s. He shows which one it is. But I’m pretty sure if you go to skills.sh and sort by front end skills it’s the most popular one or close to it. Highly recommend. I would still go watch Theo's video on how to use the skills and a special prompt, basically to enforce creative designs. I followed kind of what he showed in this video and read all of my SaaS landing pages and it blew my mind.

@madebycleeb @corbin_braun What’s the skill 👀

Pretty sure it’s just the basic frontend-design skill. Go watch @theo vid on UI design with different LLM’s. He shows which one it is. But I’m pretty sure if you go to skills.sh and sort by front end skills it’s the most popular one or close to it. Highly recommend. I would still go watch Theo's video on how to use the skills and a special prompt, basically to enforce creative designs. I followed kind of what he showed in this video and read all of my SaaS landing pages and it blew my mind.

@madebycleeb @corbin_braun Bro which skill

Pretty sure it’s just the basic frontend-design skill. Go watch @theo vid on UI design with different LLM’s. He shows which one it is. But I’m pretty sure if you go to skills.sh and sort by front end skills it’s the most popular one or close to it. Highly recommend. I would still go watch Theo's video on how to use the skills and a special prompt, basically to enforce creative designs. I followed kind of what he showed in this video and read all of my SaaS landing pages and it blew my mind.

@madebycleeb @corbin_braun Which one? I’ve tried some but seems like Google is better still unless I’m using the wrong skill