WebDevCody

enter the era of shipping fast and overwhelming users with features they'll never use or even know about

we went from nerding out about what neovim plugins each other use into nerding out about what agent skills we use

@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.

Been saying this for months now. I don't know how this happened but the GPT subset of models, not even the codex models, just the regular GPT models 5.3 and on, are miles better than any of the Anthropic models at just basic true agentic software engineering, not vibe coding, true software engineering w/ clean code. The only caveat to this I will say is that somehow Claude is miles better than pretty much everybody else at UI/UX design. My flow has been typically the following: I still use all of my plan skills and my regular workflow that was inspired by people from the Claude Code team and the people that I saw on YouTube at these conferences. It's probably 5% of me doing any manual work just to stay in the loop a little bit. Let's say I have a new feature, for example. Everything goes through CodeX. I use the CodeX app with the newest GPT model for plan execution, review, etc. Then I dump it into either Cloud Design or just Claude Code, depending on if it's a very large feature or refactor, and let Claude Code clean up the UI/UX. It doesn't touch any of the backend code or any of the code-based architecture. It simply redesigns certain components or different UI flows. Did this the other night implementing Stripe and I did it in an hour and a half with custom onboarding flow, custom paywalls, flexible billing. It's a base subscription with usage-based metering and so much that would have taken me weeks to figure out. It's actually wild the world we're living in right now.

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

@StudioZamudio far better? name them

@webdevcody Quit using premade ADEs and Claude. Build your own ADE, or use anything but Claude, and you'll be free from your insanity. You Claudettes have gone completely insane. There's more models that Opus, GPT, and Gemini in case you didn't know. Far better at coding too!

@TomasVorel I use all three Codex, Cursor, Claude Code

@webdevcody I thought you were a Cursor guy?

@_cloud_kid I mean low is implementing like 90% of my requests, but there were two bugs opus couldn't solve, I tried various effort levels, context windows, different prompts. The moment I threw the bug into codex, it fixed it.

@webdevcody What happened to this? lol I know the struggle is real. x.com/webdevcody/sta…

@ChadMoran it wasn't context rot, opus just sucks at solving some bugs, but I'm 100% going to create that skill now and use it

@webdevcody Two things. Context rot. Create a /codex skill that shells out to GPT 5.5 via the CLI.

Warp just went open source One of the best agentic terminals available is now community-driven with a public roadmap and open contribution process OpenAI is the founding sponsor and the repo is managed by Oz agents handling the coding, planning, and testing so contributors can focus on direction and ideas The terminal that already uses AI to build your code is now using AI to build itself github.com/warpdotdev/warp

@briankb I used django & python on a work project for like 6 months but then moved back to node & react land

@webdevcody are you still using or learning django? It looks like the framework of choice for agenticjumpstart.com was tanstack which is well aligned for that site. when I checked your github it only showed a single project in python.

I've been using opus 4.7 low effort all night and it's actually amazing

@NathanWilbanks_ the custom backgrounds, this guy knows how to cook

I've been running the same AI agent every day since JULY 🚨 - 297 days running 24/7 - 993,115 seconds of compute time automated - 5,020,623,362 tokens generated - 127,743 workflows ran - 605,292 tool executions Comment AGNT below to steal my setup 👇

@TheCoderBtw bruh

@webdevcody And its Friday 🙂

what's that feeling called when you wake up in the morning and see this?

@rehanhaider I just sent it a screenshot of my app and gave like one sentence description of what it did, then it gave me better looking designs

@webdevcody What was the prompt?

I used Claude Design to redesign icongeneratorai.com, and within a few hours of deploying I got my first monthly subscriber. Let's see if I can revive this 3 year old project.

@Principal_ADE I've already redesigned like 3 of my sites with it

@webdevcody Not gonna lie, I watch your videos and I want to try it. i don't know why I'm hesitating. Ill have to wake up tomorrow and just send it

I'm officially addicted to cloud agents. Who needs an IDE when I can just prompt from the web, have it be tested / verified in an isolated environment, and one click to merge the pr when it's passing. I'm about to just prompt from my phone all day now.

If these benchmarks hold in real world testing, deepseek v4 is the cheapest model for the performance we have ever seen V4 pro: $1.74 in / $3.48 out per MTok V4 flash: $0.14 in / $0.28 out per MTok 8x cheaper than gpt-5.5 on output alone 93.5% livecodebanch, codeforces 3206, 1M context, MIT license, open weights The price argument just got a lot harder to ignore

DeepSeek-V4 just dropped on Hugging Face. no announcement. no press release. just live. Pro: 1.6 trillion parameters, 49B activated. Flash: 284B, 13B activated. both with 1 million token context. MIT license. open weights. trained on 32 trillion tokens. 3.7x lower FLOPs than V3.2. Codeforces rating hitting 3206 in benchmarks which if accurate is completely unreal. DeepSeek does not do press conferences. they just ship. huggingface.co/deepseek-ai/De…