A general for vibe coding, coding agents, AI IDEs, browser builders, MCP, and shipping prototypes with LLMs.►What is vibe coding?https://x.com/karpathy/status/1886192184808149383https://simonwillison.net/2025/Mar/19/vibe-coding/https://simonwillison.net/2025/Mar/11/using-llms-for-code/►Prompting / context / skillshttps://docs.cline.bot/customization/cline-ruleshttps://docs.replit.com/tutorials/agent-skillshttps://docs.github.com/en/copilot/tutorials/spark/prompt-tips►Editors / terminal agents / coding agentshttps://cursor.com/docshttps://docs.windsurf.com/getstarted/overviewhttps://code.claude.com/docs/en/overviewhttps://aider.chat/docs/https://docs.cline.bot/homehttps://docs.roocode.com/https://geminicli.com/docs/https://docs.github.com/en/copilot/how-tos/use-copilot-agents/coding-agent►Browser builders / hosted vibe toolshttps://bolt.new/https://support.bolt.new/https://docs.lovable.dev/introduction/welcomehttps://replit.com/https://firebase.google.com/docs/studiohttps://docs.github.com/en/copilot/tutorials/sparkhttps://v0.app/docs/faqs►Open / local / self-hostedhttps://github.com/OpenHands/OpenHandshttps://github.com/QwenLM/qwen-codehttps://github.com/QwenLM/Qwen3-Coder►MCP / infra / deploymenthttps://modelcontextprotocol.io/docs/getting-started/introhttps://modelcontextprotocol.io/exampleshttps://vercel.com/docs►Benchmarks / rankingshttps://aider.chat/docs/leaderboards/https://www.swebench.com/https://swe-bench-live.github.io/https://livecodebench.github.io/https://livecodebench.github.io/gso.htmlhttps://www.tbench.ai/leaderboard/terminal-bench/2.0https://openrouter.ai/rankingshttps://openrouter.ai/collections/programming►Previous thread>>>>108351521
;)
>>108377219how does one become able to evaluate the outcome holistically enough to deem it ready for production?
>>108377253ask ai to review it, and no I'm not being sarcastic
>>108377219is claude code the best when it comes to coding? and is there anyway to access a free opus 4.6?
>>108377219>focus on syntax>typing codeThat's like 1% of the total effort when working a ticket, nice bait I'm angry now
>>108377219thanks for having an actual OP with resources and shit
>>108377296>wagegoy thinks his opinion on programming matters
>>108377296you lost
>>108377281Yes to both
>>108377320for the second yes, where?
Vibe coding is the new gambling addiction.
>>108377440sauce?
>>108377377lmarena, proxies too
>>108377462What do you need sauce for? You're repeatedly pressing the slot machines, hoping for a jackpot.
>>108377486isnt that the same for ever single action we ever take?
>>108377486It's not any different than when I was making things before this. Just faster. You come up with something. Test it. Fix the bugs. Iterate. Endlessly.
>>108377486Gamblers come up with ways to influence the result. Now I'm into gambling too.
>>108377499Yes but before you had to now you don't
computer use mcp is cool until you forget it's on the model decides to just start looking at your screen
>>108377792Before I had to manually do a bunch of shit that the machine does for me. That's what computers should be right? Automating the bitch work so I can jack off and shitpost all day long while still getting paid.
>>108377797qrd
>>108377816
>>108377828thats super cool, what are you using? is it local? api?
>>1083778585.4 is the only model that is somewhat reliable for this so it's through codex.yes it has absolutely no security features yetlel
>>108377872doesnt that burn a fuck ton of tokens?
>>108377887it can, but it's not as bad as you'd think. most captures try to happen at 1400x900 (iirc) either through downsampling or window resizing.also i believe the token use issue will mostly be solved in 6-12 monthshttps://si.inc/posts/fdm1/
>>108377219The premise of OP pic rel is wrong. A good human cover doesn't need to think about syntax or any other aspect of how to do something. They focus completely on the design.
>>108377920no
>>108377909>trying to process video using llmswhat the fuck am i reading
>>108378008>claude, drive me home, i'm drunk
>>108378008retard, computer vision has been around for eons
>>108378045you probably could use llms to do video processing, but you'd wrap it in a harness that uses at a minimum stuff like yolo and things like pyscenedetect>>108378081yes using a vastly different architecture
>>108377909>white paper >no download>no githublol
>>108378091>blessed by roonso probably getting acquired for 10bn in a couple of weeks
>>108378109that's not fair, we need the tech, can you do something about it?
>>108377828>>108377909240 frames per 200k tokensthat's awful, I can't waste that many, I need it to vibecode software
>>108377281claude opus has a massive context window (one MEELION tokens) and there’s a graph out there showing it gets WAY less dumb after a million tokens compared to other modelson the other hand, ChatGPT xhigh was able to fix a problem that Claude Code couldn’ton the other OTHER hand, Claude Code is getting retarded this weekend because Pete Hegseth is their best salesman and they’ve got an influx of new customers…plus Dario said that their limits are 2x during off-peak hoursthe fucker wrote a plan to do things and look at the code afterward and shasum the output files and regenerate to make sure that the code changes didn’t change anything…and it just up and forgot to do the shasum part of the check
>>108377253do >>108377256use a different model, toowhat are your tests like?
>>108378354what are tests?
>>108378386ways to make sure your code didn’t fuck uplike, if you’re writing a temperature converter, you want to make sure that, at the very least, it says that 32°F is 0°C, −40°F is −40°C, and 212°F is 100°Cthe above is something you’d call a “unit test” and your programming language likely already has a pretty good test framework for running testsother times, you’ll have to get creativeI’m writing a program to generate filesif I’m having my LLM change my code, then I’m going to tell it to make sure that the generate files should not change after it changes the codeso it will shasum the files before and after and if they don’t match, it fucked up somehow
So is GLM-5 currently the best open model or is it just benchmaxxed? Is the coding plan good?
>>108377237holy shit :O this guy is a genious. free claude for life if it works out.
>>108377440>Vibe coding is the new gambling addiction.dude it really is. its not even something you can GET GOOD at. its like you're always project managing a fucking idiot. the idiot gets a little smarter every 6 months and sometimes they're so fucking autistic it does near what you want. but its never perfect or clean. but it is always autistic and fast.
>>108378493it gives less tokens than gpt what are you taking bro
>>108377486remember when the cope was "i-it's a stocastic parrot"? now it's "well it's achtually gambling or something"
>>10837812720% of my context window was used for this masterpiece. highlights include:drew the face in a single actionspent 10% of context trying to use the text tooleventually gave up, generated a picture of the text via command line (i don't even know how), and pasted it into the paint window and repositioned it10/10 tokens well spent
I'm kinda anti-AI but now I figured out a valid use case: security analysis of a code base
>>108378668For some people, it is gambling.For me, it isn’t. Maybe I’m not ambitious enough.
>>108378737the people who think it's akin to a slot machine are the midwits who have never written a good spec in their lives and use generic ambiguous prompts
>>108377219Best way to use an unlimited copilot enterprise sub?I'm using the vscode copilot extension and it works best when I use the preview Claude agent mode, but it's buggy as shit as it's still in previewPls don't say opencode it's shit
