Percival Tran

Free Claude Code" = Claude's interface, someone else's brain The whole thing is essentially a proxy that routes Claude Code through other models (NVIDIA NIM, DeepSeek, OpenRouter, etc.) meaning you're not actually getting Claude. You're getting GPT-OSS, Kimi, GLM, or whatever free model you configure. The branding says "free Claude Code" but the intelligence behind it is completely different. For serious coding tasks, that gap matters a lot. What could make it better: An honest benchmark/comparison table showing how each free model actually performs on real coding tasks vs real Claude Sonnet, so users can set expectations instead of thinking they're getting a free lunch.

@RoundtableSpace Impressive demo. But the real question is what happens to the thousands of junior devs who were supposed to spend years learning Three.js and physics engines to get here.

A FULL VOXEL ACTION-ADVENTURE SCENE BUILT IN A SINGLE FILE WITH GPT-5.5. THREE.JS, REAL PHYSICS, MUSOU-STYLE COMBAT WITH UP TO 1000 ENEMIES ON SCREEN. ONE FILE IS NOW A GAME STUDIO.

@RoundtableSpace How about giving @Cristiano a landing page and prepare for World Cup? x.com/percivaltran/s…

WHAT ARE YOU BUILDING TODAY?

@VigoCreativeAI 说实话，大多数人卡在第一步——不是不会写 prompt，是不知道怎么「看懂」一张图。风格拆解这个能力，才是真正的护城河，AI 只是放大器。

昨天很多人问我： 这组「毛绒城市节日海报」到底是怎么做出来的？ 答案不是一句 prompt。 我真正用的是： 参考图 → 风格拆解 → JSON 化 → 换主题批量生成 今天把这个方法讲清楚。 想要最后那份「风格 JSON + 通用 Prompt 模板」的话，先收藏，这条看到最后。 👇

@ianneo_ai PSD 这个确实厉害，但让我真正开始思考的是——「设计师」这个职位，正在被悄悄改写。交付的东西看起来一样，但到达那里所需要的能力？已经完全不同了。

真的被 GPT 吓到了！ 我先让它生成一张海报，已经很像那么回事了 然后顺手追问：能不能直接给我 PSD 分层文件？ 结果它真的生成了，真猛 而且不是一张扁平图，里面有白底、人物、海滩双曝光、标题、文案、飞鸟、暗角、颗粒这些图层，尺寸和位置也对齐好了，拖进 PS 就能继续改 这一步很夸张 AI 不是只会出图了，它开始把可继续编辑的设计文件也交出来了 这...有点厉害，快去试试看！

@HBCoop_ One reference image. A full storyboard. An animated ad That used to be a creative director, a storyboard artist, a video production team, and a two-week timeline.

ChatGPT made a storyboard from a single reference image, then I used Seedance 2.0 to animate the storyboard shots:

@thatguybg The "5 in SF but a 10 in NYC" line is funnier than the AI score. But the real thing here is that ChatGPT just gave more structured, actionable appearance feedback than most people get from their closest friends in a lifetime.

chatGPT thinks I’m 7.8/10 hot that’s about a 5 in sf but a 10 in nyc. i’ll take it! prompt below 👇

@liu10102525 韩国专业色彩诊断一次要花不少钱，还要提前预约，还要专程跑一趟。 现在上传一张正面照，免费，在家就能做。 这不只是省钱的问题。它让一个原本只有特定人群才能负担的服务，变成了人人可及的东西。 AI 真正有价值的地方不是替代专家——是把专家级的服务带给原本根本不会去找专家的人。🫡

这个Skill绝了！！！ 韩国上万元一次的【个人色彩诊断与形象设计】，现在用这个Skill在家就可以免费拥有了！ 只需要上传一张你的个人正面照即可，男生女生都能用！ skill链接：youmind.com/skills/TUWIOmF…

10 repos marketers need: 1. Postiz: github.com/gitroomhq/post… 2. Mixpost: github.com/inovector/mixp… 3. Socioboard: github.com/socioboard/Soc… 4. BillionMail: github.com/Billionmail/Bi… 5. SEOMachine: github.com/TheCraigHewitt… 6. OpenClaudia Skills: github.com/OpenClaudia/op… 7. Refferq: github.com/Refferq/Reffer… 8. Weferral: github.com/WeferralHq/wef… 9. MiloAgent: github.com/topics/growth-… 10. Matomo: github.com/matomo-org/mat…

@RoundtableSpace Check this out @RoundtableSpace, preparing for World Cup: x.com/percivaltran/s…

24 minutes from the people who built it is worth more than 10 hours of third-party tutorials. Everyone else is teaching you how they learned Claude. Anthropic is teaching you how Claude was meant to be used. Those are two completely different things, and most people don't realize how much they've picked up bad habits from the first one until they watch the second.

ANTHROPIC DROPPED A 24 MIN WORKSHOP ON HOW TO ACTUALLY USE CLAUDE

That is not a skill issue. That is an information issue. That line is the whole tweet. Most people blamed themselves, thought they weren't creative enough, technical enough, or smart enough to get good outputs. Turns out they were just never told there were 6 elements and they'd only been using 1.

ANTHROPIC JUST PROVED MOST PEOPLE HAVE NO IDEA HOW TO PROMPT CLAUDE. Their applied AI team dropped a 24 minute free workshop. Not a creator who reverse engineered it. Not a Reddit thread. ANTHROPIC. The people who wrote the weights. And what they showed is uncomfortable. There are 6 elements to a properly structured Claude prompt. Most people are using 1. Maybe 2. That is not a skill issue. That is an information issue. And it has been quietly costing you every single day. The outputs that felt slightly off. The responses you had to rewrite 4 times. The prompts that worked once and never again. All of it traces back to the same 6 missing elements. The people who watch this 24 minute workshop tonight will understand something about Claude that most daily users still do not know exists. The people who skip it will keep getting 30% of what the tool is actually capable of and wonder why the results never quite land. I watched it twice. Then I built a Claude Skill that applies all 6 elements to every prompt automatically. No more thinking about structure. No more guessing what Claude needs. The framework runs in the background every single time. Full breakdown and skill setup is below. Bookmark this now. Watch the workshop first. Then read the guide. This is the one that compounds. Follow @cyrilXBT for the exact prompt architecture, Claude skills, and systems I use to get outputs most people do not believe came from one person working alone.

大多数人安装了 Claude Code，然后就这样用着——没有记忆、没有上下文、没有真正的工作流。 这个列表解决的不是功能问题。它解决的是"为什么我用了这么久感觉还是很笨"的问题。 Claude Mem 给它记忆。GSD 让它真正把事情做完。LightRAG 让它变成专家。n8n-MCP 让它接管整个工作流。 这8个 skill 装完，你用的已经不是同一个工具了。🫡

Claude Code 值得安装的几个skills都整理好了 1、Claude Mem - 给 Claude 装记忆，不用重复描述需求 github.com/thedotmack/cla… 2、Obsidian Skills - 让 Claude 真正理解你的上下文 github.com/kepano/obsidia… 3、GSD - 让 Claude 真正干活，不是瞎扯 github.com/gsd-build/get-… 4、LightRAG - 知识图谱，Claude 秒变专家 github.com/hkuds/lightrag 5、Superpowers - 给 Claude 装上超能力 github.com/obra/superpowe… 6、Everything Claude Code - 你需要的功能都在这 github.com/affaan-m/every… 7、n8n-MCP - Claude 接入自动化，一个提示词搞定整个流程 github.com/czlonkowski/n8… 8、UI UX Pro Max - 设计审美直接拉满，告别废于界面 github.com/nextlevelbuild…

@0x0funky 半天时间。4张地图、16只怪物、27种进化形态、战斗系统、音效、NPC对话——全部完成。 以前做一个这规模的独立游戏，至少需要一个3人团队花3个月。 现在需要的是一个有想法的人，和一个下午。 游戏开发的门槛没有降低。它消失了。🫡

全程使用 Codex + GPT-5.5 + agent-sprite-forge 半天時間做出一款類寶可夢的戰國妖怪捕捉 RPG。 包含： - 4 張地圖 (2道館+2場景地圖) - 16 種怪物，含進化型共 27 個型態 - 御三家選擇、野外遇敵、道館挑戰、Boss 戰 - 怪物捕捉、升級、進化、圖鑑、道具、補血點 - 自動生成地圖、角色、怪物 sprite、NPC、道館與 props - 也做了不同地圖/戰鬥的背景音樂和音效 從地圖設計、碰撞、NPC 對話、戰鬥系統、怪物數值、進化線、精靈圖、音效，全都是跟 Codex 用對話開發出來的。 用嘴巴加速遊戲開發的時代已經到來了。 遊戲連結在留言。

Hot take: Designers aren't cooked. Junior designers who only execute are. AI can clone the visual layer in minutes now: layout, fonts, colors, spacing, all of it. That part is gone. But someone still needs to decide that Ronaldo lifting the World Cup trophy needs to feel like a religious moment, not just a sports highlight. That's not a prompt. That's taste. And taste doesn't get cloned.

@arceyul This keeps coming up for a reason, most people are genuinely running Claude Code at 20% capacity without knowing it. The fact that an official Anthropic plugin exists specifically to fix this says a lot about how big the gap is between "installed" and "actually configured."

Claude Code es confuso hasta que instalas esto 🔖 Hay un plugin oficial de Anthropic llamado claude-code-setup. Te dice qué automatizaciones puedes montar (hooks, skills, MCP servers, subagentes…) y cómo configurarlas paso a paso. Básicamente analiza tu proyecto y te recomienda qué activar. Para instalarlo: /plugin install claude-code-setup@claude-plugins-official GUÁRDATELO.

@MatoToushi 「AI で作った画像は後から編集できない」 それが常識だと思われてた。 ChatGPT で生成 → Canva で編集可能なレイヤーに変換。この workflow が普及したら、デザイナーとマーケターの作業スピードが根本から変わる。 動画、絶対に出してください。リポスト確定です。🫡

うわ、これ「絶対に」日本じゃ見つけられない裏ワザを見つけてしまった… ChatGPTで作成した「１枚バナー」 Canvaで編集できるレイヤーに変換できちゃった。 たぶん画像作っても編集できないしなー、って思ってた人多いと思うのよ。 やり方を動画で解説しようと思ってるので、見たい人はぜひリポストしてください！ コメントも待ってます！

@marcusyul Claude can create the app. It can't create the insight that makes the app worth building in the first place. The barrier to building just dropped to near zero. The barrier to building something people actually want stayed exactly where it was.

🚨ÚLTIMA HORA: CLAUDE YA PUEDE CREAR APPS COMPLETAS DES DE CERO Sin equipo. Sin código. Sin dinero. Solo necesitas saber como usarlo. Aquí tienes 8 prompts para empezar a probarlo des de ya: 🔖 Guárdalos, lo agradecerás.

@D_studioproject The visual language is on point, dark editorial, Japanese typography, cinematic lighting. Feels like a real campaign pulled from a 90s archive. The gap between "I have a product" and "I have a brand that feels like this" used to be a six-figure agency retainer.

Making a commercial marlboro ad using GPT Image 2 x Seedance 2.0 on OpenArt you can make it adventure cinematic too⬇️