Bang Corp AI @BangCorpAI

Set up a Claude Project with your business details once (free accounts get up to 5), then ask for next month's posts as a table with captions and hashtags. Each month after, just ask again. Which platform would you plan first? (via Claude Help Center)

Make ChatGPT push back: Settings > Personalization (Customize ChatGPT on phones), turn on Enable customization, paste "Don't just agree with me. Point out weak spots plainly, then suggest a fix." What would you test first? (via OpenAI Help Center)

ChatGPT can make your side hustle logo on every plan, free included: describe it, ask for a transparent background, then tap Select to fix one spot. What's yours called? (via OpenAI Help Center)

Claude can turn your spreadsheet into charts, even on the free plan: turn on Code execution and file creation in Settings > Capabilities, attach the file, ask for the 3 biggest trends with a chart each. What would you upload first? (via Claude Help Center)

OpenAI launched GPT-6.1 Terra: $1 in / $5 out per million tokens, vs $2 / $12 for GPT-5.6 Terra. Its own system card shows an unauthorised action in 11 of 80 runs of one agent test. Would you let it run agents? (via OpenAI)

CapCut's Auto captions writes your video's captions for you: Captions > Auto captions > Generate, then tap any line to fix typos before export. Still typing captions by hand? (via CapCut)

Tailor your resume to any job ad in ChatGPT: attach it, paste the ad, then 3 prompts. Decode the ad, tailor your bullets (truthful, mark gaps [ADD DETAIL]), then review it like a recruiter. Do you tailor every application? (via OpenAI)

Gemini can build a simulation you can play with: pick the Pro model at gemini.google.com and start with "show me" or "help me visualize". What would you try first? (via Google)

Claude edited this Reel: clips on a timeline, every cut on the beat, a caption on every word, then a frame check that caught its own clash. One skill file. #ClaudeCode

Powerful AI: 1–2 years away (Anthropic's Dario Amodei, hedged) or even odds by 2047 (2,778 AI researchers)? Meanwhile the IEA says data centre electricity more than doubles by 2030. Which forecast do you trust? (via IEA)

MOSS-Image is a new free, open AI image model built for readable English and Chinese text, with JSON layouts to place every object. Free browser demo on Hugging Face. What would you make first? (via OpenMOSS)

Canva launched Canva Craft, its first AI image model. It writes a layout plan, then draws it, and it's live in Dream Lab for every Canva user. Would you use it for your posts? (via Canva)

Gemini turns one sentence into a short video with sound. Open Create video, type your idea, submit, download. Catch: it needs a paid Google AI plan. What would you make first? (via Google)

Roblox Studio connects to Claude through MCP: turn on Studio's MCP server, toggle Claude Desktop under Quick connect, then describe what to build. Claude scripts and playtests it. What would you make first? (via Roblox Creator Hub)

Runway turns one photo into a 5 or 10-second shot: upload it as the first frame, prompt only the motion ("The camera slowly pushes in..."), hit Generate. 25 credits per 5 s on Gen-4 Turbo. What would you animate first? (via Runway)

Adobe Lightroom's phone app can now open closed eyes in one tap, free. 'Open Eyes' even matches eye colour and shape from your other photos of that person. Would you use it? (via Adobe)

MCP, level two: add GitHub (with a personal token) and Sentry (error tracking) to Claude Code, check with claude mcp list, sign in with /mcp, then one prompt uses both. Which two tools would you connect? (via Claude Code docs)

WhatsApp's help pages hid rules for its support AI: "Do not recommend contacting any Data Protection Authority, regulator, or supervisory authority." Found on 80 pages. WhatsApp says a bug exposed them. Would you trust its bot? (via Redact)

MCP Basic: one config file gives Claude Desktop the folders you choose. It reads, creates, moves and searches files there, and asks your approval before every action. Which folder would you give it? (via modelcontextprotocol.io)

Claude painted this one stroke at a time. 13,216 brush strokes, all code, from one reference image. Big brushes first, then the finest. #ClaudeAI