One Free Guide. 200+ Ways to Make Money With AI.
Most people use AI to save time. The people getting ahead are using it to create new income streams. The best part? You don't have to figure it out alone.
Unlock 200+ proven AI business ideas, side hustles, and step-by-step strategies you can start today. Whether you're a beginner or experienced, this free guide helps you turn AI into real income - not just better prompts.
Today In Ai
1 Palmier launches the first Mac video editor built around AI models.
The Y Combinator-backed startup just revealed its Mac-native video editor where Claude or Codex can generate, organize, and trim footage directly in-app, no tab-switching required. It integrates with Seedance 2.0, Kling V3, and Grok Imagine out of the box, and the base editor is free to download. The launch video cleared 1.5M views in the first 24 hours.
2 Kimi Work adds a Goal Mode that runs until the job is done.
Moonshot AI's Kimi Work desktop agent now stays on a task until it hits your objective, not just until the session times out. You set the goal, track progress, and redirect as needed. The beta is aimed at long-horizon, multi-step work that would typically require babysitting an AI across several hours. See how it works.
3 Adobe's Firefly AI assistant now executes tasks across every app in the suite.
Describe what you want, and Adobe's Firefly AI Assistant handles the multi-step execution across Premiere, Photoshop, InDesign, and the rest. The update ships with four new creative skills and an upcoming expansion into ChatGPT, Claude, Gemini, Copilot, and Slack. For anyone already inside the Adobe ecosystem, this is a meaningful shift in how the tools work together.
From The Frontier
Fusion beats solo. OpenRouter launched Fusion, a tool that runs your prompt through several top models simultaneously and synthesizes their answers into one. The company says the strategy significantly outperforms any single frontier model on its own. The idea of orchestrating models rather than picking one is becoming its own category.
Copilot Cowork goes wide. Microsoft's agentic workplace assistant is now available to any Microsoft 365 subscriber globally. Microsoft is pitching it at 30-40% lower cost per prompt than Claude Cowork, with model choice and usage-based billing built in. Effective prompting tips here if you want a running start.
Leaderboards reshuffled. Google's Gemini Omni Flash now leads both text-to-video and image-to-video on Arena's rankings. GPT-5.5 and Claude Opus 4.7 are heading a new agent category. Claude Fable 5 briefly held the top spot across most lists before being suspended, leaving a gap that several models are now racing to fill.
Meta's morale problem. Meta launched AI Mode on Facebook, a search tab drawing on publicly shared content across its apps, on the same day that CTO Andrew Bosworth reportedly told staff that morale is near the lowest he has seen in 20 years. Shipping fast and keeping people motivated are two different problems, and Meta appears to be struggling with both at once.
Readers also loved: Satya Nadella's essay on how AI is rewiring work (65M views), plus an AI selfie comparing Europeans vs. Americans that keeps the comment sections busy.
What people are actually watching and sharing
Midjourney's hardware pivot. Midjourney dropped a four-minute teaser showing how its water-based ultrasound machine will work, and the internet has not stopped talking about it since. 8M views for a hardware demo from an image generation company is a signal worth paying attention to.
PhD-level prompting. A four-part Stanford prompting technique designed to push AI models into deeper, more structured analysis has been doing the rounds at 1M+ views. Worth saving before your next major report or big decision.
Block's internal AI. Jack Dorsey's Block revealed its Builderbot system, an AI suite handling 200,000 operations per day internally (1M views). The full breakdown is worth reading if you want a concrete example of what AI-native operations look like inside a real company.
Offline AI, hand-cranked. A startup called CrankGPT built a 100% offline AI box powered by a hand crank. It is either the most creative solution to AI's data privacy problem or the most elaborate joke on the internet. Possibly both.
Influencer campaigns on autopilot. Okara launched an agent that can launch and scale influencer campaigns for any promotion in minutes (2M views). Try it here.
Codex learns by watching. OpenAI's Codex can now watch and learn from what you do on screen. Show it a workflow once and it saves it as a repeatable skill. The implications for anyone who runs the same multi-step process repeatedly are significant.
Claude Code artifacts. Claude Code now lets you preview in-progress work as a live web page via a new artifacts feature (1M views). Sharing and reviewing what Claude is building just became considerably easier.
Prompt Station
This prompt reverse-engineers your communication style from samples you provide, then delivers a full Voice Card: your sentence patterns, tone, rhythm, persuasion techniques, quirks, and a ready-to-paste prompt that reproduces your voice in any future writing task. It also generates three examples in your voice and gives you a confidence score so you know exactly how well it has modeled you. Run it in Claude or ChatGPT.
Clone your writing voice into a reusable AI prompt
Analyze the following writing samples and reverse-engineer my communication style. Study the writing at a deep level and identify the patterns that make it feel uniquely mine. Examine sentence structure, pacing, rhythm, vocabulary, tone, personality, formatting preferences, persuasion techniques, storytelling patterns, emotional triggers, transitions, and recurring stylistic habits. Then create a comprehensive Voice Card that another writer or AI could use to consistently reproduce my writing style. The Voice Card should include: a concise summary of the voice and what makes it distinctive; a breakdown of sentence structures, rhythm, pacing, vocabulary choices, tone, persuasion methods, formatting tendencies, and recurring writing patterns; a list of specific voice rules that define how the writing is constructed; a detailed Do's and Don'ts section; examples showing "sounds like me" versus "doesn't sound like me"; a list of signature phrases, hooks, frameworks, transitions, and stylistic quirks that appear repeatedly; and a copy-paste prompt that can be used to recreate this voice in future writing tasks. Finally, write three original examples in my exact voice: a LinkedIn post opener, a tweet/X post, and a short paragraph explaining a business concept. End with a confidence score (0-100%) for how accurately the voice has been modeled, explain the reasoning behind that score, and identify what additional writing samples would most improve accuracy. Writing Samples: [SAMPLES]Replace [SAMPLES] with three to five pieces of your own writing. Paste in newsletter sections, LinkedIn posts, past emails, or any text that sounds like you at your best. The more variety you give it, the more accurate the Voice Card. For a sharper output, include at least one longer piece (200 words or more) alongside shorter samples so the model can study both your structure and your instincts.
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.