Which AI niche looks hottest for the next 12 months?

By sacrificing the Sora video generator and walking away from a billion-dollar Disney deal to free up compute, OpenAI is betting everything on a massive new model codenamed "Spud." In this video, @Julian Goldie breaks down: 🔥 The "Spud" reveal and why Sam Altman says it will accelerate the entire economy ⚡ Why OpenAI KILLED Sora and the Disney partnership to save GPU power 🧠 The rebuilt audio architecture with sub-800ms latency for natural human conversation 🚀 The shift to "AGI Deployment"—what it means when OpenAI renames their entire product team 💡 Native Multimodality and how "Spud" understands text, audio, and images simultaneously Whether you’re a builder, a content creator, or an AI entrepreneur, understanding this massive shift in OpenAI’s strategy is a total game-changer for your workflow. P.S. If you want to go deeper than just tutorials and actually build a business with AI tools, then join the AI Profit Boardroom. This is where you’ll find the 30-day roadmaps, daily accountability, and advanced automation workflows that don’t make it to YouTube. When "Spud" finally drops, we’ll be running dedicated sessions in the Boardroom to show you exactly how to build it into your business fast. Check out the AI Profit Boardroom here: https://clinthermanlikes.com/aipb

The full stack, the breakdowns, and the 3 things I wish someone had told me before I started 👇 Tuesday, 9 PM. I open the terminal to set up Hermes Agent (an open-source agentic framework by Nous Research). By 3 AM, I had: → Spent 2 hours just fixing my broken terminal config before I could even start (broken $PATH, no brew, basic Unix commands missing — the works) → Watched Hermes take 10+ minutes to respond to a single "who are you?" — and then time out → My ANTHROPIC_API_KEY was sitting unused in .env while Hermes kept silently calling a different account I'd already burned through→ A growing suspicion that "agentic AI is here" was a marketing line I closed the laptop. Slept badly. Thursday, 7 PM. Here's what's running for BVMBY SEO without my hands on the keyboard: 🇩🇪 German writer #1 — 1 article/day at 5 AM UTC 🇫🇷 French writer — 1 article/day at 6 AM 🇩🇪 German writer #2 — 1 article/day at 7 AM 🔍 Keyword researcher (FR+DE) — every Monday 8 AM 📊 GEO citation monitor — every Wednesday 9 AM (Perplexity API) ✉️ Cold email drafter — Mon→Thu, 5 drafts/day for manual review 📈 Weekly digest — every Monday 8 AM, dropped in my inbox Total: 7 agents · ~$22/month in API · roughly 5 hours/week of my time saved. The 3 things that unblocked me 1. Don't run agents on a "local" model that isn't built for agents. Hermes Agent loads 89 skills + 28 tools into the system prompt. The default local model (hermes3:8b) physically can't process that — it spends 10+ minutes thinking about a single "hi". Switch the runtime to Claude / GPT / Gemini. "100% local" stays a fantasy until inference gets cheaper. 2. Kill the OAuth cache. Hermes caches Claude Code OAuth in ~/.hermes/auth.json and silently overrides your API key from .env. You'll watch your Claude Max quota evaporate while wondering why your API key budget isn't moving. Just delete the file. 3. Don't pick the newest Claude model.claude-sonnet-4-6 has 1M context and requires a paid "long context" tier opt-in (which, surprise, you don't have on a fresh API account). Switch to claude-sonnet-4-5-20250929 (200K context). Same quality, no opt-in.