🧠 Agents, Bosses, and Budgets: The Catch-Up Has Begun

 🔥🔥🔥 Three Highly Curated AI Updates

1. Microsoft says AI agents will reshape how companies work

Your next direct report might not have a pulse. Microsoft's Bryan Goode just made the clearest case yet for why org charts are about to be replaced by something he calls "work charts" — and the shift is already happening at companies paying attention.

The idea is simple: instead of hiring people to handle every task, professionals will increasingly orchestrate agents — AI systems that handle specific workflows autonomously.

For instance, a marketing manager doesn't just write campaigns; she manages the agent that drafts them, the agent that schedules them, and the agent that reports on them. Her job becomes directing and judging, not executing. Which means most professionals become what Bryan calls “agent bosses.”

Earlier this year, Kushank sat down for a conversation with Bryan Goode, where they break down what agents look like in practice, how adoption happens top-down and bottom-up, and why unlearning old ways of working is the real challenge. This one is a must watch if you lead people, run a business, or just want to understand what the next two years of work actually look like before they arrive.

2. GitHub Copilot moves to usage-based pricing

There are no free lunches, people. GitHub just confirmed that starting June 1st, Copilot will shift from flat-rate pricing to a usage-based billing model. This comes as Copilot evolves beyond simple autocomplete into more agent-like workflows that require significantly more compute.

In early May, users will start seeing a preview bill experience. This gives visibility into projected costs before the pricing change actually hits.

This matters because the “all-you-can-eat AI” era is starting to crack. Flat pricing worked when usage was predictable. It breaks when people start building full workflows, agents, and apps on top of these tools. Expect more tools to follow this exact path.

And if you’re using Copilot at work, log in and check the preview billing experience in early May before June 1st surprises you

3. Obsessed with ChatGPT image prompts? Yeah, same here

This prompt caught our attention because it basically turns ChatGPT into a magazine designer.

The idea is simple: You take a normal photo and layer it with handwritten-style annotations, magazine headlines, sticky-note elements, and structured layout rules. The output looks like a mix of editorial design and casual social content.

What’s interesting here is the level of control. The prompt dictates everything. Title placement, margins around the subject, handwritten texture, layout flow, even where decorations are allowed. You’re not just generating an image. You’re directing composition.

And the results are kind of wild. With the right prompt, you can get SNS-style visuals that look like they were designed manually, not generated.

🔥🔥 Two AI Tips Worth Trying

1. 🧠 AI being used for what it’s actually meant to do

There’s a viral story going around about a 14-year-old who learned Claude Code from YouTube and built a multiplayer math game.

It’s amazing that vibe coding lets you describe what you want instead of building everything from scratch. The AI handles most of the heavy lifting. You step in to guide, fix, and refine. This is best for people who understand what they want to build but don’t want to get stuck writing every line themselves. It shifts your role from “writer of code” to “director of outcomes.”

2. 🎨 Prompting like a creative director

There’s a new wave of prompts going around where people are generating full luxury ad campaigns. Think Ralph Lauren-style editorials or surreal Crocs campaigns that look like they came straight out of a magazine:

The interesting part isn’t the visuals. It’s how detailed the prompts are. They control posture, lighting, typography, and even brand-style composition. So you’re not just generating images. You’re directing a full creative shoot and it comes ready with text embedded.

🔥 One Thing You Didn’t Know AI Can Do

How close are we to real-world, campaign-ready quality?

People are using Seedance 2.0 to generate hyper-realistic ad campaigns that look like they came straight out of a papparazzi or a brand shoot.

The reason it works is simple. You start with a clear aesthetic or persona. Then you write a detailed prompt that covers the subject, setting, lighting, styling, and composition. Seedance handles the rendering, and you refine by iterating on small details.

What stands out is the realism. The lighting, textures, and proportions feel much closer to actual photography than most tools right now.

This trend is still early. We’re actively looking for the exact prompt formats behind the best results and will share them once we find something consistent.