AI isn’t confusing because it’s new. It’s confusing because every week another model shouts about being “the most advanced.” After watching Jeff Su break down what tools actually do best, I’m convinced the only strategy that works is simple: pick tools for their superpower, not their sizzle. My view is clear. Stop chasing the loudest launch and start matching tools to jobs.
My Take on Jeff Su’s Message
Jeff Su cuts through the noise by mapping each tool to a clear edge. That approach works. It respects time and lowers risk. He shows that the right question isn’t “Which model is smartest?” It’s “What does this tool do better than the rest?”
Native integration beats raw IQ. Jeff’s example inside Google Workspace is the strongest proof. He shows how Gemini can pull emails, docs, and calendar notes into one query and draft debriefs from it. That’s impact, not hype.
“Gemini is the only AI that can search and synthesize across your entire Google Workspace… from emails to docs to calendar invites in a single query.”
Action inside your system is a cheat code. Notion AI doesn’t just draft copy. It creates new pages, populates fields, and moves linked items. That’s real agent behavior inside a real workflow.
“If you need AI that can build and edit inside your workspace and not just search it, Notion AI is the only option so far that actually takes action.”
Voice removes friction. With Whisper Flow, Jeff shows how a 30-second brain dump beats a trimmed text prompt. More context means better results. He flags two caveats: the weak iPhone flow and the risk that a larger platform can ship similar features fast.
“The transcription… is not perfect, but it’s 95% there… it makes a material impact on how we interact with AI.”
Creative Work: Control, Precision, and Memory
Creative teams need different edges. Jeff sorts them cleanly.
- MidJourney: maximum control for power users through syntax and parameters.
- Google’s Nano Banana Pro: precise text edits and easy iteration without starting over.
- OpenAI’s image model: better consistency across a sequence of images.
- Google Flow: generate images and animate between them inside one app.
These edges matter in practice. If your brand mascot must look the same in 15 scenes, Jeff shows OpenAI’s image model keeps hair streaks and textures steady while others drift.
“If you need multiple related images where characters or visual elements need to stay consistent, ChatGPT is the more reliable choice.”
For presentation-ready graphics, he proves Nano Banana Pro can tweak fonts, colors, and layout with plain English and keep building on prior outputs instead of restarting from scratch. And for motion, Flow turns two static frames into a smooth transition without extra tools.
What This Means for Leaders and Teams
I build systems that turn customers into superfans. The lesson I’m taking from Jeff is blunt. Tool choice should serve your repeatable play, not your curiosity. If your work lives inside Gmail, Docs, and Calendar, start with Gemini because it knows your files. If your team runs on Notion, use Notion AI to act inside databases and templates. If you make a lot of decks or training content, pick your creative model based on the output constraint that matters most: control, precision, or consistency.
There’s also a risk check here. Jeff warns that third-party features can vanish the moment big platforms ship something similar. That’s not a reason to wait. It’s a reason to build portable processes and keep content stored in tools your team already uses.
How I’d Operationalize This Now
Here’s a simple playbook you can run this week to turn noise into results.
- Map your top five repeat tasks to a single tool’s superpower.
- Standardize two prompts per task: a short text version and a voice version.
- Document where final outputs live and who approves them.
- Assign a “tool owner” for each superpower to track updates.
- Revisit once a quarter; keep the workflow, swap tools only if the edge moves.
This isn’t about being first. It’s about being clear. Customers don’t care which model you used. They care that your email was accurate, your deck was sharp, and your imagery felt on-brand across every touchpoint.
The Bottom Line
Jeff Su is right: stop trying to crown one model. Win by picking the right superpower for the job and turning it into a repeatable system. Start with where your work already lives. Choose one tool per job. Make your prompts easy to teach and easy to audit. Your team will move faster, and your brand will look smarter.
Act this week. Audit your stack. Pick three superpowers. Train one workflow per superpower. Then ship. Fans are made in the moments you deliver with speed and consistency.
