AI isn’t stuck in the lab anymore. It’s building book covers, grading marketing, and shipping live sites in minutes. After watching Kipp Bodnar and Kieran Flanagan walk through Perplexity Computer, my take is simple: super agents are becoming the main way we’ll work—and most people will fail unless they focus.
The Big Shift: From Apps To Agents
Kipp and Kieran laid it out with refreshing clarity. The race isn’t about yet another tool. It’s about a single agent that plugs into your stack, learns your skills, and executes across apps. Email. YouTube. Analytics. Design. Research. These agents wire into it all and run real jobs end to end.
“Every single thing is converging on… a super agent. You give it tools and it has a bunch of skills.”
That’s the correct read. And it matches what I’ve seen across crypto, social, and online business: the edge goes to the operator who sets one agent up well, then iterates it into a weapon. Dabblers will drown in options.
Perplexity’s Play: Speed, Skills, And “Disposable Web”
The demo that hit me hardest wasn’t just the model talk. It was outcomes. Perplexity Computer scraped top business book covers, analyzed patterns, wrote a reusable skill, and spit out polished mockups—then launched a clean website to share them. That’s not a party trick. That’s a new workflow.
“It built a website… so I could just send you the link.”
I’ve preached this for years: disposability drives speed. Spin up, test, toss, repeat. Perplexity’s model-agnostic routing and parallel sub-agents make that loop fast enough to change how teams ship.
“Let me launch four parallel sub-agents, each analyzing 25 book covers in depth.”
Yes, you could get more control in a local setup. But for most teams, ease beats precision—if the output is good enough to ship.
One Big Problem: The AI Clutter Trap
Kieran voiced the headache everyone feels.
“There’s too many super agents with tools and skills. I don’t know which one to use.”
That’s the trap. People build dozens of half-baked skills, share them in Slack, and never use them again. The dopamine hit wins. Work doesn’t get better.
“AI clutter is the new thing.”
My advice as a builder: pick one agent and go deep for 90 days. Make it your default for research, copy, QA, and analysis. If you switch weekly, you’ll never form the habits that create leverage.
What To Actually Do This Month
Here’s a simple system I use and teach. It works for founders, PMMs, and growth teams.
- Choose one agent: Perplexity Computer, Claude Co-Work, or Manna.
- Record a single workflow with a Loom. Narrate every choice.
- Turn that into a skill file. Run it. Track the number of edits to final.
- Iterate until changes drop under 5%. Then lock it and teach the team.
- Move to the next workflow. One at a time. No hoarding skills.
This sequence keeps you shipping while you learn. It also gives you real metrics. If edits aren’t dropping, your skill isn’t ready.
Why Marketing Wins Here
Kipp and Kieran nailed the bigger point: taste is back. With access to the same models and connectors, your advantage is judgment—what to ask, what to ship, what to discard.
“Never been a better time to be a creative person with taste.”
OpenClaw’s lobster vibe shows how brand can punch above tech. The tool isn’t easier; the story is stronger. That’s a reminder to product folks: your agent needs a point of view, not just features.
Receipts That Matter
The best use cases weren’t fluff. They were practical and money-adjacent:
- Live sites pulling PolyMarket data for real-time political or sports signals.
- Reverse engineering growth of under-the-radar B2B players.
- Auditing product pages with a custom scoring model and concrete fixes.
- Comp prep that sources pay bands, RSUs, and drafts a smart counter.
That’s work people usually burn weekends on. Now it’s an afternoon.
Costs, Trade-offs, And A Hard Truth
Perplexity Computer isn’t cheap at $200 a month and it’s desktop-only for now. Still, if you run a team or ship revenue-driving work, one win pays for itself. The bigger trade-off is depth versus speed. If you need full control, go local. If you need outcomes fast, start here.
My Bottom Line
As someone who’s built across crypto, media, and SaaS, my bet is firm: super agents will replace a chunk of “click work,” and the winners will be the few who master one stack and grind their skills to 95%+ quality.
“Do one workflow at a time… default to the AI way of doing it.”
Start now. Pick your agent. Capture one workflow. Build the skill. Measure edits. Iterate to near-zero changes. Then repeat. That’s how you turn demos into revenue, not clutter.
Call to action: This week, record your highest-impact task and turn it into a reusable skill. Next week, run it end to end and force yourself to ship with it. Momentum beats tinkering.
