Founder @ ChiefAIO | Helping Companies Automate & Streamline Operations
Fort Lauderdale, FL
Joined January 2014
- Tweets1.5K
- Following605
- Followers1.4K
- Likes1.5K
Most AI bills aren't a model problem. They're a nobody-checked problem.
About the same tasks solved, 2.8x the cost per solved task. Most small teams would never catch that.
Pick your 3 most common jobs and check cost per finished task monthly.
What's yours costing?
Keep Opus 5.5 as your coding default in Claude Code. On 52 coding tasks in Claude Code 2.1.322, Fable 5.2 solved about as many (82.7% vs 84.6%) but cost 2.80x as much per solved task.
Opus already solves most of these tasks, though, so telling the two apart will take harder ones.
github.com/garrytan/g-harnes…
4
Small shops don't need a cyber budget to start. They need a list.
Someone on your team's probably pasting customer info into an AI tool you never approved.
Ask everyone what they use. Cut what no one owns.
You can't protect what you don't know you're running.
What's on yours?
The FAA nationwide ground stop reminds me that the best business to be in right now is cybersecurity
15
Shipping more isn't the flex now.
67 things shipped, 4 moved a number. Two winning tests sat for 2 weeks because nobody made the call.
Before you ship, write the number it should move and a check date.
Busy isn't progress. It just feels like it.
What's waiting on your call?
I have a confession to make.
I've been building like a maniac with Claude. 738 pull requests since September. 😵
I was so proud of myself. 😁
So I asked Claude to grade it. Which of these actually significantly improved the business?
4.
Out of 67 things we actually shipped, 4 clearly moved a number.
Only 8% of it was even for AppSumo. The business that pays for all of it.
Then it went through my messages. 2,045 of them. I asked "did that work?" 25 times.
And the worst part. We ran 2 tests on AppSumo that found winners. They've been sitting there for 2 weeks because nobody made the call. Nobody = me. 🤦
Building used to be the hard part. Now it's basically free. So I was doing whatever felt productive instead of what moved things forward.
New rule: nothing ships unless I know what number it's supposed to move.
Prompt I used 👇
"Look at everything I shipped since [date] and every message I sent you. For each change: what number was it supposed to move, did it move, and what did I never check? Rank what I should stop, keep and finish."
19
After Andre came home, his family started asking a hard question: "If he can change and he can become successful, what's our excuse?"
@andrenorman on what one comeback does to everyone around you.
Session 2 of The Comeback Method: comebackmethod.org/climb
1
21
Robert Kanaat retweeted
Give Grok @Bot your most laborious tasks!
I asked @bot the best way to move 2000 photos in @Dropbox from Camera Uploads to Folders. She said, “would you like me to do it for you?” I was surprised as I had booked out the week to start sorting through them, a painful task. She did it in five minutes. I went to my Folders and saw they were there. Now I can spend my time scrolling on @X 🤓🤓🤣🤣
The build is the easy part now.
The hard part is the call that comes in at 7 PM and nobody picks up. Whoever owns that handoff owns the niche.
Pick one job. Answer, book, hand it to a human.
Owners don't buy stacks. They buy fewer missed calls.
Which niche would you pick?
One of the best businesses to build right now: a Grok Bot for one industry.
Pick a niche like dentists and build the AI agents for them so they don’t have to figure any of it out.
The stack:
- T3 Code + Opus 5.5/Fable 5.2 to build it
- Paper to design every screen first
- Orgo to give each agent its own computer
- AgentMail for an inbox
- AgentPhone for a phone number
- Composio for every connector
- CleanShot to screenshot bugs and say “fix this”
Dentists fix teeth and you set up the agents.
Really good way to learn how to become AI native while getting paid for it too.
2
2
75
Same goes for the software nobody uses.
The tool your team quietly works around isn't harmless. It's teaching everyone the real system is the spreadsheet.
Ask your team which tool they'd cut tomorrow. Then cut it.
Dead tools still cost you. Every month.
What's yours?
Bizarre truth: I’ve never met a business owner who regretted firing an underperformer too soon. Only too late.
16
A spending cap your agent can override isn't a cap. It's a suggestion.
The fix isn't a tighter number. It's deciding who gets pinged when it's crossed.
Over budget? It pauses and texts a person. A two-minute yes beats a surprise bill.
Who gets that text at your company?
Sh*t Our Agents Do:
We set a hard $100/day cap on AI spend.
An agent blew right through it to fix a P0.
It was probably the right call, so the cap turned out to be more of a suggestion.
16
Today's Microsoft outage is a free fire drill.
If your email's down all afternoon, how do customers reach you?
Write the backup down. A forwarding phone line, a text number, a pinned note on your site.
One tool's handy. It's also one point of failure.
What's your plan B?
Robert Kanaat retweeted
SpaceX community benefits
SpaceX builds the infrastructure of the future. And we’re committed to making sure the communities where we build share in the benefits → spacex.com/updates#building-…
Your AI doesn't need more freedom. It needs a sign-off step.
SaaStr's agents wanted to email every past speaker. An approval rule stopped it.
Pick the one thing your AI could embarrass you with. Make it ask first.
Speed's cheap. Trust isn't.
What's yours allowed to do alone?
Our agents wanted to email every SaaStr Annual speaker ever. That's why nothing happens without approval.
"They have to come to me before they take any action. They surface what they want to do, I approve it, then they can go do it."
Amelia Lerutte, on this week's The Agents.
63
Robert Kanaat retweeted
The worst thing you can do is aim low and hit.
Because now you're stuck.
Most people who are stuck want a miracle.
They want a lottery ticket.
They don't want a hard hat and some boots.
What are you aiming at?
Your team isn't slow. Your handoffs are.
At AppSumo, 78% of ideas now ship with the person who posted them, up from 8% a year ago. Shipped ideas went from 24 to 74.
This week, let whoever spots a fix own it start to finish.
Fewer handoffs, more done.
Who owns your ideas?
October 2025 - 24
October 2026 - 74
Ideas that went from Slack to live at AppSumo. 📈
About 20 people. None of it was on the project list.
Everyone says AI isn't making teams more productive. Ayman on our team saw this chart and said one word: "Bullshit."
What changed:
1- One owner, start to finish.
78% of ideas now go from Slack thread to live with the same person. A year ago: 8%.
No more "oh, you should do this." Whoever posts the idea ships it.
2- About 1 day.
That's the typical time from Slack thread to live.
Ship it before the thread goes cold.
3- Fast reviews.
Two of our engineers cut the friction in code review and deploys, and kept things safe and stable.
Faster builders need faster reviewers.
4- No permission.
Nobody waits for a ticket or a planning meeting.
If you see it, ship it.
Steal the measurement. Paste this into Claude with Slack and GitHub connected:
"Find every idea posted in our Slack in the last 12 months. Match each one to the code change that shipped it. By month, show: ideas shipped, whether the person who posted it also shipped it, and days from thread to live. Show me where ideas die."
Props to the team. 3x in a year.
ALT AppSumo bar chart titled 'Slack ideas brought to life, each month', Jul 2025 to Oct 2026, stacked bars: yellow = same owner from Slack thread to main, gray = handed off. Labeled totals: Oct 2025 = 24, Oct 2026 = 74. Approximate unlabeled totals: Jul 9, Aug 21, Sep 20, Nov 14, Dec 6, Jan 10, Feb 6, Mar 14, Apr 22, May 18, Jun 9, Jul 17, Aug 27, Sep 68. The yellow share is tiny until Apr 2026, then dominates Aug to Oct 2026 (about 22, 50 and 58). Stat cards: 3x more Slack ideas live each month (Oct '26 vs Oct '25); 78% had the same owner from Slack to main (was 8%); ~1 day typical time from Slack thread to main.
1
24
Most CRE data dies the day the deal closes.
In MGC CRM, closed deals become comps on their own, and each suite runs its own lease clock with rent roll, WALT, and a 24-month rolloff.
Built inside a working Bellingham brokerage with @muljat.
Where does your deal data go?
1
1
32